Analysis of a “Split-and-Stuttering” Module of an Assembly Line Polyketide Synthase

Notwithstanding the “one-module-one-elongation-cycle” paradigm of assembly line polyketide synthases (PKSs), some PKSs harbor modules that iteratively elongate their substrates through a defined number of cycles. While some insights into module iteration, also referred to as “stuttering”, have been derived through in vivo and in vitro analysis of a few PKS modules, a general understanding of the mechanistic principles underlying module iteration remains elusive. This report serves as the first interrogation of a stuttering module from a trans-AT subfamily PKS that is also naturally split across two polypeptides. Previous work has shown that Module 5 of the NOCAP (nocardiosis associated polyketide) synthase iterates precisely three times in the biosynthesis of its polyketide product, resulting in an all trans-configured triene moiety in the polyketide product. Here we describe the intrinsic catalytic properties of this NOCAP synthase module. Through complementary experiments in vitro and in E. coli, the “split-and-stuttering” module was shown to catalyze up to five elongation cycles, although its dehydratase domain ceased to function after three cycles. Unexpectedly, the central olefinic group of this truncated product had a cis configuration. Our findings set the stage for further in-depth analysis of a structurally and functionally unusual PKS module with contextual biosynthetic plasticity. TOC/Abstract Graphic

the NOCAP (nocardiosis associated polyketide) synthase iterates precisely three times in the biosynthesis of its polyketide product, resulting in an all trans-configured triene moiety in the polyketide product. Here we describe the intrinsic catalytic properties of this NOCAP synthase module. Through complementary experiments in vitro and in E. coli, the "split-and-stuttering" module was shown to catalyze up to five elongation cycles, although its dehydratase domain ceased to function after three cycles. Unexpectedly, the central olefinic group of this truncated product had a cis configuration. Our findings set the stage for further in-depth analysis of a structurally and functionally unusual PKS module with contextual biosynthetic plasticity.

