G. Schneider, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, 171 77 Stockholm, Sweden Fax: +46 8327626 Tel: +46 852487675 E-mail: firstname.lastname@example.org
The glycosyltransferase SnogD from Streptomyces nogalater transfers a nogalamine moiety to the metabolic intermediate 3′,4′-demethoxynogalose-1-hydroxynogalamycinone during the final steps of biosynthesis of the aromatic polyketide nogalamycin. The crystal structure of recombinant SnogD, as an apo-enzyme and with a bound nucleotide, 2-deoxyuridine-5′-diphosphate, was determined to 2.6 Å resolution. Reductive methylation of SnogD was crucial for reproducible preparation of diffraction quality crystals due to creation of an additional intermolecular salt bridge between methylated lysine residue Lys384 and Glu374* from an adjacent molecule in the crystal lattice. SnogD is a dimer both in solution and in the crystal, and the enzyme subunit displays a fold characteristic of the GT-B family of glycosyltransferases. Binding of the nucleotide is associated with rearrangement of two active-site loops. Site-directed mutagenesis shows that two active-site histidine residues, His25 and His301, are critical for the glycosyltransferase activities of SnogD both in vivo and in vitro. The crystal structures and the functional data are consistent with a role for His301 in binding of the diphosphate group of the sugar donor substrate, and a function of His25 as a catalytic base in the glycosyl transfer reaction.
Database The atomic coordinates and structure factors have been deposited with the RCSB Protein Data Bank under accession numbers 4AMB, 4AMG and 4AN4
Anthracylines, produced mainly by soil dwelling Gram-positive bacteria of the genus Streptomyces, are aromatic polyketides that often exhibit high cytostatic potency. Several members of this group, for example epirubicin and doxorubicin, are among the most used anti-cancer agents worldwide . Anthracycline biosynthesis in Streptomyces starts with production of the polyketide scaffold, which is formed by iterative additions of malonate extender units to acetate or propionate primers by type II polyketide synthase in concert with cyclases, aromatases and ketoreductases . The aglycone backbone is further modified by tailoring enzymes including hydroxylases, methylases and aminotransferases, leading to the large diversity of metabolites. An important class of tailoring enzymes are the glycosyltransferases, as glycosylation of the carbon skeleton is required in many cases to form biologically active polyketides [3,4]. These glycosyltransferases use nucleotide diphosphate sugars as donors, typically TDP-conjugated sugars, with significantly greater diversity amongst the sugar acceptors .
Streptomyces nogalater produces the aromatic polyketide nogalamycin (1) (Fig. 1), which contains the carbohydrate moieties nogalose attached to carbon 7 and nogalamine attached to carbon 1 . The attachment of nogalamine to the aglycone is remarkable as it involves formation of a carbon–carbon bond linking atoms C5′′ of the sugar and C2 of the polyketide scaffold. A second link is formed via an O-glycosidic bond to carbon 1 of nogalamine. Attachment of sugar moieties to natural products via C–C bonds has been described in a few cases, for instance urdamycin , gilvocarcin , hedamycin  and granaticin , but the combination of both types of linkages in nogalamycin is rather unique and poses interesting questions regarding the order of the attachment and the chemistry of this glycosylation step.
The gene cluster responsible for the biosynthesis of nogalamycin in S. nogalater has been cloned, and the functions of most of the encoded enzymes have been annotated by sequence comparisons and/or subsequent biochemical and structural studies [11–15]. The cluster contains three genes that encode glycosyltransferases –snogD, snogE and snogZ, which are related to each other by amino acid sequence identities of ∼ 30%. A recent study showed that SnogE is responsible for transfer of nogalose to the C7 hydroxyl group of nogalamycinone (2) (Fig. 1), whereas SnogD is involved in transfer of the nogalamine moiety to the glycosylated metabolic intermediate 3′,4′-demethoxynogalose-1-hydroxynogalamycinone (3) (Fig. 1) . The glycosyltransferase SnogZ appears to be redundant as it is not necessary for heterologous formation of double glycosylated anthracyclines in the expression host Streptomyces albus.
