Molecular Basis for Olefin Rearrangement in the Gephyronic Acid Polyketide Synthase

Abstract

Polyketide synthase (PKS) are a rich source of natural products of varied chemical composition and biological significance. Here, we report the characterization of an atypical dehydratase (DH) domain from the PKS pathway for gephyronic acid, an inhibitor of eukaryotic protein synthesis. Using a library of synthetic substrate mimics, the reaction course, stereospecificity, and tolerance to non-native substrates of GphF DH1 are probed via LC-MS analysis. Taken together, the studies establish GphF DH1 as a dual-function dehydratase/isomerase that installs an odd-to-even double bond and yields a product consistent with the isobutenyl terminus of gephyronic acid. The studies also reveal an unexpected C2 epimerase function in catalytic turnover with the native substrate. A 1.55-Å crystal structure of GphF DH1 guided mutagenesis experiments to elucidate the roles of key amino acids in the multi-step DH1 catalysis, identifying critical functions for leucine and tyrosine side chains. The mutagenesis results were applied to add a secondary isomerase functionality to a non-isomerizing DH in the first successful gain-of-function engineering of a PKS DH. Our studies of GphF DH1 catalysis highlight the versatility of the DH active site and adaptation for a specific catalytic outcome with a specific substrate.

INTRODUCTION

Polyketide synthases (PKS) produce a plethora of bioactive and chemically diverse compounds. The bacterial modular type I PKS are among the most versatile, as each module contains a set of distinct catalytic domains individually responsible for a single step during the elongation and modification of polyketide intermediates. In cis-AT pathways, each module includes a ketosynthase (KS), an acyltransferase (AT), and an acyl carrier protein (ACP) at a minimum, whereas in trans-AT pathways, a separate AT polypeptide delivers a common building block to all modules. A module may also contain one or more modification domains such as ketoreductase (KR), dehydratase (DH), O- methyltransferase (OMT), C-methyltransferase (CMT), and enoylreductase (ER).

The DH domain generates a 2-enoyl thioester intermediate via syn-elimination of a 3-hydroxy group, resulting in the production of alkenes in even-to-odd positions in the final polyketide structure^1–9. DHs are notable for a strict stereospecificity of the 3-hydroxy substrate, which has been correlated with the geometry of the product double bond¹–^3,10. Despite high-resolution crystal structures of several type-I PKS DHs, the structural basis for substrate specificity and product selectivity remains unclear^{4, 11–14}. While “A” and “B” sequence motifs within KR domains predict the stereochemical configuration of the 3-hydroxy group¹⁵, no such motifs have been identified in PKS DHs^{3, 11}. To date, only DHs that act on A-type KR products have been shown to produce cis double bonds³, whereas B-type KR reduction followed by DH dehydration leads to trans double bonds in most cases^{1, 2, 4, 5}, with notable exceptions^{4, 6, 16, 17}

Like other PKS enzymes, DH homologs exist in both type I and type II fatty acid synthase (FAS). The type II FAS of α- and γ-proteobacteria, including Escherichia coli, employ two DHs, FabA and FabZ¹⁸. FabZ catalyzes a general syn-elimination reaction like PKS DHs, acting on a variety of fatty acyl chain lengths to produce an 2-enoyl intermediate that is subsequently reduced and subject to further elongation cycles. FabA is specific for decanoyl substrates, and initially produces a trans double bond ((E)-2-decenoyl, even-to-odd) in a manner analogous to PKS DHs and FabZ. However, FabA also catalyzes a subsequent double-bond migration to produce an odd-to-even olefin in the cis configuration ((Z)-3-decenoyl, Scheme S1)⁷. The final products of type II FAS with FabA include fatty acids with a cis, odd-to-even alkene situated six carbons from the terminal methyl group. An analogous system from Enterococcus faecalis has been characterized in which FabN, a dehydratase more similar to FabZ than FabA, dehydrates and isomerizes a 3- hydroxydecanoyl-ACP substrate to an odd-to-even (Z)-3-decenoyl product¹⁹. FabZ, FabA and PKS DHs possess conserved histidine and aspartate (or glutamate) catalytic amino acids and similar active site structures^{4, 11–14, 20, 21}. The catalytic groups presumably interact identically with the 3-hydroxy substrate in all these enzymes, but the structural basis for their differing activities is unknown.

