Loss of Single‐Domain Function in a Modular Assembly Line Alters the Size and Shape of a Complex Polyketide

Abstract The structural wealth of complex polyketide metabolites produced by bacteria results from intricate, highly evolved biosynthetic programs of modular assembly lines, in which the number of modules defines the size of the backbone, and the domain composition controls the degree of functionalization. We report a remarkable case where polyketide chain length and scaffold depend on the function of a single β‐keto processing domain: A ketoreductase domain represents a switch between diverging biosynthetic pathways leading either to the antifungal aureothin or to the nematicidal luteoreticulin. By a combination of heterologous expression, mutagenesis, metabolite analyses, and in vitro biotransformation we elucidate the factors governing non‐colinear polyketide assembly involving module skipping and demonstrate that a simple point mutation in type I polyketide synthase (PKS) can have a dramatic effect on the metabolic profile. This finding sheds new light on possible evolutionary scenarios and may inspire future synthetic biology approaches.

Abstract: The structural wealth of complex polyketide metabolites produced by bacteria results from intricate,h ighly evolved biosynthetic programs of modular assembly lines,i n which the number of modules defines the sizeofthe backbone, and the domain composition controls the degree of functionalization. We report aremarkable case where polyketide chain length and scaffold depend on the function of as ingle b-keto processing domain:Ak etoreductase domain represents as witch between diverging biosynthetic pathwaysl eading either to the antifungal aureothin or to the nematicidal luteoreticulin. By ac ombination of heterologous expression, mutagenesis,m etabolite analyses,a nd in vitro biotransformation we elucidate the factors governing non-colinear polyketide assembly involving module skipping and demonstrate that as imple point mutation in type Ip olyketide synthase (PKS) can have ad ramatic effect on the metabolic profile.T his finding sheds new light on possible evolutionary scenarios and may inspire future synthetic biology approaches.
Modular type Ip olyketide synthases (PKSs) assemble ab road range of ecologically and pharmaceutically relevant molecules. [1] In contrast to the chemical synthesis of complex polyketides,w hich often requires challenging, multistep transformations, [2] type IP KSs connect and process simple acyl and malonyl thioester building blocks in af ully programmed fashion. [1] Structural diversity results from variations in the biosynthetic program. Ty pically,a na ctivated starter unit is loaded onto the PKS and propagated by modules that minimally consist of ak etosynthase (KS) domain catalyzing aC laisen condensation for CÀCb ond formation, an acyltransferase (AT) domain selecting and loading the malonyl extender unit, and an acyl carrier protein (ACP) domain serving as an anchor for the growing acyl chain. Thep olyketide chain is passed from one module to another until it is released from the PKS,usually catalyzed by at hioesterase (TE) domain. [3] In each module optional ketoreductase (KR), dehydrogenase (DH), and enoylreductase (ER) domains determine to which degree the b-keto groups of the intermediates are processed. Owing to the unidirectional propagation of the nascent chain bound to the assembly lines,the structures of the polyketide backbones and the corresponding modular PKSs are typically colinear. [1] Thus,i ti sg enerally feasible to predict the basic polyketide structures based on the number and architectures of the PKS modules,a nd vice versa. (Figure 1) This co-linearity rule has been successfully applied for genome mining [4] and rational biosynthetic engineering approaches. [5] Furthermore,i tp rovides am odel for the evolution of polyketide diversity; whereas gene duplications or deletions would lead to different chain lengths (number of modules), the gain or loss of encoded domain functions would influence the degree of bketo processing (composition of modules). [6] To gain insight into such evolutionary processes and to emulate possible recombination scenarios we have studied in detail the biosynthesis of structurally related bacterial polyketides that share characteristic nitroaryl and pyrone moieties. [7] Functional analyses of the assembly lines for the prototype of this family,t he antifungal and antiproliferative agent aureothin (aur, 1) [8] and ah igher homologue,n eoaureothin (syn. spectinabilin), [9] revealed that the involved PKSs breach with the colinearity rule. [10] Specifically,the first modules and the penultimate AT domains are used itera- tively. [11] Based on the deduced biosynthetic program of the aur PKS we have engineered an artificial pathway for the nematicide luteoreticulin [12] (syn. griseulin, [13] 2). [14] This congener has ar educed chain length compared to 1,a n isomeric pyrone ring, and an altered substitution pattern. Although it has been feasible to morph the aur PKS into an artificial luteoreticulin (alut)a ssembly line (Supporting Information, Figure S1), [14] in the plethora of sequenced bacterial genomes ag enuine gene cluster coding for luteoreticulin biosynthesis has not yet been detected. Here we elucidate the true biosynthetic origin of luteoreticulin and show the unexpected impact of single loss-of-function mutations of amodular PKS on the metabolite scaffold.