Main
Polyketide synthases (PKSs) are multifunctional enzymes responsible for the biosynthesis of structurally diverse polyketide products, many of which are clinically used as medicines (1). Therefore, understanding their catalytic diversity and mechanisms has been of high interest to researchers in the hope of engineering novel pharmacologically useful molecules. One class of PKSs, called assembly line PKSs, consists of multiple catalytic modules each of which is responsible for a well-defined set of chemical transformations on the growing polyketide chain. Minimally a PKS module harbors a ketosynthase (KS), acyltransferase (AT), and acyl carrier protein (ACP) domain (2). Some PKS modules (hereafter designated cis-AT PKSs) include the AT as an integral domain of a multifunctional polypeptide, whereas other modules (designated trans-AT PKSs) engage stand-alone AT proteins (3).
Canonically, each module of an assembly line PKS catalyzes a single elongation and modification cycle, where after the growing polyketide chain is either translocated onto the KS domain of the next module or it is off-loaded from the assembly line (4,5). Such a collinear architecture has facilitated prediction of structural features of the resulting polyketide products (6). However, not all PKSs operate according to a collinear one-module-one-elongation-cycle model (3). For example, the NOCAP synthase (7,8) harbors a trans-AT module (Module 5) that catalyzes three successive elongation and modification cycles on the growing polyketide chain yielding products 1 and 2 ( Figure 1) (9,10). After each cycle, the elongated and modified polyketide product is passed from the ACP domain of Module 5 back to its own KS domain (11). This type of programmed module iteration, also known as "stuttering" (12,13), has been identified in a few other assembly line PKSs (14) such as the stigmatellin (15), lankacidin (16), borrelidin (17), neoaureothin and aureothin (18) PKS pathways.
While the governing principles that control programmed iteration are not well understood, some insights have been gained through in vitro and in vivo analysis of cis-AT PKSs. Module 1 of the aureothin synthase (cis-AT PKS), which catalyzes two elongation and modification cycles in the context of the complete assembly line PKS, was expressed in Streptomyces lividans (19). The resulting strain produced a spectrum of compounds that had undergone up to 4 rounds of elongation, although products resulting from 3 and 4 rounds of iteration were orders of magnitude less abundant. Additionally, an in vitro analysis of Module 5 of the borrelidin PKS revealed that the KS domain of this stuttering module accepted non-native substrates corresponding to the chain lengths of its natural intermediates (20). Collectively, these and other findings have led to a mechanistic model in which the KS specificity at least partly dictates the number of iterations observed. However, the downstream module also has been suggested to exhibit a gatekeeping role (21). Given the lack of prior studies on stuttering modules from trans-AT PKSs, we interrogated Module 5 of the NOCAP synthase. Of particular interest to us was its additionally unusual "split-and-stuttering" feature, with its KS and ACP domains distributed across separate proteins (Figure 1).
In order to investigate the potential gatekeeping role of downstream modules in the NOCAP synthase we sought to develop a truncated assembly line excluding Modules 6-8-TR. Fortunately, a system for functionally expressing the entire NOCAP synthase in E. coli BAP1 (22) was recently developed which was adapted for the present study (10). Three plasmids with unique antibiotic resistance markers were introduced into E. coli BAP1 (Figure 2, Figure S1). Plasmid pCK-KPY178 encodes the malonyl-CoA synthetase MatB from Streptomyces coelicolor (23), the Rhizobium leguminosarum malonate carrier protein MatC (24), the loading module (Module L) and trans-AT-TEII (tAT-TEII) of the NOCAP synthase. Plasmid pCK-KPY222 encodes Modules 1-3 with engineered docking domains to facilitate protein-protein interactions (25). Finally, pKMG14 or alternatively pKMG15 encodes Module 4 and the DH-ACP-KR or the DH-ACP domains of Module 5, respectively. In the absence of a specific product release mechanism in these truncated assembly lines, we relied on spontaneous hydrolysis of polyketide products from ACP5.
One to three liters of E. coli BAP1 [pCK-KPY178/pCK-KPY222/pKMG14] supernatant was subjected to solid phase extraction (SPE) followed by high performance liquid chromatography (HPLC) (Figure S2). Encouragingly, two compounds 3 and 4 with distinct UV signatures were identified (Figure 2, Figures S3-S4). Control strains harboring pKMG15 in lieu of pKMG14 or lacking either plasmid failed to produce either 3 or 4 (Figure S5). High-resolution mass spectrometry (HRMS) analysis revealed that 3 had a molecular formula of C20H28O5 (m/z:  (Figure S6). Similar to the relationship between 1 and 2, these predicted formulas differed by a mass shift consistent with an unsaturated two-carbon (C2H2) moiety. To verify the anticipated biosynthetic relationship between 3 and 4, an E. coli BAP1 strain lacking Module 3 [pCK-KPY178/pCK-KPY102/pKMG14] was engineered (SI Methods). The resulting strain yielded 4 but not 3 (Figure S5), confirming a requirement of Module 3 for the biosynthesis of the latter compound. However, because these HRMS values were inconsistent with our expectation that Module 5 catalyzed three elongation and modification cycles, we turned to NMR spectroscopy to elucidate the structures of 3 and 4.
The structures of 3 and 4 were uncovered by extensive 1-D and 2-D NMR analysis, including 1 H 1D, 13 C 1D, COSY, HSQC, HMBC, and ROESY experiments (Figure 3, Figures  S7-S31, Table 1). The structures deduced from NMR analysis suggested that Module 5 catalyzed five elongation cycles in the biosynthesis of these compounds, but that the final two cycles were not accompanied by dehydration of the -hydroxyacyl intermediate (Figure 4,   Figure S32). Furthermore, the presence of a cross-peak in the ROESY spectrum implied the existence of a cis-olefinic bond between C8-C9 in contrast to the all-trans configuration of 1 and 2 (Figures S28 and S31). To our knowledge this is the first reported instance of "imperfect stuttering", where the programmed sequence of chain modification reactions is not faithfully catalyzed in every iteration. The unexpected chain length and configuration of 3 (and 4) raised the possibility that endogenous E. coli enzymes played a role in its biosynthesis. For instance, FabA from the fatty acid synthase of E. coli iteratively catalyzes the dehydration of ACP-bound -hydroxyacyl intermediates generated during fatty acid biosynthesis. FabA can also reversibly isomerize trans-2-decenoyl-ACP to cis-3-decenoyl-ACP (26). Therefore, to rule out the possibility of artifactual involvement of host enzymes, we sought to functionally reconstitute the truncated hexamodular derivative of the NOCAP synthase in vitro using purified proteins.
The following proteins were expressed and purified via immobilized metal ion chromatography followed by anion exchange chromatography: DH5-ACP5-KR5, DH5-ACP5 and tAT-TEII from the NOCAP synthase, and S. coelicolor MatB (Figure S33, SI Methods).  Figure S34). 4 was not synthesized in detectable amounts in vitro. These results largely support our proposed biosynthetic pathway. However, despite considerable effort, these in vitro reactions did not yield sufficient material for ROESY verification of 3's geometry.
In summary, this is the first description of a trans-AT "split-and-stuttering" module that preferentially iterates for additional cycles beyond those observed in the context of the full PKS assembly line. We therefore propose that Module 6 of the NOCAP synthase regulates the overall polyketide chain length by preferentially recognizing the ACP5-bound intermediate after it has undergone three elongation and modification cycles on Module 5. The inability of DH5 to catalyze dehydration after three elongation cycles revealed an unprecedented feature of this PKS domain. Future efforts will aim to elucidate the underlying mechanisms that govern module 5's particularly unusual iterative behavior. Overall, our findings demonstrate the remarkable contextual biosynthetic plasticity of Module 5 of the NOCAP synthase while highlighting the opportunity for future polyketide engineering through a deeper understanding of iterative PKS modules.