While many modular PKS pathways include DHs, the products of these pathways rarely contain odd-to-even unsaturated products. Nevertheless, a number of interesting natural products of unknown biosynthetic origin contain odd-to-even double bonds including the tedanolides, candidaspongiolides and myriaporones^22–24. Two distinct mechanisms for generation of odd-to-even unsaturation have been identified in trans-AT PKS. In the related pathways for chloronotil and anthracimycin, the formation of odd-to-even unsaturation has been attributed to multifunctional DHs²⁵. In other pathways, DH-related enoyl isomerases (EIs) generate 3-enoyl intermediates from 2-enoyl substrates^{26, 27}. The best characterized is the EI from module 14 of the bacillaene pathway (PksEI14), which catalyzes isomerization using only a catalytic histidine²⁷, in stark contrast to FabA, where isomerization requires both the catalytic histidine and aspartate²⁸.

Olefin rearrangement has been characterized in only two cis-AT PKS systems^{8, 29}. In the ansamitocin pathway, feeding studies implicated DH-containing module 3 in the formation of a conjugated 2,4-dienoyl product via a proposed vinylogous dehydration²⁹. An interesting DH from the ambruticin pathway catalyzes a novel three-step reaction in which a 3,4-olefin is generated only after epimerization of a 4-methyl substituent⁸. To date, no structures are available for any PKS DH predicted to catalyze both dehydration and isomerization.

Gephyronic acid (1) is a eukaryotic translation inhibitor produced by a cis-AT PKS in the myxobacterium Cystobacter violaceus³⁰. An odd-to-even double bond in 1 is apparently installed by the first extension module of the Gph PKS (Figure 1A). The GphF tri-module polypeptide includes module 1, which extends an isobutyryl starter unit from the loading module to produce an intermediate containing a 3-enoyl (odd-to-even) intermediate (Figure 1A). In addition to the DH, module 1 contains CMT and KR modification domains, which act first and present to the DH a 2-methyl-3-hydroxy substrate³¹. Here we determine that the module 1 DH domain (GphF DH1) is the source of the odd-to- even unsaturation in gephyronic acid. Biochemical characterization and a crystal structure of GphF DH1 establish its role as a bifunctional dehydratase-isomerase and reveal an unexpected C2 epimerization activity. Additionally, we identify tyrosine and leucine amino acids that are key to the isomerase activity.

Figure 1:

A: Domain architecture of GphF, the first gene in the gephyronic acid PKS. Gephyronic acid 1 contains an unusual odd-to-even double bond, apparently installed by module 1. B: Library of N-acetylcysteamine (NAC) substrate mimics used to assess the function of GphF DH1. The library was designed specifically to test the stereospecificity of the DH by sampling each diastereomer of the predicted substrates (2, 3, 4, 5).