Sequence comparisons of SnogD (GenBank accession number AAF01811.1, EC 2.4.1, UniProt ID Q9RN61) with related enzymes show that it belongs to the glycosyltransferase type 1 family . These enzymes display the GT-B fold , and are characterized by a SN2-type mechanism of sugar transfer with accompanying inversion of the anomeric configuration at C1 of the donor sugar. Here we present the X-ray crystal structure of SnogD in the apo-form and in complex with the nucleotide 2′-deoxyuridine-5′-diphosphate (dUDP). We further report the results of in vitro and in vivo mutagenesis studies of the conserved active-site residues His25 and His301. The structural and functional data are consistent with crucial roles for these residues in the mechanism of SnogD as a catalytic base and in nucleotide–sugar binding, respectively.
Structure determination of SnogD
The crystal structure of recombinant SnogD, comprising residues 13–390 of the native enzyme, was determined in two space groups to resolutions of 2.6 and 2.7 Å (Table 1). Crystals of wild-type SnogD (SnogD-wtdUDP) of sufficient diffraction quality (< 3 Å resolution) could not be reproduced despite numerous attempts. Reductive methylation of the protein sample proved critical for reproducibility of crystallization and the diffraction quality of the crystal. Chemical modification of SnogD resulted in a mass of 42 023.5 Da determined by electrospray mass spectrometry, which compares well to the theoretical mass (42 024.0 Da) of the SnogD construct including eight methyl groups. This data thus suggest almost complete dimethylation of the N-terminal nitrogen atom and the four lysine residues.
Table 1. Statistics on data collection and refinement. Numbers in parentheses are for the highest-resolution shell.
a At the European Synchrotron Radiation Facility.
Cell axes (Å)
64.9, 171.0, 69.4
64.6, 70.1, 176.0
66.7, 179.8, 70.2
Cell angles (degrees)
90.0, 90.0, 90.0
90.0, 91.7, 90.0
90.0, 90.0, 90.0
Number of reflections
69 886 (6703)
192 976 (21 392)
77 517 (6780)
Number of unique reflections
23 310 (2915)
42 845 (5800)
26 051 (3158)
Wilson B-factor (Å2)
Number of non-hydrogen atoms/mean B-factor (Å2)
R.m.s.d. bond lengths (Å)
R.m.s.d. bond angles (degrees)
Residues in favoured regions (%)
Residues in allowed regions (%)
Residues in disallowed regions (%)
The structure of the enzyme was determined in three forms: wild-type SnogD in space group P21212 (SnogD-wtdUDP) and methylated SnogD in space groups P21212 and P2 (SnogD-mdUDP and SnogD-m). The orthorhombic space group contains two chains in the asymmetric unit (51% solvent content). The refined models comprise all residues of the construct (residues 13–390), with a few exceptions. Weak or no electron density was observed for two loop regions in most of the chains of all three reported structures, which were thus modelled to various extents. These peptide segments are in the vicinity of the acceptor substrate binding site of the enzyme, and include residues 74–93 (loop FL1) and 323–327 (loop FL2). The asymmetric unit of the monoclinic crystal form contains four polypeptide chains, corresponding to a solvent content of 40%. In addition to the above-mentioned two loops, several other loop regions are disordered in at least one of the subunits of the corresponding model. These include residues 152–156, 196–199, 235–244, 275–276 and 339–350.
During refinement, residual difference electron density at the nucleotide binding site of chain B in SnogD-wtdUDP and chains B and D of SnogD-mdUDP indicated ligand binding. The shape of this electron density and the hydrogen bonding interactions with the enzyme were consistent with that of a diphosphorylated nucleoside lacking the 2′-hydroxyl group, and was subsequently modelled as dUDP (Fig. 2). The ligand originates from the expression host as it was not added during crystallization. Details of the refinement and the model quality are given in Table 1.
Overall structure of SnogD
Structure of the subunit
SnogD displays the twin-domain architecture of the GT-B fold, consisting of two Rossman-fold domains with the active site located at the domain interface (Fig. 3). The N-terminal domain (residues 1–209) consists of a seven-stranded parallel β-sheet, eight α-helices and two 310 helices. The sheet itself is sandwiched between four of the helices on one side and three helices on the other side. The C-terminal domain (residues 228–390) is of similar topology, with a six-stranded parallel β-sheet flanked by six α-helices and four 310 helices. The last C-terminal helix, with a kink between residues Glu374 and Pro377, a common feature of the GT-B fold, crosses over to complete the N-terminal domain through residues Pro378–Gly390.
