From a cosmid library of Streptomyces cyanogenus S136, DNA fragments encompassing approximately 35 kb of the presumed landomycin biosynthetic gene cluster were identified and sequenced, revealing 32 open reading frames most of which could be assigned through data base comparison.
Streptomyces cyanogenus S136 produces landomycins, consisting of a benz[a]anthraquinone-type aglycone moiety and a varying phenol-glycosidically linked oligosaccharide chain . Landomycin A, the principal metabolite of S. cyanogenus, is the largest angucycline so far described, and the most active landomycin . Its deoxysugar moiety is an unusual hexasaccharide consisting of four d-olivose and two l-rhodinose residues (Fig. 1). Over the past decade, the gene clusters encoding many structurally related aromatic polyketide synthases (PKSs) have been cloned and sequenced, among them the clusters for the angucycline antibiotics urdamycin A and jadomycin B [3–5]. All these studies paved the way for the construction of hybrid synthases by mixing PKS subunit genes from various parent microorganisms [6–8]. Many aromatic polyketides are modified by subsequent enzymes of the pathway, such as oxygenases and glycosyltransferases. These post-PKS tailoring enzymes are often responsible for converting inactive precursors into biologically active metabolites . From investigations of the biosynthetic gene cluster for the landomycins, we expected to obtain interesting genes for combinatorial biosynthetic studies.
2Materials and methods
2.1Bacterial strains, plasmids and growth conditions
S. cyanogenus S136 (DSM 5087) was grown at 28°C on CRM medium for DNA isolation. Escherichia coli XLI-Blue MRF (Stratagene, La Jolla, CA, USA) was used for the preparation of the cosmid library and routine subcloning. pBluescript SK− (pSK−) was obtained from Stratagene and cosmid pOJ446 was a gift from Eli Lilly Co. (Indianapolis, IN, USA).
2.2DNA manipulation and southern hybridization
Chromosomal DNA from S. cyanogenus S136 and plasmid DNA from E. coli were prepared using standard protocols [10, 11]. A cosmid library was prepared as described . Restriction endonuclease digestions, alkaline phosphatase treatments, ligation, Southern hybridization and other DNA manipulations were performed according to standard procedures as specified by the manufacturers (Amersham, Braunschweig; Boehringer, Mannheim).
2.3DNA sequencing and computer assisted sequence analysis
DNA was sequenced with an automated laser fluorescence sequencer (Vistra, Molecular Dynamics) according to the supplier's instructions using double-stranded DNA as template purified with Nucleobond AX-20 columns. Computer-assisted sequence analysis was carried out using the DNASIS software package (version 2, 1995; Hitachi Software Engineering, San Bruno, CA). BlastX analyses  were run with the GenBank CDC translations+PDB+SwissProt+Spupdate+PIR, release 2.0. Open reading frames were identified by using the CODONPREFERENCE program .
3Results and discussion
3.1Screening of a cosmid library
It was likely that a dTDP-glucose 4,6-dehydratase, catalyzing the deoxygenation reaction at C-6 of TDP-glucose, is one of the first enzymes involved in the biosynthesis of TDP-olivose and TDP-rhodinose . Therefore approximately 2000 colonies containing recombinant cosmids were probed by colony hybridization using an internal fragment of a dNDP-glucose 4,6-dehydratase gene as a probe . Cosmid DNA from seven hybridizing clones was analyzed by restriction mapping and Southern blot experiments. These revealed that the cosmids contained overlapping DNA encompassing approximately 70 kb of the S. cyanogenus S136 genome (Fig. 2). One cosmid, H2-26, was used for further experiments.
3.2Sequence analysis of the putative landomycin biosynthetic gene cluster
CODONPREFERENCE analysis of 35-kb DNA revealed 30 open reading frames (ORFs). Two additional putative ORFs (lanU, lanZ6) were detected with a large number of codons with A or T in third position (Fig. 3, Table 1). The GenBank/EMBL/IDDBJ accession number for the nucleotide sequence is AF080235.