Prompted by the surprising observation that the aureothin producer strain Streptomyces thioluteus produces minute amounts of 2,w er evisited reported luteoreticulin producers (S. luteoreticuli [12] and S. griseus [15] )and noted that 1 was also detected in their fermentation broths.T ot est the possibility that 2 is as ide product of the aur pathway we investigated aheterologous host exclusively expressing the aur biosynthesis gene cluster (S. albus::pHJ48).
[7a] By metabolic profiling of an up-scaled culture of this designated producer strain we detected small amounts of 2.G iven that both polyketide metabolites (1 and 2)d iffer in both size and shape,i ti s remarkable that they seem to derive from the same modular assembly line.
Thef ormation of 2 as ab yproduct of the aur PKS ( Figure 2B)c ould be rationalized by erratic substrate processing or by non-functional catalytic domains.Aretrobiosynthetic analysis suggested that the pyrone ring would result from the cyclization of an enol intermediate.T he requisite carbonyl group would require an impaired b-keto reduction after three rounds of elongations.T hus,w e postulated that the ketoreductase domain (KR2) in module 2( AurB) constitutes ab ranching point for the lut and aur pathways.
To probe this hypothesis we scrutinized the KR2 domain. Alignment with other known KR domains ( Figure S2) indicated that KR2 belongs to B1-type KRs,w hich are characterized by adiagnostic LDD motif (VDD in KR2) and the absence of aT rp eight residues upstream of the catalytic Ty r. [3a] We also identified the catalytic Lys, Ty r, and Ser moieties that aid in binding and reducing polyketide intermediates.The only deviation is that leucine is replaced by Va l in the conserved LDD motif.H owever,t his replacement occasionally appears in other active KR domains [16] (e.g., Nys3 and Nys12, Figure S2) and is thus unlikely to influence the reductive activity.S ince KR2 is obviously functional, an impaired b-keto reduction could result from alimited supply of the essential cofactor NADPH or from as low turnover rate.
To test whether ac omplete shutdown of KR2 has an impact on the production of 2,w ei nactivated the KR2 domain. Therefore,w er eplaced tyrosine with phenylalanine in the catalytic motif YA AAN by site-directed mutagenesis of the PKS gene cloned into plasmid pHJ48 (Figures 2Band S3). Theresulting mutated plasmid (pHY127) was introduced into expression host S. albus by conjugation to generate the KR2 null mutant (S. albus::pHY127).
Them etabolic profiles of S. albus::pHY127 and S. albus::pHJ48 were compared. As trong yellow pigmentation indicated an altered metabolite spectrum of the KR2 null mutant ( Figure 2C). HPLC-MS analyses of the culture extracts showed that the production of 2 increased dramatically in the KR2 null mutant (50.8 mg L À1 ), whereas 1 could only be detected in trace amounts (0.36 mg L À1 )( Figure 2D, trace iv and Figure S7). Notably,p roduction of 2 in the KR2 null mutant was 500-fold higher than in the strain with the artificial lut (alut)P KS (0.1 mg L À1 ). [14] This finding strongly suggests that luteoreticulin production could result of asingle loss of function in AurB.A saconsequence of the impaired ketoreduction in module 2t he polyketide intermediate obviously skips adownstream module.
To gain insight into the impact of incomplete b-keto processing we generated two additional mutants,o ne that is deficient in dehydration, and another one that lacks af unc- tional ER domain. Therefore,wereplaced the catalytic His of the HVVLGSTLVP motif [3a] of DH2 by Phe,y ielding DH2 null mutant S. albus::pHY140 ( Figure S4). To obtain the ER2 null mutant, S. albus::pHY147 ( Figure S5), we changed the conserved NADPH-binding motif GGVGMA [3a] to SPVGMA in ER2. HPLC analysis of the metabolic profile of the DH2 null mutant showed that in lieu of 1 and 2,which can only be detected in trace amounts,s everal new compounds (3-5)a re produced ( Figure 3B,t race iii). These metabolites were isolated in pure form by preparative HPLC to obtain 3 (6.4 mg), 4 (24.7 mg), and 5 (3.6 mg) from 1Lculture of DH2 null mutant (S. albus::pHY140), and their structures were elucidated by 1D and 2D NMR analysis ( Figure 3A,F igures S8-S28, and Tables S1-S3). Theabsolute configurations of 3 and 4 were determined by the combination of in silico analyses [17] and chemical derivatization (Figures S29-S43, Tables S4-S6).