Structural relationships to related proteins
A search for structural homologs of SnogD using the DALI server  resulted in several GT-1 glycosyltransferases with significant structural similarity despite low sequence homology. The closest structural relatives are CalG3  (Z-score 42.5, rmsd 2.3 Å, 38% sequence identity), UrdGT2  (Z-score 37.4, rmsd 2.6 Å, 28% sequence identity), SpnG  (Z-score 37.4, rmsd 2.6 Å, 30% sequence identity), CalG1  (Z-score 37.1, rmsd 3.5 Å, 30% sequence identity), EryCIII  (Z-score 35.8, rmsd 2.9 Å, 34% sequence identity) and OleI  (Z-score 33.2, rmsd 2.7 Å, 25% sequence identity).
A pBLAST search returned the amino acid sequence of the hypothetical protein ML5_4439 (Uniprot ID E8RWV0) from Micromonospora sp. L5 as the closest relative, showing 63.1% amino acid sequence identity.
Analysis of the crystal packing by manual inspection and using the PISA server  revealed that SnogD is a dimer in both crystal forms. This agrees well with the results from analytical gel-filtration experiments that indicated a dimeric structure in solution. The subunits in the SnogD dimer are oriented head-to-tail (Fig. 3), and related by a twofold non-crystallographic symmetry axis. The dimer interface covers an area of 1430 Å2, and includes three salt bridges and 16 hydrogen bonds between the subunits. Part of the interface is an extension of the seven-stranded β-sheet of the N-terminal domain by an additional parallel β-strand comprising amino acids 215–217 from the long interdomain peptide segment (210–227) linking the N-terminal to the C-terminal domain of the neighboring subunit. Superimposition of the two subunits forming the dimer results in an rmsd of 0.5–0.8 Å. The overall structures of the dUPD-free and dUDP-bound subunits are very similar, irrespective of crystal from, and superposition results in rmsd values in the range 0.6–0.8 Å. Structural deviations are restricted to the nucleotide binding site and are described below.
Reductive methylation stabilizes crystal packing
Reductive methylation significantly improved the reproducibility in obtaining better-diffracting crystals. However, electron density corresponding to a methylated lysine side chain was only observed for one (Lys384) of the four lysine residues in chain A. This residue was well defined in electron density in the two crystal forms, but the side chains of other lysine residues (including Lys384 from chains B, C and D in the asymmetric unit) were disordered to various extends. Methylated Lys384 forms an intramolecular stacking interaction with the side chain of Trp204, and most importantly is also involved in a salt bridge to Glu374* from the polypeptide chain related by crystallographic symmetry. This salt bridge, only observed for chain A, is not present in crystals of wild-type SnogD, and the additional interaction in the crystal lattice is most likely involved in improved crystallization behavior upon reductive methylation.
Nucleotide binding to SnogD
dUDP is bound in a cleft at the interface between the two domains of the SnogD subunit. Hydrogen bonds to the ligand involve residues from the C-terminal domain and one residue of the interdomain linker (Fig. 4). The uracil moiety packs against the side chain of Trp285. Atoms O2 and N3 of the pyrimidine ring are hydrogen-bonded to the backbone amide of Leu288 and carbonyl oxygen of Ile286, respectively, and the O4 atom is in close proximity (3.4 Å) to the backbone amide of the latter residue. The 3′-hydroxyl group and the O5′ atom of the deoxyribose moiety form hydrogen bonds to the side chains of Asn212 and Thr306, respectively.
The α-phosphate is located at the N-terminal end of the helix (303–315) and is linked via hydrogen bonds to the amide groups of Ser304 and Gly305 (Fig. 3). Both residues are part of the conserved diphosphate-binding PPi motif  represented in SnogD by GGSG (Fig. S1). The highly conserved His301 is appropriately positioned to form hydrogen bonds to both α- and β-phosphate groups. A modest movement brings the β-phosphate group within hydrogen bonding distance to the main chain amide and side chain hydroxyl of Ser236. Such an interaction was observed in one of the nucleotide-binding subunits of SnogD-mdUDP.