Table 1. Deduced functions of ORFs in the landomycin biosynthetic gene cluster
Size of the protein (aa)
Size of the protein (aa)
chain length factor
acyl carrier protein
3.3Characterization of the deduced amino acid sequences
The deduced amino acid sequences encoded by lanA, lanB and lanC strongly resemble gene products of type II iterative polyketide synthases. The closest resemblance was found to UrdA (LanA: 81% identity), UrdB (LanB: 72%) and UrdC (LanC: 63%) from S. fradiae involved in the biosynthesis of urdamycins  and to Orf1 (LanA: 82% identity), Orf2 (LanB: 72%) and Orf3 (LanC: 70%) involved in jadomycin production . The deduced LanD protein sequence has high similarity to ketoreductases from aromatic gene clusters that are involved in reduction of the C-9 keto group of the nascent polyketide chain. Again the highest resemblance was found to proteins of the urdamycin (UrdD, 85% identity) and jadomycin (ORF5, 86%) producers [3, 4]. Comparison of the putative products of lanN and lanV to proteins in the data bases suggests that both belong to the short-chain alcohol dehydrogenase family of proteins. The highest resemblance was found to an ORF upstream of jadR2 in the jadomycin cluster (LanN 40% identity; LanV 47%). Short-chain alcohol dehydrogenases, which are proteins of about 250 amino acids, exhibit a strictly conserved SX10–13YXXXK motif which is believed to form part of the active site. This motif is also present in LanN (SRGAVEPVEPKAAVYAMAK) and LanP (SSGLTRVASPDQVTYGMSK). The deduced protein products of two further genes (lanO and lanZ4) contain the motif VxVxPxxxxSxPxxxKxxxD which has been found in a NAD(P)H-dependent dehydrogenase from Bacillus subtilis (EMBL accession number: Z99121). The deduced amino acid sequence of lanP shows 53% identity to MmdA, a methylmalonyl-CoA decarboxylase from Veillonella parvula, and also a strong resemblance to several propionyl-CoA carboxylases from different organisms. All these proteins catalyze carboxyl transfer from methylmalonyl-CoA to protein-bound biotin to yield propionyl CoA and the carboxybiotin derivative, or catalyze carboxyl transfer in the reverse reaction. LanP might catalyze a decarboxylation step during the biosynthesis of the angucyclinone skeleton. The putative protein encoded by lanF is very similar to UrdF (73% identity)  and also similar to TcmI (38%) . TcmI is believed to be involved in cyclization of the fourth ring during tetracenomycin biosynthesis. The deduced amino acid sequence of lanL resembles proteins designed as aromatases in different organisms. The closest resemblance was to JadORf4 (76% identity) , believed to catalyze aromatization of the first ring of jadomycin and actinorhodin respectively. The deduced amino acid sequences encoded by lanE and lanM strongly resemble proteins believed to be involved in oxygenation steps. Again, proteins from the urdamycin  and jadomycin  producers are most similar. Because the deduced amino acid sequence of lanZ5 shows significant similarity to ActVA-ORF5 , lanZ5 could be a third gene of the landomycin biosynthetic gene cluster encoding an oxygenase. The amino acid sequences encoded by lanG and lanZ2 are highly similar to GraD (LanG: 69% identity, LanZ2: 46%) and StrD (LanG: 52%, LanZ2: 42%), suggesting that they encode dNDP-hexose synthetases. The deduced amino acid sequence of LanH resembles several dehydratases (e.g., 65% identity to GraE from S. violaceoruber Tü22 ) suggesting that lanH encodes a dTDP-glucose 4,6-dehydratase involved in C-6 dehydroxylation. LanQ strongly resembles AscC from Yersinia pseudotuberculosis (53% identity)  and proteins belonging to the SMAT (secondary metabolic aminotransferase) family . AscC is involved in C-3 dehydroxylation. As the biosynthesis of l-rhodinose requires dehydroxylation at position C-3 LanQ might be involved in this reaction. The predicted LanS enzyme is 46% identical to DnmT from S. peucetius. This enzyme may be involved as 2,3-dehydratases in deoxygenation at position C-2 of