Stereospecific Formation of Z-Trisubstituted Double Bonds by the Successive Action of Ketoreductase and Dehydratase Domains from trans-AT Polyketide Synthases

Abstract

Incubation of (±)-2-methyl-3-ketobutyryl-SNAC (3) and (±)-2-methyl-3-ketopentanoyl-SNAC (4) with BonKR2 or OxaKR5, ketoreductase domains from the bongkrekic acid (1) and oxazolomycin (2) polyketide synthases, in the presence of NADPH gave in each case the corresponding (2R,3S)-2-methyl-3-hydroxybutyryl-SNAC (5) or (2R,3S)-2-methyl-3-hydroxypentanoyl-SNAC (6) products, as established by chiral GC-MS analysis of the derived methyl esters. Identical results were obtained by BonKR2- and OxaKR5-catalyzed reduction of chemoenzymatically prepared (2R)-2-methyl-3-ketopentanoyl-EryACP6, (2R)-2-methyl-3-ketobutyryl-BonACP2 (12), and (2R)-2-methyl-3-ketopentanoyl-BonACP2 (13). The paired dehydratase domains, BonDH2 and OxaDH5, were then shown to catalyze the reversible syn dehydration of (2R,3S)-2-methyl-3-hydroxybutyryl-BonACP2 (14) to give the corresponding trisubstituted (Z)-2-methylbutenoyl-BonACP2 (16).

Graphical abstract

Double bonds are ubiquitous structural features found in thousands of known bacterial polyketide natural products. While the vast majority are disubstituted or trisubstituted E double bonds, numerous polyketides harbor isomeric Z double bonds, such as those found in bongkrekic acid (1)¹ and oxazolomycin A (2)² (Figure 1), as well as in fostriecin,^3–5 phoslactomycin,⁶ borrellidin, ^{7, 8} bacillaene,^{9, 10} curacin,^11,12 difficidin,¹³ chivosazol A,¹⁴ disorazol A,¹⁴ lactimidomycin,¹⁴ and macrolactin, ¹⁴ among many others. Modular polyketide synthases (PKSs) typically generate each double bond by the coupled action of a ketoreductase (KR) domain, which carries out the stereospecific, NADPH-dependent reduction of a 3-ketoacyl-acyl carrier protein (ACP) or a 2-methyl-3-ketoacyl-ACP chain elongation intermediate, and a paired dehydratase (DH) domain, which catalyzes the syn dehydration of the reduced 3-hydroxyacyl-ACP or 2-methyl-3-hydroxyacyl-ACP intermediate.

Figure 1.

Domain organization of the trans-AT polyketide synthase modules that generate (Z)-trisubstituted double bonds in bongkrekic acid (1) and oxazolomycin (2). The indicated methyl groups are introduced by C-methyl transferase (MT) domains.

A variety of DH domains from modular PKSs have been expressed as discrete proteins and their reactions characterized biochemically. Thus EryDH4, from module 4 of the erythromycin PKS,¹⁵ and NanDH2, from module 2 of the nanchangmycin PKS,¹⁶ each catalyze the syn dehydration of a (2R,3R)-2-methyl-3-hydroxyacyl-ACP substrate to the corresponding trisubstituted (E)-2-methylenoyl-ACP. By contrast, RifDH10, from module 10 of the rifamycin PKS, catalyzes the analogous syn dehydration of the diastereomeric (2S,3S)-2-methyl-3-hydroxyacyl-ACP to an (E)-2-methylenoyl-ACP product, in spite of the fact that the derived trisubstituted double bond in the ultimately formed rifamycin has the Z configuration, suggesting that the geometry must be altered subsequent to double bond formation.¹⁷ PicDH2, from module 2 of the picromycin PKS,^{18, 19} and FosKR1, from module 1 of the fostriecin PKS,⁵ each catalyze the dehydration of a (3R)-3-hydroxyacyl-ACP substrate to the corresponding disubstituted (E)-2-enoyl-ACP. The predicted syn stereochemistry of the latter two dehydration reactions has yet to be confirmed experimentally. We recently reported that FosDH2, from module 2 of the fostriecin PKS, catalyzes the reversible dehydration of a (3S)-3-hydroxyacyl-ACP to the corresponding disubstituted (Z)-2-enoyl-ACP product.⁵ This report was the first to document the in vitro DH-catalyzed formation of any (Z)-enoyl-thioester product. Indeed, for a variety of reasons, all prior attempts to demonstrate the predicted formation of disubstituted or trisubstituted (Z)-double bonds by recombinant PKS DH domains had been unsuccessful, resulting instead in the exclusive generation of the isomeric (E)-double bonds.^{7, 11, 12, 17, 20}

We now report that BonDH2, from module 2 of the bongkrekic acid trans-AT PKS, and OxaDH5, from module 5 of the oxazolomycin trans-AT PKS (Figure 1), each catalyze the stereospecific syn dehydration of (2R,3S)-2-methyl-3-hydroxyacyl-ACP substrates, generated by the paired ketoreductases, BonKR2 and OxaKR5, respectively, to give the corresponding trisubstituted (Z)-2-methylenoyl-ACP products.