In UDP-utilizing glycosyltransferases, recognition of the ribose moiety is provided by hydrogen bonds of a glutamic acid residue located in the α-helix directly after the PPi motif [24,27,28]. In SnogD, this glutamic acid is replaced by Thr309, which does not interact with the deoxyribose group. Instead the 3′-hydroxyl group of the carbohydrate moiety forms a hydrogen bond to the side chain of Asn212. The different mode of hydrogen bonding interactions of the deoxyribose group in SnogD compared to the UDP-donor specific glycosyltransferases results in a difference in conformation of parts of the interdomain linker (residues 210–227). These conformational differences may be described as a shift of this peptide segment closer to the nucleotide binding site (Fig. S2), a feature also recently observed in TDP-donor specific SpnG .
Another sequence motif, D/E-Q, which is frequently encountered in the nucleotide binding region of glycosyltransferases  and is involved in hydrogen bonds to the C2′′–C4′′ hydroxyl groups of the donor sugar is not conserved in SnogD (Fig. S1). However, within the FL2 loop, polar residues are present that may potentially form hydrogen bonding interactions to the 3′′-amino and 4′′-hydroxyl groups of the donor sugar nogalamine. This loop is disordered in several of the subunits of SnogD, most likely due to flexibility in the absence of the activated sugar donor.
In SnogD-wtdUDP and SnogD-mdUDP, nucleotide binding is observed in only one of the two subunits of the dimer. Comparison of nucleotide-free with nucleotide-bound subunits revealed that ligand binding is accompanied by rearrangement and/or stabilization of two nearby residue segments. One of these segments (n1) comprises part of the binding site for the α-phosphate (residues 235–237). In the nucleotide-free subunits, these residues are either disordered or adopt different conformations, and are located more distantly from the ligand binding site. Local structural differences extend to amino acids up to position 249, including unwinding of the 310 helix (residues 242–246) in nucleotide-free subunits A and C of SnogD-m (Fig. 5). In the structure of the binary complex, loop n2 (residues 266–268) is shifted away from the binding site to accommodate the pyrimidine ring of the bound nucleotide.
The binding of nucleotides to only one of the active sites in the dimer observed for SnogD-wtdUDP and SnogD-mdUDP may be directly linked to participation of segments 235–249 (n1) and 266–277 (n2) in crystal packing interactions in both crystal forms. It is thus likely that ‘half-occupied’ and ‘empty’ dimers are preferably selected for incorporation into the crystal lattice, as they interact more optimally with neighboring molecules. This may also explain the difficulties in obtaining crystals of SnogD with fully occupied binary complexes.
Model of SnogD with activated sugar donor and acceptor substrates
Based on the structures of the SnogD–dUDP binary complexes and the complex of a plant 3-O-glucosyltransferase with the sugar donor analog uridine-5′-diphosphate-2-deoxy-2-fluoro-α-d-glucose and the acceptor kaempferol , we modeled binding of an activated sugar donor to the active site of SnogD (Fig. 6). The position of the donor sugar nogalamine is restricted by the covalent link to the dinucleotide, the small pocket for the carbohydrate moiety present in SnogD, and the requirement for the C1′-hydroxyl to be in an axial orientation so that glycosyl transfer can occur.
In the absence of experimental data on SnogD–acceptor substrate complexes, the precise orientation of acceptor substrates in the binding pocket of SnogD is more difficult to define, due to the large size of the hydrophobic pocket and the lack of structural data for the FL1 loop. Acceptor substrate binding to SnogD probably involves hydrophobic residues lining the cleft between the domains and closure of the FL1 loop over the acceptor substrate as seen in CalG3 . The acceptor substrate 3 (Fig. 1) may be modeled in the predominantly hydrophobic substrate cleft in such a way that the C1 hydroxyl group of the acceptor is located suitably to act as nucleophile and attack the C1 atom of the activated nogalamine sugar (Fig. 6). In this model, His25 is close to the C1 hydroxyl group of the aglycon, which is the attacking nucleophile in the glycosyl transfer reaction.
