A putative catalytic triad consisting of tyrosine, serine, and lysine residues was identified in the ketoreductase (KR) domains of modular polyketide synthases (PKSs) based on homology modeling to the short chain dehydrogenase/reductase (SDR) superfamily of enzymes. This was tested by constructing point mutations for each of these three amino acid residues in the KR domain of module 6 of the 6-deoxyerythronolide B synthase (DEBS) and determining the effect on ketoreduction. Experiments conducted in vitro with the truncated DEBS Module 6+TE (M6+TE) enzyme purified from Escherichia coli indicated that any of three mutations, Tyr --> Phe, Ser --> Ala, and Lys --> Glu, abolish KR activity in formation of the triketide lactone product from a diketide substrate. The same mutations were also introduced in module 6 of the full DEBS gene set and expressed in Streptomyces lividans for in vivo analysis. In this case, the Tyr --> Phe mutation appeared to completely eliminate KR6 activity, leading to the 3-keto derivative of 6-deoxyerythronolide B, whereas the other two mutations, Ser --> Ala and Lys --> Glu, result in a mixture of both reduced and unreduced compounds at the C-3 position. The results support a model analogous to SDRs in which the conserved tyrosine serves as a proton donating catalytic residue. In contrast to deletion of the entire KR6 domain of DEBS, which causes a loss in substrate specificity of the adjacent acyltransferase (AT) domain in module 6, these mutations do not affect the AT6 specificity and offer a potentially superior approach to KR inactivation for engineered biosynthesis of novel polyketides. The homology modeling studies also led to identification of amino acid residues predictive of the stereochemical nature of KR domains. Finally, a method is described for the rapid purification of engineered PKS modules that consists of a biotin recognition sequence C-terminal to the thioesterase domain and adsorption of the biotinylated module from crude extracts to immobilized streptavidin. Immobilized M6+TE obtained by this method was over 95% pure and as catalytically effective as M6+TE in solution.

Cassette replacement of acyltransferase (AT) domains in 6-deoxyerythronolide B synthase (DEBS) with heterologous AT domains with different substrate specificities usually yields the predicted polyketide analogues. As reported here, however, several AT replacements in module 4 of DEBS failed to produce detectable polyketide under standard conditions, suggesting that module 4 is sensitive to perturbation of the protein structure when the AT is replaced. Alignments between different modular polyketide synthase AT domains and the Escherichia coli fatty acid synthase transacylase crystal structure were used to select motifs within the AT domain of module 4 to re-engineer its substrate selectivity and minimize potential alterations to protein folding. Three distinct primary regions of AT4 believed to confer specificity for methylmalonyl-CoA were mutated into the sequence seen in malonyl-CoA-specific domains. Each individual mutation as well as the three in combination resulted in functional DEBSs that produced mixtures of the natural polyketide, 6-deoxyerythronolide B, and the desired novel analogue, 6-desmethyl-6-deoxyerythronolide B. Production of the latter compound indicates that the identified sequence motifs do contribute to AT specificity and that DEBS can process a polyketide chain incorporating a malonate unit at module 4. This is the first example in which the extender unit specificity of a PKS module has been altered by site-specific mutation and provides a useful alternate method for engineering AT specificity in the combinatorial biosynthesis of polyketides.

Geldanamycin, a polyketide natural product, is of significant interest for development of new anticancer drugs that target the protein chaperone Hsp90. While the chemically reactive groups of geldanamycin have been exploited to make a number of synthetic analogs, including 17-allylamino-17-demethoxy geldanamycin (17-AAG), currently in clinical evaluation, the "inert" groups of the molecule remain unexplored for structure-activity relationships. We have used genetic engineering of the geldanamycin polyketide synthase (GdmPKS) gene cluster in Streptomyces hygroscopicus to modify geldanamycin at such positions. Substitutions of acyltransferase domains were made in six of the seven GdmPKS modules. Four of these led to production of 2-desmethyl, 6-desmethoxy, 8-desmethyl, and 14-desmethyl derivatives, including one analog with a four-fold enhanced affinity for Hsp90. The genetic tools developed for geldanamycin gene manipulation will be useful for engineering additional analogs that aid the development of this chemotherapeutic agent.

Two new geldanamycin derivatives produced by genetic engineering of Streptomyces hygroscopicus strain K309-27-1 were isolated and characterized. Removal of the 8-methyl group of geldanamycin was achieved by replacing the AT4 domain of the polyketide synthase with a malonyl AT domain. The resulting strain produced 8-demethyl geldanamycin (2) and 4,5-epoxy-8-demethylgeldanamycin (3). The structures of both molecules were elucidated through interpretation of 1D and 2D NMR data as well as comparison with authentic geldanamycin derivatives. Compounds 2 and 3 displayed moderate cytotoxicity against the human breast cancer cell line SK-BR-3.

The megalomicin and erythromycin polyketide synthases (PKSs) produce the same aglycon product, 6-deoxyerythronolide B (6-dEB). Both PKSs were examined in an Escherichia coli strain metabolically engineered to support complex polyketide biosynthesis. Production of 6-dEB in shake flask fermentations was undetectable by mass spectrometry in the strain expressing the megalomicin (Meg) PKS genes, whereas 31 mg/L 6-dEB was produced by the strain with the erythromycin (DEBS) PKS. The genes for each of the three subunits comprising the PKSs were expressed in different combinations from three compatible expression vectors (e.g., DEBS1, DEBS2, and MegA3) to identify two Meg PKS subunits, MegA1 and MegA3, which conferred lower 6-dEB titers than their DEBS counterparts. Comparison of protein expression levels and 6-dEB titers by engineered hybrid DEBS/Meg PKS genes further defined regions within modules 2 and 6 of MegA1 and MegA3, respectively, which limit protein expression and 6-dEB production in E. coli. Meg module 2 + TE (M2 + TE) and a hybrid DEBS M2/Meg M2 + TE protein were engineered and purified for in vitro comparisons with DEBS M2 + TE. The specific activity of the hybrid M2 + TE was approximately 16-fold lower than DEBS M2 + TE and only twice as high as the Meg M2 + TE enzyme in diketide elongation assays. Since the hybrid M2 worked comparably to DEBS M2 in vivo, this suggests that boosting subunit concentration could serve as a useful approach to overcome enzyme deficiencies in heterologous polyketide production.