RESULTS AND DISCUSSION

Synthesis of Gph1 DH Substrate Candidates

To investigate both the reactivity and specificity of GphF DH1, we systematically examined features of both the native diketide and the potential intermediates crucial for recognition and processing. Four potential 2-methyl-3-hydroxy substrates (2-5), a potential (E)-2-methyl-2-enoyl dehydration product (6), and two potential 2-methyl-3-enoyl isomerization products (rac-7, 8) were synthesized in an effort to explore the effect of stereochemistry on GphF DH function (Figure 1B). To explore the importance of the 2-methyl group, a 3-hydroxy substrate (9), a potential (E)-2-enoyl dehydration product (10), and a potential 3-enoyl isomerization product (11) were synthesized (Fig 1B). Finally, as some DHs can act on (Z) configured alkenes^{3, 10}, (Z)-2-methyl-2-enoyl (12) and (Z)-2-enoyl (13) compounds were synthesized. All thioester analogs were synthesized by coupling the corresponding carboxylic acids to N-acetylcysteamine (NAC) except potential isomerization product rac-7, which was synthesized in an expedient manner via a palladium-catalyzed alkenylation followed by routine conversion to the NAC thioester. Full synthetic procedures and characterization data are described in the supplemental material. All synthetic substrates had distinct retention times in chromatographic separation excepting the 2/3 enantiomers, the 4/5 enantiomers and the E/Z regioisomers 6/12 and 11/13 (Figure S1, Table S1).

Catalytic assay

Amino acids 1697–1984 of GphF were selected as the boundaries of the DH1 domain based on homology to PKS DH domains of known structure^{4, 11–14}, and the excised DH was produced in E. coli. The purified, recombinant GphF DH1 was monomeric in solution with an apparent molecular weight of 29 kDa by size-exclusion chromatography (calculated molecular weight 32.6 kDa), in contrast to other module-embedded DH domains, which are dimeric in solution (Figure S2)^{4, 11–13.} The DH N-terminal sequence, where dimer formation is expected, is not markedly different from that of other PKS DHs, and we predict that the GphF DH1 contributes to the dimer interface in the full module. GphF DH1 was assayed using NAC-linked test substrates 2-13 (Figure 1B). Following removal of protein by methanol precipitation, reaction mixtures were separated by HPLC and analyzed by both absorbance at 225 nm and mass spectrometry (Tables 1–3, Table S2). Quantitation based on 2-13 authentic standards (Figure S1) yielded highly consistent results with A₂₂₅ and LC/MS analysis, but owing to the low ultraviolet absorbance of 2-13, the MS analysis had greater precision. Total ion counts (TIC) of the LC elution profiles of the test substrates and reaction mixes were monitored, and extracted ion chromatograms (EICs) were generated from individual peaks and used to quantitate the extent of catalysis (Tables 1–3, S1). GphF DH1 catalyzed reactions with all compounds except 5 and 8.

Table 1:

Activity of Wild-Type GphF DH1

	Percent distribution of substrate & products* (cmpd)
Starting material	3-hydroxy	2-enoyl	3-enoyl
2	71.6 ± 0.5 (2)	27.7 ± 0.5 (6)	0.6 ± 0.1 (14)
3	81.6 ± 0.4 (3)	18.0 ± 0.3 (6)	0.3 ± 0.1 (14)
4	99.4 ± 0.1 (4)	0.5 ± 0.1 (6)	< 0.01 (14)
5	99.9 ± 0.1 (5)	< 0.01 (6)	< 0.01 (14)
6	1.3 ± 0.1 (2)	89.4 ± 0.1 (6)	9.3 ± 0.1 (14)
rac-7	< 0.01 (2)	12.9 ± 0.2 (6)	87.0 ± 0.2 (rac-7) (37 + 50) (14 + 8)
8	< 0.01 (2)	< 0.01 (6)	99.9 ± 0.1 (8)
9	87.5 ± 0.2 (9)	5.4 ± 0.1 (10)	7.1 ± 0.1 (11)
10 & 11 (2:1)	5.9 ±0.3 (9)	38 ± 2 (10)	56 ± 2 (11)
12	0.3 ±0.1 (2)	96.6 ± 0.1 (6 & 12)	3.1 ± 0.1 (14)
13	6.3 ±0.2 (9)	37.0 ± 0.4 (10 & 13)	56.7 ± 0.6 (11)

* Quantitation of substrate/product distribution is based on LC/MS separation and extracted ion counts for each species, as described in Methods. The starting substrate concentration in all assays was 200 μM.

