Engineering of Chimeric Polyketide Synthases Using SYNZIP Docking Domains

Abstract

Engineering of assembly line polyketide synthases (PKSs) to produce novel bioactive compounds has been a goal for over 20 years. The apparent modularity of PKSs has inspired many engineering attempts in which entire modules or single domains were exchanged. In recent years, it has become evident that certain domain-domain interactions are evolutionarily optimized and, if disrupted, cause a decrease of the overall turnover rate of the chimeric PKS. In this study, we compared different types of chimeric PKSs in order to define the least invasive interface and to expand the toolbox for PKS engineering. We generated bimodular chimeric PKSs in which entire modules were exchanged, while either retaining a covalent linker between heterologous modules or introducing a noncovalent docking domain, or SYNZIP domain, mediated interface. These chimeric systems exhibited non-native domain-domain interactions during intermodular polyketide chain translocation. They were compared to otherwise equivalent bimodular PKSs in which a noncovalent interface was introduced between the condensing and processing parts of a module, resulting in non-native domain interactions during the extender unit acylation and polyketide chain elongation steps of their catalytic cycles. We show that the natural PKS docking domains can be efficiently substituted with SYNZIP domains and that the newly introduced noncovalent interface between the condensing and processing parts of a module can be harnessed for PKS engineering. Additionally, we established SYNZIP domains as a new tool for engineering PKSs by efficiently bridging non-native interfaces without perturbing PKS activity.

Figure 1.

Schematic architecture of DEBS. (A) The three polypeptides (DEBS1–3) and the encoded modules (M1–M6), as well as the loading didomain (LDD), the thioesterase domain (TE), and the final product (1) are depicted. Polyketide intermediates are shown as attached to the respective acyl carrier protein (ACP). Black tabs depict docking domains. Domain annotations are as follows: AT, acyltransferase; KS, ketosynthase; ACP, acyl carrier protein; KR, ketoreductase; DH, dehydratase; ER, enoylreductase. Module 3 has a ketoreductase-like domain (denoted in lowercase) that lacks NADPH-dependent oxidoreductase activity but harbors C-2 epimerase activity. (B) The intramodular chain elongation and intermodular chain translocation steps are shown for M1. The ACP of this module carries the α-carboxyacyl-chain (methylmalonyl-ACP) or the (2S,3R)-2-methyl-3-hydroxy-diketide condensation product.

Figure 2.

Use of SYNZIP domains in the design of catalytically efficient PKSs harboring a split module. (A) Design of a bimodular DEBS derivative comprised of LDD(4), intact (5)M1(2) or a split version thereof, and (3)M2-TE. (B) Model of DEBS M1 with the natural AT-KR linker (top) and the SYNZIP-containing variant (bottom). This model was built based on SAXS analysis of DEBS M3. Fusion sites of the SYNZIP domains to either the AT or KR are indicated by blue and pink dots, illustrating the parallel orientation of the heterospecific coiled-coil. An eight-residue flexible Gly-Ser linker is used to connect the SYNZIP domain to the PKS protein (see protein sequence Table S2). (C) Turnover rates of bimodular PKSs employing M1 or a split M1. Except for intact M1, at least two independently purified protein preparations were evaluated (1 or 2): each preparation of the N-terminal part is shown in a separate column, whereas preparations of the C-terminal part are indicated as black and white dots. All initial rate data was obtained at 2 μM enzyme concentration and nonlimiting concentrations of propionyl-CoA, methylmalonyl-CoA, and NADPH. Measurements were performed in triplicate, and the grand mean is indicated.

Figure 3.

Comparison of chain translocation and elongation rates of intact versus split modules. The occupancy of individual proteins by growing polyketide chain precursors was measured in the absence (A) or presence (B) of methylmalonyl-CoA, and in modules containing chimeric KS:ACP interfaces (C, with methylmalonyl-CoA). All measurements included LDD(4) plus the designated split or intact module combinations. Labeling of the different proteins was quantified at different time points (1, 3, 5, and 10 min), and the resulting counts were normalized based on the maximal occupancy. Combined counts of both the KS and ACP containing proteins are also presented for split-module systems. In the case of RIFS Module 1, the N- and C-terminal fragments could not be separated by SDS-PAGE gel; hence only combined counts are reported (panel C). All measurements were performed in triplicate using 2 μM enzyme concentration and nonlimiting concentrations of propionyl-CoA, methylmalonyl-CoA, and NADPH.

Figure 4.

Turnover rates of bimodular chimeric PKSs harboring a chimeric chain translocation interface. Design and turnover rates of bimodular chimeric PKSs. LDD(4) was always used as a separate protein, whereas DEBS M1 and variable acceptor modules were interfaced using either noncovalent docking domains (A), SYNZIP domains (B), or a combination of covalent fusions and SYNZIP domains (C). Construct design and the predicted triketide lactone products are shown next to the measured turnover rate of each system. All initial rate data was obtained at 2 μM enzyme concentration and nonlimiting concentrations of propionyl-CoA, methylmalonyl-CoA, and NADPH. In all cases, two independently purified protein preparations were evaluated (data shown as black and white dots). Each set of measurements was performed in triplicate, and the grand mean is indicated. LC-MS analysis of bimodular PKSs newly generated in this study (B, C) was performed after overnight incubation to verify product identity (Figure S3). The expected products were detected for all reactions except for the bimodular PKS with SZ4-M3-TE as the acceptor. For product analysis of the reference system, refer to Klaus et al.

Figure 5.

Turnover rates of bimodular chimeric PKSs harboring a chimeric chain elongation interface. Bimodular PKSs comprised of LDD(4),(5) KS1-AT1-SZ3, and variable SZ4-KR_n-ACP_n-Module_n+1-TE constructs. Construct design and the predicted triketide lactone products are shown next to the measured turnover rate of each system. All initial rate data was obtained at 2 μM enzyme concentration and nonlimiting concentrations of propionyl-CoA, methylmalonyl-CoA, malonyl-CoA (RIFS M2 is malonyl specific), and NADPH. Turnover analysis was performed on two individually purified proteins per construct (back and white dots). Measurements were performed in triplicate, and the grand mean is indicated.