In vitro characterization of active-site mutants of SnogD
Based on the modeled ternary complex of SnogD with the nucleotide sugar donor and acceptor substrates, we selected the highly conserved residues His25 and His301 (Fig. S1) as targets for site-directed mutagenesis. The mutations His25Ala, His25Asn and His301Ala were introduced, and the mutant proteins were purified according to the protocol used for wild-type SnogD. The CD spectra of mutants and wild-type enzymes were very similar, indicating that the mutants were correctly folded. The natural sugar donor TDP-nogalamine was not available, and the enzymatic reaction was therefore studied in the reverse direction using the assay described previously  (Fig. 7A). The in vitro assays with the mutants His25Ala, His25Asn and His301Ala showed a significant reduction in UDP-dependent deglycosylation of nogalamycin F (4) to 3 to 1.5–2.4% of the wild-type enzyme activity (Fig. 7B and Table 2).
Table 2. Relative in vitro activities of SnogD mutants.
Relative activity of triplicates (%)
Standard deviation (%)
In vivo studies of active-site mutants
In order to probe the glycosyl transfer reaction in the forward direction, the mutated genes were expressed in vivo in S. albus using a two-plasmid system. The polyketide acceptor 3 and the nucleotide diphosphate deoxysugar donor substrates were produced using cosmid pSnoΔgD, which contains the majority of the nogalamycin gene cluster and from which snogD has been deleted . The native and mutated snogD genes were expressed in the heterologous host using the high copy number vector pIJ486 from their own promoter sequences as described in Experimental procedures. The production profiles of the strains indicated that formation of the aminoglycosylated SnogD reaction product, nogalamycin R (5) (Fig. 8A), was impaired in the mutant strains (Fig. 8B), consistent with the in vitro activity measurements.
Previous studies on the glycosylation reactions in nogalamycin biosynthesis have shed some light on the order of these steps and the enzymes involved . Gene inactivation experiments have shown that glycosyl transfer of nogalose to the C7 hydroxyl group of nogalamycinone by SnogE precedes transfer of nogalamine to the C1 hydroxyl by SnogD. The studies further suggested that formation of the C1 O-glycosidic linkage precedes formation of the C–C bond between the C5′′ carbon of nogalamine and the C2 carbon of the polyketide scaffold. The nogalamine moiety has been shown to be involved in DNA–nogalamycin interactions , which are important for poisoning of human topoisomerases , one of the major biological activities of nogalamycin. Therefore, detailed understanding of these biosynthetic steps and the enzymes involved may facilitate engineering of novel anthracyclines with improved activities.
The structure analysis of SnogD provides a framework for the last glycosylation step, transfer of nogalamine to 3. The enzyme belongs to the type 1 branch of the GT-B fold family of glycosyltransferases. Fortuitously, the enzyme crystallized with bound dUDP derived from the expression host, allowing structural characterization of nucleotide binding to SnogD. Sugar donors for the glycosylation reactions in anthracycline biosynthesis in Streptomyces species are often formulated as activated TDP sugars, although biochemical evidence in the case of nogalamycin biosynthesis is lacking. SnogD can utilize both TDP and UDP in the in vitro assay, the deglycosylation of 4 . This is consistent with structural data, as the binding pocket may accommodate uracil and thymidin rings. However, SnogD lacks the glutamic acid residue that is characteristic of UDP-donor specific glycosyltransferases , and thus may not be able to discriminate efficiently between UDP and TDP donors.
As attempts to obtain crystals of ternary complexes of SnogD with acceptor substrates were not successful, the structure of SnogD with bound dUDP was used as the template to model the Michaelis complex of SnogD with UDP-nogalamine and 3. This model suggests the existence of one active-site residue, His25, close to the C1′′ atom of the activated sugar, that may be able to facilitate the enzymatic reaction. This residue may be involved in proton abstraction, thus enhancing the nucleophilicity of the acceptor substrate and facilitating the attack on the C1′′ of the donor sugar and subsequent glycosyl transfer. Replacement of His25 by alanine or asparagine resulted in mutants that are severely impaired in catalytic activity in vitro. Analysis of the production profiles in vivo using these SnogD mutants show the same trend, with a significant reduction in the amounts of 5 produced. These data demonstrate a critical role for this residue in the enzyme reaction, and are consistent with a function as catalytic base.