Table 3:

Activity of GphF DH1/Y1856F

	Percent distribution of substrate & products (cmpd)
Starting material	3-hydroxy	2-enoyl	3-enoyl
2	83.7 ± 0.6 (2)	16.2 ± 0.7 (6)	< 0.01 (14)
3	80.5 ± 0.4 (3)	19.5 ± 0.4 (6)	< 0.01 (14)
4	99.4 ± 0.1 (4)	0.6 ± 0.1 (6)	< 0.01 (14)
5	99.9 ± 0.1 (5)	< 0.01 (6)	< 0.01 (14)
6	2.3 ± 0.3 (2)	97.7 ± 0.3 (6)	< 0.01 (14)
rac-7	< 0.01 (2)	< 0.01 (6)	99.9 ± 0.1 (rac-7)
8	< 0.01 (2)	< 0.01 (6)	99.9 ± 0.1 (8)
9	61.4 ± 0.6 (9)	31.9 ± 0.5 (10)	6.6 ± 0.3 (11)
10 & 11 (2:1)	31.1 ± 0.3 (9)	38.5 ± 0.4 (10)	30.4 ± 0.2 (11)
12	0.6 ± 0.1 (2)	99.3 ± 0.1 (6 & 12)	< 0.01 (14)
13	33.8 ± 0.1 (9)	39.9 ± 0.2 (10 & 13)	26.3 ± 0.2 (11)

Dehydratase Activity

Based on the biosynthetic pathway deduced from the gph gene cluster³⁰ and on the established intermediates and products of homologous PKS modules, the GphF DH1 natural substrate is predicted to be (2R,3R)-3-hydroxy-2,4-dimethylpentanoyl-ACP, as the GphF CMT has been shown to act before the GphF KR1³¹, and the “B-type” KR1 should produce an (R)-3-hydroxy product¹⁵. PKS DHs are expected to catalyze syn-elimination^{1, 2, 32}, as could occur with the (2R,3R)-2-methyl-3-hydroxy substrate 2 or its enantiomer 3 to yield the (E)-2-methyl-2-enoyl product 6. The prediction is based on PKS DH similarity to FAS DHs, where syn-elimination has been demonstrated^{33, 34}, and on the activity of DH domains with 2-methyl substituents^{4, 32}. Consistent with the predicted substrate stereochemical configuration, recent studies on the stereospecificity of CMT domains from several trans-AT PKS pathways found that CMT domains installed (R)-2-methyl groups³⁵. However, based on the structure of 1, the methyl group installed by GphF MT1 is predicted to be in the (S)-configuration³⁰. Thus, we investigated GphF DH1 reactivity with NAC-linked thioesters of all four diastereomers of the predicted natural substrate (2, 3, 4, 5) (Figure S3, Table 1). Dehydrated products were detected for the enantiomers 2 and 3, which are competent for production of an (E)-configured dehydrated product via syn-elimination. Greatest activity was detected with 2 (28% conversion to 6), whereas 3 was less reactive (18% conversion to 6) under our assay conditions. For the other test substrates, where syn-elimination would yield the (Z)-2-methyl-2-enoyl product 12, only trace amounts of dehydrated product were detected with 4, and none with 5. This indicates that the natural substrate for GphF DH1 is likely (2R,3R)-3-hydroxy-2,4-dimethylpentanoyl-ACP.

Isomerase Activity

Reaction of 2 or 3 with GphF DH1 resulted in two dehydrated products with retention times of 4.47 min for the major product, and 4.35 min for the minor product (Figure S3, Table 1). The masses and elution times of these products exactly match those of authentic 6 and rac-7 prepared independently (Figure S4). Thus, GphF DH1 possesses both dehydration and isomerization activities, and dehydration of a 2-hydroxy substrate may be followed by isomerization of the resulting 2-enoyl 6 to 3-enoyl rac-7 (8 or 14), similar to the catalytic route of FabA (Scheme S1). The isomerase activity was easily detected, but less 3-enoyl than 2-enoyl product accumulated under our assay conditions. In the reaction of GphF DH1 with 4, we detected only trace amounts of the 2-enoyl product (retention time 4.47 min, Table S1) and the low turnover may have precluded detection of any 3-enoyl product.