l-daunosamine. Amino acid residues 47–243 of LanS could be aligned to residues 262–461 (35% identity), indicating two similar substrate or cofactor binding sites in this protein. LanT is most related to RdmF from S. purpurascens and a glucose-fructose oxidoreductase (Gfor) from Zymomonas mobilis. The latter can catalyze oxidation of glucose to gluconolactone and the concomitant reduction of a fructose to sorbitol, during which the pyridine nucleotide cofactor remains bound to the protein. As a reduction and an oxidation step might be necessary for C-2 deoxygenation two enzymes (LanS and LanT) could be involved. LanR is most closely related to a UDP-glucose 4-epimerase (ExoB) from Azospirillum brasilense (31% identity)  and also shows some resemblance to a dTDP-6-deoxy-l-mannose dehydrogenase (RhsD) from Sphingomonas strain S88 (28% identity) . LanZ3 also resembles 4-ketoreductases. Here the highest resemblance was found to DnmV from S. peucetius (43% identity) involved in the biosynthesis of l-daunosamine as the sugar component of daunorubicin . LanR and LanZ3 are probably involved in reducing 4-keto intermediates to NDP-d-olivose and NDP-l-rhodinose respectively. LanZ1 resembles DnmU from S. peucetius (45% identity) . This enzyme is related to a large family of sequences that have been postulated to be involved in biosynthesis of l-sugars. LanGT1, LanGT2, LanGT3 and LanGT4 show significant similarity to putative glycosyltransferases , indicating an important function for these proteins in the biosynthesis of the sugar side chain. LanK resembles proteins involved in regulation of antibiotic biosynthesis. The closest similarity was found to JadR2 from S. venezuelae (29%)  and to TcmR from S. glaucescens (29%) [28, 29]. Homologous regions were especially found in the N-terminal regions of these proteins. LanJ resembles hydrophobic proteins known as proton-dependent transporters of different drugs . The closest resemblance was found to Brm3 from B. subtilis (EMBL accession number: D50098) and a hypothetical protein from Mycobacterium tuberculosis (EMBL accession number: Z97139). Interestingly all genes except lanK are transcribed in the same direction. The fact that lanK, a putative regulatory gene, is transcribed divergently from lanJ provides the possibility that LanJ could be regulated by LanK at the level of transcription .
The function of two further genes in the cluster is completely unknown. LanX is similar to some proteins known to be B12-dependent. Resemblance was found to a methylmalonyl-CoA mutases (McmA2) from Archaeoglobus fulgidus (EMBL accession number: AE000952). LanZ6 resembles Nsh-ORFC, a protein with unknown function from S. actuosus. No significant matches were found for LanU.
We have sequenced a region of DNA from S. cyanogenus S136 spanning about 35 kb. The analysis of the DNA sequence is clearly consistent with the possibility that it could encode genes for the biosynthesis of landomycin but absolute proof was not possible, since all attempts to insertionally inactivate genes of the cluster through gene disruption or gene replacement failed. Introduction of isolated cosmids into S. lividans and S. coelicolor CH999 by transformation gave apramycin-resistant colonies, and some of these colonies produced compounds with a similar color and Rf value to some of the landomycins (TLC). However, because of the instability of the cosmids we could not produce enough material of these compounds necessary for NMR analysis. We also could not prove that the whole landomycin biosynthetic gene cluster has been cloned, but after sequencing an additional 0.7 kb on both sides (upstream of lanE and downstream of lanZ6) of the cluster no further putative related gene could be identified.
We thank the Deutsche Forschungsgemeinschaft (SFB 323) and the European Community (BIO4-CT96-0068) (grants to A.B.). Work at the John Innes Centre was supported by the John Innes Foundation and by the Biotechnology and Biological Science Research Council (grants to D.A.H.).