In agreement with the domain set of the mutated module 2(DH2 null), compound 3 is acongener of 1 with ahydroxyl group at C8 that results from incomplete b-keto processing. Compound 4 has an additional hydroxyl group at C7, which is likely introduced by the cytochrome P450 monooxygenase AurH. [18] Compound 5 differs from 3 in ak eto group at C8. This keto group might result from a) impaired ketoreduction, or b) oxidation of 3,p ossibly by AurH. To elucidate the biogenesis of these oxygenated compounds,w eh eterologously produced AurH in E. coli used recombinant AurH for biotransformation experiments.I nv itro assays showed that AurH transforms 3 into 4 and 5 (Figures 3C and S58). Consequently, 4 results from AurH-mediated C7-hydroxylation, and 5 is formed by the oxidative route, [19] which is plausible since an impaired KR at this stage would channel the intermediate into the lut pathway.
From the HPLC profile of the ER2 null mutant, we detected two new compounds, 6 and 7 ( Figure 3B,t race vi). Thes tructures of 6 and 7 were determined by 1D and 2D NMR analyses of the isolated and purified compounds ( Figures 3A and S44-S57, and Tables S7 and S8). Both 6 and 7 feature double bonds at the respective C7-C8 positions that result from ketoreduction and dehydratation, and differ only in the configuration of the O-methylated pyrone ring (aor g-position), which is mediated by the methyltransferase AurI. [20] It is remarkable that trace amounts of 2 are detectable in all mutants.The formation of 2 can be explained by the non-quantitative reduction of the b-keto group in module 2asobserved in the wild-type PKS (S. albus::pHJ48). Tr aces of 1 in the DH2 null mutant can be explained by nonenzymatic dehydration of the hydroxy intermediate,orbythe involvement of al ong-range acting DH domain as in isomigrastatin biosynthesis. [21] Thek ey message of these mutational experiments is, however, that the biosynthesis of 2 primarily depends on the dysfunction of the KR. Furthermore,itisremarkable that no module skipping takes place when the b-keto group is processed into either b-hydroxyl or enoyl groups.This finding indicates that the unreduced b-keto group in module 2i s essential for the skipping process required for lut biosynthesis. Since AT4i ss maller than typical AT domains,a nd the function of AT4c an be substituted by AT3f rom the penultimate module, [14] module 4islikely skipped. To test this hypothesis,wereplaced AT4with AT3i nthe KR2 null mutant. HPLC-MS analysis revealed that 2 was still produced as the main product of the KR2 null + aAT4 mutant (S. albus::pHY145) ( Figures 4A-C and S6). Consequently, the AT domain exchange does not affect the skipping of the fourth module ( Figure 4C and D). Ap lausible explanation for this observation is that the KS4 domain solely plays arole as agatekeeper that recognizes the structure of the polyketide chain. Ther equired b,d-diketo thioester intermediate for pyrone formation could be generated by the iterative use of module 3(aur pathway) or by the sequential single elongation of mutated module 2a nd module 3( lut pathway). While substrate specificities of KS domains are best studied for trans-ATP KSs, [22] it is astonishing that the aur KS2 domain downstream of iterative module 1e xhibits as imilar gatekeeping function. [23] Seminal studies have shown that mutagenesis and domain swaps of modular PKS lead to mainly predictable derivatives of the parent polyketide backbones. [5a,24] These genetic manipulations have led to ab road range of compounds with diverse substitution patterns,yet the size and overall scaffold of the polyketides has remained unaffected. We report ar adically different scenario where as ingle loss of domain function or even as ingle point mutation of am odular PKS leads to ap roduct with altered size and shape.T his is an unusual mechanism by which structural diversity of biologically active polyketide products is created and illustrates an overlooked ability of PKS to breach with the colinearity rule on the basis of the intermediates redox state.T ogether with recently reports on keto-processing PKS domains that are toggled on and off in an iterative module [25] or have become dysfunctional during evolution [26] to create structural diversity,t hese insights may guide future approaches to generate polyketide diversity by PKS engineering.  ::pHY145). B) Amino acid sequences of original and mutated sites (introduced by restriction enzyme HpaI). C) HPLC profiles of (i) aureothin reference (1), (ii)l uteoreticulin reference (2), and (iii) metaboilites of KR2 null + aAT4 mutant. UV detection is at 350 nm. D) Module skipping is not influencedb yA T4 activity.