When exposed to an (E)-2-enoyl NAC substrate, the E. coli FAS DH FabA produced an equilibrium mixture of 68% 3-hydroxy, 29% (E)-2-enoyl and 3% (Z)-3-enoyl products⁷. In an analogous manner, we used 6 and rac-7 to assess the mixture of dehydration and isomerization products of GphF DH1 (Figure S5, Table 1). Incubation with 6 resulted in 9% conversion to the 2-methyl-3-enoyl 14 or 8 (retention time 4.35 min), but only 1% re-hydration to substrate 2 or 3 (retention time 3.78 min). Similarly, GphF DH1 isomerized 13% of rac-7 to 6, but 2/3 was not detected. These results provide additional support for a sequential mechanism of dehydration followed by isomerization and demonstrate that the reaction equilibrium strongly favors the dehydration product. Additionally, the enzyme must overcome a significant thermodynamic barrier in isomerizing the 2,3-unsaturated intermediate, which is conjugated with the thioester, to the unconjugated 3-enoyl product.

Specificity at C2 and Epimerase Activity

We used the (R)-configured 8 to assess GphF DH1 stereospecificity at C2 (Figure S4, Table 1). No products were detected in reactions of GphF DH1 with 8 (Figures S5, S6, Table 1), indicating that isomerization of 6 yielded the (S)-2-methyl product 14. This was confirmed by chiral-LC/MS analysis (Figure S6, Table S1). The rac-7 standard was partially resolved into a species with an identical retention time to 8 (10.21 min) and a second species (10.41 min) corresponding to its enantiomer 14 (Figure S6, Table S1). When incubated with 6, GphF DH1 generated a product whose retention time matched that of 14, and co-injection of 8 with the 6 + GphF DH1 reaction mixture clearly demonstrated that the isomerization product was 14 and not 8 (Figure S6, Table S1). These results confirm a C2 epimerase function in the overall transformation of 2 to 14. The effective epimerization requires either different catalytic residues for dehydration and epimerization, or substrate rearrangement in the active site to position 6 with the pro-R face of C2 towards the presumed proton donor (His1735).

We also tested the tolerance of GphF DH1 for non-methylated substrates using 9, which resulted in products with retention times of 4.23 min and 4.13 min, identical to those of authentic standards for dehydration product 10 and isomerization product 11, respectively, which were available as a 2:1 mixture (Figures S1, S3, S5, Table 1). Thus, the dual-functionality of GphF DH1 is not strictly dependent on the presence of a 2-methyl substituent. Nevertheless, GphF DH1 preferred 2-methyl substrates: 9 was a poorer dehydration substrate than were 2 and 3 (13% dehydration for 9 vs. 28% for 2 and 18% for 3, Table 1). This is consistent with the demonstration that 2-methylation by GphF CMT precedes KR1 reduction³¹. While lower levels of dehydrated intermediate 10 accumulated with non-methylated substrate 9 (5% 10, compared to 28% 6 from the methylated 2), the isomerization was more effective, as 9 yielded significant quantities of the 3-enoyl product 11 (7%, compared to <1% 14 from the methylated 2 or 3). Additionally, GphF DH1 converted the 67%:33% mixture of 10:11 to a 38%:56%:6% mixture of 10:11:9 (Figure S5, Table 1).