Another histidine residue, His301, which is close to the diphosphate group of the donor sugar, is also essential for catalysis, as substitution of this residue by alanine resulted in significant loss of enzyme activity in vitro. This residue is involved in binding the diphosphate group and may further stabilize the negative charge that develops at the β-phosphate of the leaving group . These two residues are conserved in many glycosyltransferases, and their counterparts have been implicated in the catalytic mechanisms of these enzymes [19,24,27–29,33].
Evidence for the function of SnogD as a glycosyltransferase involved in formation of the C1 O-glycosidic bond to nogalamine is now rather compelling. However, it is much less clear whether or not SnogD is directly involved in formation of the C–C linkage between C5′′ of nogalamine and the C2 atom of the polyketide. This reaction requires formally abstraction of hydrogen, and the structure of SnogD does not contain any catalytic groups and/or co-factors that might support such chemistry. This step in nogalamycin biosynthesis remains enigmatic, and most likely requires another enzyme, either working in concert with SnogD or acting after the glycosyl transfer reaction.
Cloning, expression and purification of recombinant SnogD
Cloning, expression and purification of recombinant SnogD were performed as described previously . In short, eight constructs were designed, all with an N-terminal His tag, amplified from genomic S. nogalater DNA by PCR, and cloned using ligation-independent cloning . The production of one construct, comprising residues 13–390, using Escherichia coli as the expression host, was performed as described previously . For production of SnogD used for subsequent methylation experiments, cells were grown in culture medium supplemented with a mixture of metal salts, with final concentrations of 40 μm CaCl2, 20 μm MnCl2, 20 μm ZnSO4, 4 μm CoCl2, 4 μm CuSO4, 4 μm Na2MoO4, 4 μm H3BO3, 4 μm NiSO4, 4 μm Na2SeO4 and 4 μm FeCl3. Recombinant SnogD was purified using affinity, anion-exchange and size-exclusion chromatography. The purity of the protein was monitored by SDS/PAGE, and the identity of the recombinant protein was confirmed by ESI-MS.
Residues for mutagenesis were selected based on the structure of SnogD and protein sequence alignment to structural homologs (Fig. S1). The mutations were introduced by PCR using the full-length expression plasmid (encoding residues 1–390 of SnogD cloned as described previously ) as template and Phusion DNA polymerase (Finnzymes, Espoo, Finland) to amplify the mutated DNA fragments. The primers used were His25Alaf (5′-GCCATCCTGCCGACGGTGCCGCTGGC-3′), His25Alar (5′-GGCGCTGAGCCCGGGTGAAGTGATGAACAACGCA-3′), His25Asn_f (5′-AACATCCTGCCGACGGTGCCGCTGGC-3′), His25Asn_r (5′-GTTGCTGAGCCCGGGTGAAGTGATGAACAACGCA-3′), His301Alaf (5′-GCCGGGGGCAGCGGCACACTGCTGACG-3′) and His301Alar (5′-GGCATGGATGATCGCGTCGCACGTCTCCAGCAG-3′), and vector-specific primers 5′-GAATAATTTTGTTTAACTTTAAGAAGGAGATATACATATGCACCATCA-3′ (forward) and 5′-TGCGGCCGCAAGCTTGTCGACGG-3′ (reverse). These fragments were subsequently used to generate the full dsDNA sequence of SnogD harboring the mutation, as well as vector-compatible restriction sites. After restriction enzyme digestion (NdeI and HindIII), ligation using T4 ligase (all enzymes from New England Biolabs, MA or Fermentas International Inc., Burlington, Canada) was performed to insert the dsDNA into the vector pET28-pNic BsaI (Structural Genomics Consortium, Oxford, UK). The sequences of the mutants were verified by DNA sequencing.
Reductive methylation of SnogD
Purified SnogD was diluted to a final concentration of 1 mg·mL−1 in a solution containing 45 mm Na3PO4, 0.3 m NaCl and 10% v/v glycerol pH 8.0. Reductive methylation was performed as described previously  with two exceptions: the formaldehyde solution was prepared freshly by dissolving paraformaldehyde in 50 mm Na3PO4 pH 8.0 and heating to 333 K over 4 h, and the methylation reaction was quenched by addition of 1 m glycine to a final concentration of 0.1 m. After filtration, methylated SnogD was purified using anion-exchange and size-exclusion chromatography. The extent of methylation of SnogD was determined by ESI-MS.