BonKR2 and BonDH2, from Burkholderia gladioli pathovar cocovenenans, and OxaKR5 and OxaDH5, from Streptomyces albus, were each expressed in Escherichia coli as N-terminal His₆-tagged proteins using codon-optimized synthetic genes, based on consensus PKS domain boundaries (Figures S1–S4). Each of these four recombinant proteins was purified to homogeneity by immobilized Ni²⁺ affinity chromatography. The purity and M_r of each recombinant protein were assessed by SDS-PAGE and confirmed by LC–ESI(+)–MS analysis (Figure S5 and Table S1).

In common with the vast majority of polyketide synthase KR domains, all of which belong to the superfamily of short chain dehydrogenase/reductase (SDR) proteins, ^{21, 22} BonKR2 and OxaKR5 each harbor the highly conserved active site triad of Ser, Tyr, and Lys residues (Figure 2).^{23, 24} On the other hand, both KR domains, along with other KR domains from trans-AT PKSs such as BaeKR9, from module 9 of the bacilllaene PKS, and DifKR6, from module 6 of the difficidin PKS, lack the diagnostic sequence markers, such as the conserved Trp residue (A-Type KR domain) or Leu-Asp-Asp motif (B-Type KR domain), that are normally correlated with the respective (3S)- or (3R) configuration of the resultant 3-hdyroxy acyl thioester reduction product (Figure 2).^23–26 Moreover, none of the additional conserved KR sequence features that are normally diagnostic of the (2R)- or (2S)-methyl configuration of the reduced 2-methyl-3-hydroxyacyl-thioester (KR subtypes 1 or 2) are evident in the amino acid sequences of these trans-AT KR domains. As a consequence, the intrinsic stereospecifity of BonKR2 or OxaKR5 cannot be deduced from simple consensus sequence alignments. (Although the Leu-Val-Asp triad of OxaKR5 might have suggested classification of OxaKR5 as a B-Type KR domain which would be predicted to generate a (3R)-3-hydroxyacyl group, the experiments described below establish firmly that OxaKR5 is an A1-Type KR.)

Figure 2.

Mega3.0 (http://www.megasoftware.net) sequence alignment of PKS ketoreductase domains that reduce 2-methyl-3-ketoacyl-ACP intermediates, showing the four different stereochemical classes of KR domains. Conserved K, S, and Y residues constitute the canonical active site catalytic triad of SDR proteins. The BaeKR9, BonKR2, DifKR6, and OxaKR5 domains lack the conserved sequence motifs diagnostic of established KR Types. PKS source: Amp, amphotericin; Ave, avermectin; Bae, bacillaene; Bon, bongkrekic acid; bor, borrelidin; Con, concanamycin A; Dif, difficidin; Ery, erythromycin; Lan, lankamycin; Lip, lipomycin; Meg, megalomicin; Mei, meilingmycin; Nys, nystatin; Oxa, oxazolomycin; Pic, picromycin; Pla, pladienolide; Tyl, tylactone.

The reductase activities of BonKR2 and OxaKR5 were each confirmed and the steady-state kinetic parameters were determined using the model N-acetylcysteamine thioester substrates (±)-2-methyl-3-ketobutyryl-SNAC (3) and (±)-2-methyl-3-ketopentanoyl-SNAC (4)²⁷ and NADPH (Scheme 1a, Figures S6 and S7, Table S2). The stereochemistry of the resulting (2R,3S)-2-methyl-3-hydroxyacyl-SNAC products 5 and 6 was then established by chiral GC-MS analysis of the derived methyl esters 7-Me and 8-Me, including direct comparison with authentic synthetic standards of all four diastereomers of each product (Figures S8–S11, Tables S3 and S4).²⁸ We also used ACP-bound substrates to confirm the stereospecificity of the BonKR2- and OxaKR5-catalyzed reductions. Thus, in one set of experiments, (2R)-2-methyl-3-ketopentanoyl-EryACP6, generated by incubation of propionyl-SNAC with Ery[KS6][AT6], the ketosynthase-acyltranferase didomain from module 6 of the erythromycin PKS, and EryACP6 plus methylmalonyl-CoA,²⁸ was reduced in separate experiments with BonKR2 or OxaKR5 in the presence of NADPH to generate (2R,3S)-9 (Scheme 1b). After basic hydrolysis and methylation with TMSCHN₂, chiral GC-MS analysis confirmed the exclusive formation of methyl (2R,3S)-2-methyl-3-hydroxypentanoate (8-Me) (Figures S12 and S13, Table S3).

Scheme 1.

Stereochemistry of BonKR2- and OxaKR5-Catalyzed Reduction of 2-Methyl-3-ketoacyl Thioesters.

