Quantitative analysis of loading and extender acyltransferases of modular polyketide synthases

The acyltransferase (AT) domains of modular polyketide synthases (PKSs) are the primary determinants of building block specificity in polyketide biosynthesis and are therefore attractive targets for protein engineering. Thus far, investigations into the fundamental biochemical properties of AT domains have been hampered by the inability to produce these enzymes as self-standing polypeptides. Here we describe an alternative, generally applicable strategy for overexpression and analysis of AT domains from modular PKSs as truncated didomain proteins (approximately 60 kDa). Recently, we reported the expression and reconstitution of the loading didomain of 6-deoxyerythronolide B synthase (Lau, J., Cane, D. E., and Khosla, C. (2000) Biochemistry 39, 10514-20). By replacing the AT domain of this protein with a methylmalonyl-CoA specific AT domain from module 6 of the 6-deoxyerythronolide B synthase, or alternatively a malonyl-CoA specific AT domain from module 2 of the rapamycin synthase, each of these extender unit AT domains could be overproduced and purified to homogeneity. Using acyl-CoA substrates as acyl group donors and N-acetylcysteamine as the thiol acceptor, we devised a steady-state kinetic assay to probe the properties of these three didomain proteins and selected mutants. Propionyl-CoA was the preferred substrate of the loading didomain, although acetyl- and butyryl-CoA were also accepted with approximately 40-fold-lower specificity. In contrast to the relatively relaxed specificity of the loading AT domain, the methylmalonyl- and malonyl-specific AT domains had high specificity (>1000-fold) toward their natural substrates. The acyl transfer reaction was inhibited by coenzyme A (CoASH) with both a competitive and a noncompetitive component. Use of an exogenous holo-acyl carrier protein (ACP) as an acceptor thiol did not increase the rate of acyl transfer relative to the reaction involving N-acetylcysteamine, suggesting that either the on-rate of the acyl group is rate-limiting or that the apo-ACP component of the didomain protein precludes effective docking of a second ACP onto the AT active site. Mutation of Trp-222 in the loading AT domain to an Arg residue that is universally conserved in all extender unit AT domains failed to enable the loading AT domain to accept methylmalonyl-CoA as an alternative substrate. In contrast, mutation of the equivalent Arg residue in an extender AT domain resulted in a protein with no activity. Together, these results provide a foundation for future structural and mechanistic investigations into the properties of AT domains of modular PKSs.