RifDH10, the dehydratase domain from the terminal module of the rifamycin polyketide synthase, catalyzed the stereospecific syn dehydration of the model substrate (2S,3S)-2-methyl-3-hydroxypentanoyl-RifACP10, resulting in exclusive formation of (E)-2-methyl-2-pentenoyl-RifACP10. RifDH10 did not dehydrate any of the other three diastereomeric, RifACP10-bound, diketide thioester substrates. On the other hand, when EryACP6, from the sixth module of the erythromycin polyketide synthase, was substituted for RifACP10, RifDH10 stereospecifically dehydrated only (2R,3R)-2-methyl-3-hydroxypentanoyl-EryACP6 to give exclusively (E)-2-methyl-2-pentenoyl-EryACP6, with no detectable dehydration of any of the other three diastereomeric, EryACP6-bound, diketides. An identical alteration in substrate diastereospecificity was observed for the corresponding N-acetylcysteamine or pantetheine thioester analogues, regardless of acyl chain length or substitution pattern. Incubation of (2RS)-2-methyl-3-ketopentanoyl-RifACP10 with the didomain reductase-dehydratase RifKR10-RifDH10 yielded (E)-2-methyl-2-pentenoyl-RifACP10, the expected product of syn dehydration of (2S,3S)-2-methyl-3-hydroxypentanoyl-RifACP10, while incubation with the corresponding EryACP6-bound substrate, (2RS)-2-methyl-3-ketopentanoyl-EryACP6, gave only the reduction product (2S,3S)-2-methyl-3-hydroxypentanoyl-EryACP6 with no detectable dehydration. These results establish the intrinsic syn dehydration stereochemistry and substrate diastereoselectivity of RifDH10 and highlight the critical role of the natural RifACP10 domain in chaperoning the proper recognition and processing of the natural ACP-bound undecaketide substrate. The 1.82 Å-resolution structure of RifDH10 revealed the atomic resolution details of the active site and allowed modeling of the syn-dehydration of the (2S,3S)-2-methyl-3-hydroxyacyl-RifACP10 substrate. These results suggest that generation of the characteristic cis double bond of the rifamycins occurs after formation of the full-length RifACP10-bound acyclic trans-unsaturated undecaketide intermediate, most likely during the subsequent macrolactamization catalyzed by the amide synthase RifF.

Ketoreductase (KR) domains from modular polyketide synthases catalyze reduction of 2-methyl-3-ketoacyl-ACP substrates and in certain cases epimerization of the 2-methyl group as well. The structural and mechanistic basis of epimerization is poorly understood, and only a small number of such KRs been studied. We have studied three recombinant KR domains with putative epimerase activity – NysKR1 from module 1 of the nystatin PKS whose stereospecificity can be predicted from both protein sequence and product structure, RifKR7 from module 7 of the rifamycin PKS whose stereospecificity cannot be predicted from protein sequence, and RifKR10 from module 10 of the rifamycin PKS whose specificity is unclear from both sequence and structure. Each KR was individually incubated with NADPH and (2R)- or (2RS)-2-methyl-3-ketopentanoyl-ACP generated enzymatically in situ or via chemo-enzymatic synthesis, respectively. Chiral GC-MS analysis revealed that each KR stereospecifically produced the corresponding (2S,3S)-2-methyl-3-hydroxypentanoyl-ACP in which the 2-methyl substituent had undergone KR-catalyzed epimerization. Thus, our results have led to the identification of a prototypical set of KR domains that generate (2S,3S)-2-methyl-3-hydroxyacyl products in the course of polyketide biosynthesis.

No abstract