Finally, in a complementary series of incubations, a mixture of BonKS2, the ketosynthase from module 2 of the bongkrekic acid PKS, and either acetyl-SNAC or propionyl-SNAC, was combined with malonyl-BonACP2, generated in situ as previously described by treatment of apo-BonACP2 with malonyl-CoA and the surfactin pantetheinyl transferase Sfp, so as to yield 3-ketobutyryl-BonACP2 (10) or 3-ketopentanoyl-BonACP2 (11) (Scheme 1c).²⁹ Stereospecific methylation of 10 or 11 was achieved as previously described²⁹ by treatment with a mixture of BonMT2, the C-methyl transferase from module 2 of the bongkrekic acid PKS, and S-adenosyl methionine (SAM), in the presence of S-adenosylhomocysteine (SAH) nucleosidase to prevent potent product inhibition by the co-product SAH, yielding (2R)-2-methyl-3-ketobutyryl-BonACP2 (12) or (2R)-2-methyl-3-ketopentanoyl-BonACP2 (13). In the presence of NADPH, incubation of 12 with either BonKR2 or OxaKR5 gave 14, while reduction of 13 with BonKR2 gave 15. Chiral GC-MS analysis of the derived methyl esters 7-Me and 8-Me, confirmed the exclusive formation of the corresponding (2R,3S)-2-methyl-3-hydroxybutyryl-BonACP2 (14) or (2R,3S)-2-methyl-3-hydroxypentanoyl-BonACP2 (15) (Figures S14–S16, Tables S3 and S4).

Having firmly established the stereochemistry of the BonKR2- and OxaKR5-catalyzed reductions, we next determined the stereospecificity of the paired dehydratase reactions. Chemoenzymatically prepared (2R,3S)-2-methyl-3-hydroxybutyryl-BonACP2 (14) was incubated in separate experiments with BonDH2 or OxaDH5 (Scheme 2a). Hydrolysis of the resultant acyl-ACP thioester (Z)-16, the product of syn dehydration, by treatment with PICS TE, the thioesterase from the picromycin PKS,³⁰ gave exclusively (Z)-2-methylbutenoic acid (17), as established by GC-MS analysis and direct comparison with authentic standards of both (Z)- and (E)-2-methylbutenoic acid (Figures S17 and S18). In a negative control, NigDH1, from module 1 of the nigericin PKS, did not dehydrate 14.³¹ (Although NigDH1 naturally acts only as a 2-methyl-3-ketobutyryl-ACP epimerase, we have shown that it also harbors a cryptic dehydratase activity capable of converting (2R,3R)-2-methyl-3-hydroxypentanoyl-ACP to (E)-2-methyl-2-pentenoyl-ACP.)³¹ BonDH2 and OxaDH5 also catalyzed the reverse reaction, resulting in the stereospecific hydration of chemoenzymatically prepared (Z)-2-methylbutenoyl-BonACP2 (16) to yield exclusively the syn hydration product (2R,3S)-2-methyl-3-hydroxybutyryl-BonACP2 (14) (Scheme 2b). LC-ESI(+)-MS analysis after removal of the DH proteins confirmed the addition of water to (Z)-16, as established by the increase in mass of M+18 of each of the acyl-ACP products (Figure S19, Table S5). Finally, chiral GC-MS analysis established the exclusive formation of the derived methyl (2R,3S)-2-methyl-3-hydroxybutyrate (7-Me) (Figure S20). In the negative control, treatment of (Z)-16 with NigDH1 did not result in the formation of detectable hydration product.

Scheme 2.

Stereochemistry of BonDH2- and OxaDH5-Catalyzed Dehydration/Hydration Reactions

The above-described experiments establish conclusively that the dehydratase domains from the bongkrekic acid and oxazolomycin polyketide synthases, BonDH2 and OxaDH5, each catalyze the syn dehydration of (2R,3S)-2-methyl-3-hydroxyacyl-ACP substrates, which are exclusively generated by their respective paired BonKR2 and OxaKR5 domains, to give the corresponding (Z)-2-methylenoyl-ACP products. These findings constitute the first experimental demonstrations of the DH-catalyzed formation of a Z-trisubstituted double bond, in spite of a number of unsuccessful prior efforts, cited above, that have been directed toward this surprisingly elusive goal. Protein sequence alignments indicate that all of these dehydratases retain a conserved set of active site His and Asp residues but harbor no obvious DH sequence motifs that can be correlated with the E- or Z-configuration of their characteristic disubstituted or trisubstituted enoyl-ACP products (Figure S21).³² All PKS DH-catalyzed dehydration and/or hydration reactions for which the stereochemistry has been determined involve a reversible syn elimination/addition of water,^15–17 in common with the established stereospecificity of the closely related FabA and FabZ dehydratase domains of E. coli fatty acid biosynthesis^{33, 34} as well as the analogous action of the yeast fatty acid synthase.³⁵ The ultimate E or Z enoyl-ACP product geometry must reflect important differences in the active site conformation of the bound 3-hydroxyacyl-ACP substrates that are not obviously correlated with any conserved amino acid sequence motifs.