Crystallization of SnogD
Wild-type SnogD (SnogD-wtdUDP) was crystallized by hanging drop vapour diffusion against a reservoir containing 26% w/v poly(ethylene glycol) (PEG) 350 and 0.1 m citrate pH 5.2. Droplets of the protein solution (10 mg·mL−1 SnogD-wtdUDP in 50 mm Tris pH 7.4, 0.2 m NaCl) were mixed with equal volumes of the reservoir solution. Crystals belonging to space group P21212 appeared after 3 days equilibration at 293 K.
As diffraction quality crystals were difficult to reproduce using wild-type enzyme, screening for crystallization conditions was repeated with methylated SnogD, which, in combination with streak seeding, reproducibly yielded diffraction-quality crystals under two conditions: the SnogD-mdUDP and SnogD-m data were collected from crystals obtained by vapour diffusion against a reservoir of 15% w/v PEG 8000, 0.16 m calcium acetate, 0.08 m sodium cacodylate pH 6.8, and 16% w/v PEG 3350, 0.2 m MgCl2, 0.1 m BisTris pH 5.7, respectively.
Data collection and structure determination
SnogD-wtdUDP crystals were briefly transferred to a solution of 30% v/v glycerol, 28% w/v PEG 3350, 0.1 m Bis/Tris pH 5.3 prior to flash-freezing in liquid nitrogen. Crystallographic data were collected at 100 K and a wavelength of 0.934 Å at beamline ID14-EH1 of the European Synchrotron Radiation Facility (Grenoble, France), to a resolution of 2.6 Å. Crystals of methylated SnogD were cryo-protected using mineral oil. Data were collected from two crystals to 2.7 Å (SnogD-mdUDP) and 2.6 Å (SnogD-m) resolution, respectively, at a wavelength of 1.072 Å at beamline ID23-EH1 at the European Synchrotron Radiation Facility. Data processing was performed using the programs MOSFLM and SCALA from the CCP4 package . Statistics for the diffraction datasets are given in Table 1.
The SnogD-m crystal belonged to space group P21212, with similar unit cell dimension as the SnogD-wt crystal, and contained two enzyme subunits per asymmetric unit. In contrast, the SnogD-mdUDP crystal belongs to space group P2, with four molecules (two dimers) per asymmetric unit.
The crystal structure of SnogD-wtdUDP was determined by molecular replacement using PHASER  and an ensemble of monomer structures of five homologous enzymes (PDB accession codes 1F0K, 1NLM, 2IYA, 2P6P and 3D0Q) as the search model, of which CalG3 from Micromonospora echinospora shows the highest sequence identity to SnogD (36%). The structures of methylated SnogD were determined by molecular replacement using PHASER (SnogD-mdUDP) or MOLREP  (SnogD-m) and the refined ligand-free subunit of SnogD-wt as the search model.
Model building and refinement were performed using WinCOOT  and REFMAC5 , employing tight main chain non-crystallographic symmetry (NCS) restraints in the first and automatically determined local NCS restraints in the final cycles of refinement. Randomly selected reflections (5%) were used to monitor Rfree. Water molecules were added in WinCOOT. The presence of residual electron density indicated ligand binding to the active sites of one subunit per dimer for SnogD-wtdUDP and SnogD-mUDP. Based on the shape of the electron density and the hydrogen bonding pattern, the ligand was identified and modeled as dUDP, an analog of one of the products of the SnogD-catalyzed reaction.
The final model of SnogD-wtdUDP contains residues 13–389 of SnogD, one residue from the affinity tag for chain A and residues 13–390 for chain B, two residues from the affinity tag, a total of 115 water molecules and one dUDP molecule. Two loop regions, residues 76–84 and 325–327, are disordered to various extents in the two subunits. The refined model of SnogD-mdUDP contains four subunits. In this space group, several loop regions were disordered to various extents in the four subunits and were not included in the model. The structural model contains 48 water and two dUDP molecules. The final model of SnogD-m contains residues 13–389 (except for the following loop regions, which are disordered in the electron density maps: A79–A90, A239–A241, B78–82 and B324–327) and 67 water molecules. Refinement statistics are given in Table 1.