E vs. Z Selectivity

To further explore the relaxed substrate selectivity, we also evaluated the tolerance of GphF DH1 for (Z)-2-enoyl substrates 12 and 13 (Table 1, Figure S5). The (Z)-configured 13 was converted to a mixture of isomerization product 10 (57%) and hydrated product 9 (6%). In contrast, GphF DH1 had very low activity with the (Z)-2-methyl substrate 12 (3% isomerized to 7, <1% hydrated to 2/3), illustrating again that a 2-methyl substituent is detrimental to isomerization and effectively blocks hydration of 2-enoyl substrates. Activity with a (Z)-configured substrate is additional evidence for the relaxed substrate stereospecificity and intermediate regiospecificity of GphF DH1, although it is unlikely that the activity observed here represents a biologically relevant function, as neither 12 nor 13 is predicted to be a native substrate or intermediate³⁰. Some PKS DHs are predicted to produce a (Z)-2-enoyl products, and this has been demonstrated in one case³.

GphF DH1 Crystal Structure

Purified, recombinant GphF DH1 yielded single crystals that diffracted to 1.55 Å, but they were difficult to reproduce and grew slowly (Table S3, Figure 2A). A P1711L variant had identical activity to the wild type DH (Figure S7) and resulted in reproducible growth of crystals of a different form that diffracted to 1.85 Å (Table S3). GphF DH1 has the canonical double-hotdog fold of PKS DHs (Cα superposition results in RMSD values of 1.0 – 2.1 Å, Figure 2B)^{4, 11–14}, but it does not form a dimer in crystals, in agreement with the monomeric state in solution (Figure S2). The active site, identified by the catalytic His1735 and Asp1898 side chains, is at the start of a hydrophobic substrate pocket (~7 Å wide and ~14 Å long) that extends from the catalytic His-Asp dyad along the central hotdog helix, and has a bi-lobed shape that could accommodate the substrate terminal methyl groups. The GphF DH1 pocket is shorter than the pockets of the curacin and erythromycin DHs, which have longer substrates^{11, 12} (Figure S8).

Figure 2:

GphF DH1 structure. A: Overall structure. B: Superposition of GphF DH1 (green), PksEI14 (cyan), and several non-isomerizing PKS DHs (DEBS yellow, Rif tan, CurK red, CurH white, CurJ purple, AmbDH3 light green). The catalytic His of GphF DH1 and Pks EI is further from the catalytic Asp than in the non-isomerizing DHs, correlated with a branched-chain amino acid in lieu of Pro. Protrusion of the Leu or Val into the active site also influences the Asp position. C. Superposition of WT GphF DH1 (Green) with L1744P GphF DH1 (Cyan). A modest shift in the position of the catalytic histidine is observed in the L1744P variant, reducing the distance between the catalytic His and Asp to 5.7 Å from 6.4 Å. Although this difference appears minor from a structural perspective, the L1744P variant is inactive as an isomerase, while retaining dehydration activity. D. The active site architecture of PksEI with relevant amino acids shown as sticks. The active site of PKSEI is very similar to that of GphF DH1. E. Active site architecture of the CurK DH with relevant amino acids shown as sticks. CurK DH is representative of a typical monofunctional PKS DH. The catalytic His and Asp are much closer than in GphF DH1, and the distance between the catalytic Asp and a conserved Tyr is well outside of hydrogen-bond length.

Superposition of active site structures revealed distinct differences between PKS enzymes capable of isomerization (GphF DH1 and PksEI14²⁷) and the non-isomerizing DHs (Figure 2B)^{4, 11–13}. Like the FAS DHs, all these enzymes have a hydrogen bond from the catalytic His-Nδ to a backbone carbonyl (Val1742), making the His-Ns atom available for acid-base chemistry. The most striking difference is a nearly 2 Å greater separation of the catalytic His and Asp functional groups in the isomerizing enzymes than in non-isomerizing PKS DHs^{4, 11–13}, and 1.3 Å longer than in FabA²⁰. The extended distance may be due to a branched-chain amino acid near the catalytic His (GphF DH1 Leu1744, PksEI14 Val27, Figures 2C, 2D). As noted previously²⁷, a branched amino acid exists at this position in most of the presumed isomerizing enzymes, whereas it is proline in non-isomerizing PKS DHs (Figure S9). To investigate the importance of Leu1744 to the active site structure, we solved a crystal structure of GphF DH1/L1744P (Table S3, Figure 2C). In this structure, the catalytic His1735 Nε is 0.7 Å closer to Asp1898 than in the wild-type DH, but still further than in non-isomerizing DHs (Figure 2E). Thus, the bulkier Leu side chain does not solely account for the longer distance between the catalytic side chains of GphF DH1.