The atomic coordinates and crystallographic data have been deposited with the Protein Data Bank under accession numbers 4AMB (SnogD-wtdUDP), 4AMG (SnogD-m), and 4AN4 (SnogD-mdUDP).
Quaternary structures, crystal packing and interfaces were analyzed using the PISA server at the European Bioinformatics Institute . Database searches for structural homologs were performed using Dali . Sequence alignments were performed using Clustal W . Figures were prepared using PyMol .
In vitro enzyme assays
Full-length SnogD and point mutants were used for the activity assays. As activated TDP-nogalamine or UDP-nogalamine were not available as sugar donors, the reaction was monitored in the reverse direction, with UDP as acceptor and 4 as donor, as described previously . Briefly, 8.1 μg of SnogD was added to 40 μL of a mixture of 20 mm UDP as acceptor and 0.5 mm of 4 as donor substrate in 50 mm Tris, 200 mm NaCl, pH 7.4, and incubated in the dark at 295 K for 13h. The reactions were quenched by addition of chloroform, and the anthracyclines were extracted from the resulting mix of chloroform and water. The chloroform phases were analyzed by HPLC to detect glycosyl transferase activity, which was defined as the ratio between formed product 3 over remaining substrate 4, after correction for 3 present in minor amounts in the substrate solution.
In vivo activity assays
The mutated snogD genes were moved as NdeI–EcoRI fragments into digested and calf intestinal alkaline phosphatase (CIAP)-treated pBluescript SK(-) (Stratagene, La Jolla, CA, USA) harboring the PCR-amplified and sequenced snogD promoter region. PCR was performed using pSnogaori  as the template and the following primers with vector-compatible restriction sites (EcoRI/HindIII) and an added NdeI site downstream of the promoter region: FWgDp (5′-AATAAGCTTAGGTCTCCGGGCGGGTCA-3′) and REVgDp (5′-TTAGAATTCTTACATATGGACGGCGCCTTCTGTTGC-3′) (restriction sites underlined). The promoter and gene region were moved as an EcoRI–HindIII fragment (1.5 kb) into a digested and CIAP-treated pIJ486 vector . The plasmids were amplified in Streptomyces lividans TK24, verified by digestion analysis and moved to S. albus/pSnoΔgD by PEG-induced protoplast transformation . The wild-type SnogD plasmid was constructed as described previously .
For analysis of the activity of SnogD and mutated SnogD, the strains were cultivated in 250 mL Erlenmeyer flasks with 30 mL NoS-soyE1  supplemented with 50 μg·mL−1 thiostrepton (Calbiochem, San Diego, CA) and 50 μg·mL−1 apramycin (Sigma-Aldrich, St Louis, MO, USA) for 5 days at 303 K with vigorous shaking. The metabolites produced were collected by binding to 20 g·L−1 Amberlite® XAD-7 for 2 days (Rohm and Haas, Philadelphia, PA, USA), after which the medium was decanted and the XAD-7 was washed with water. The metabolites were extracted using MeOH and filtered through 0.2 μm polytetrafluoroethylene (PTFE) (VWR, Radnor, PA, USA) prior to analysis by HPLC (Shimadzu SCL-10Avp, Kyoto, Japan) using a reversed-phase column (SunFire™ C18; 3.5 μm, 2.1 × 150 mm, Waters, Milford, MA, USA) using a gradient from 15% acetonitrile in 0.1% formic acid to 100% acetonitrile.
We thank Dr Jarmo Niemi (Department of Biochemistry and Food Chemistry, University of Turku, Finland) for DNA encoding SnogD. We acknowledge access to synchrotron radiation at the European Synchrotron Radiation Facility (Grenoble, France) and the European Molecular Biology Laboratory Outstation, Deutsches Elektronen Synchrotron (Hamburg, Germany) and thank the beamline staff for helpful support. This work was supported by a grant from the Swedish Research Council and the Academy of Finland (grant number 136060).