A second striking difference of GphF DH1 to the non-isomerizing DHs is a hydrogen bond from the conserved Tyr1856 hydroxy to the catalytic Asp1898 carboxylate, a unique feature of GphF DH1 (Figures 2C, 2E). The hydroxy-carboxylate contact does not occur in the non-isomerizing DHs due to the position of the Asp side chain, which in GphF DH1 is affected by the branched side chain at Leu1744. An analogous hydrogen bond also does not exist in the PksEI14 where the Asp is replaced with Asn and the Tyr with Trp (Figures 2B, 2D, S9)²⁷ The Tyr1856 - Asp1898 hydrogen bond is especially interesting in light of the unusual GphF DH1 chemistry to effect an epimerization during sequential dehydration and isomerization reactions.

Assessment of Dual-Enzyme Functionality

We first tested the role of the His-Asp dyad in the dual DH activities. When the catalytic His1735 was substituted with Gln, no products were detected with 2, 6, 9, or 10 & 11, and substitution of the catalytic Asp1898 with Asn was similarly unreactive with 2 or 6. Thus, like FabA, both His1735 and Asp1898 are required for each of the dehydration and isomerization reactions (Figure S10)²⁸. In contrast, the isomerase activity of the PksEI14 required only the catalytic His²⁷.

We next investigated the role of the branched-chain amino acid near His1735. GphF DH1/L1744P had substantially increased dehydration activity relative to wild type DH1 (Table 2), but it lacked detectable isomerase activity with any of the 2-methyl substrates (2, 3, 6, rac-7, 12; Figure S11; Table 2). In contrast, isomerase activity was retained with the substrates lacking a 2-methyl substituent (9, 10, 11, 13). Thus, the effect of the branched side chain in the DH1 active site is specific to 2-methyl substrates. Additionally, DH1/L1744P had greater rehydration activity with 6 (13% vs. 1%) and with the 10/11 mixture (35% vs. 6%) than did the wild type. The increase in dehydration and rehydration with DH1/L1744P indicates that the branched side chain at position 1744 enables isomerization at the cost of reduced dehydration and rehydration activities.

Table 2:

Activity of GphF DH1/L1744P

	Percent distribution of substrate & products (cmpd)
Starting material	3-hydroxy	2-enoyl	3-enoyl
2	51.3 ± 0.4 (2)	48.7 ± 0.4 (6)	< 0.01 (14)
3	50.3 ± 0.6 (3)	49.6 ± 0.5 (6)	< 0.01 (14)
4	99.3 ± 0.1 (4)	0.6 ± 0.1 (6)	< 0.01 (14)
5	99.9 ± 0.1 (5)	< 0.01 (6)	< 0.01 (14)
6	12.9 ± 0.2 (2)	87.1 ± 0.2 (6)	< 0.01 (14)
rac-7	< 0.01 (2)	< 0.01 (6)	99.9 ± 0.1 (rac-7)
8	< 0.01 (2)	< 0.01 (6)	99.9 ± 0.1 (8)
9	58.9 ± 0.1 (9)	39.1 ± 0.5 (10)	2.0± 0.1 (11)
10 & 11 (2:1)	35.1 ± 0.2 (9)	35.8 ± 0.8 (10)	29.1 ± 0.5 (11)
12	4.3 ± 0.1 (2)	95.6 ± 0.1 (6 & 12)	< 0.01 (14)
13	39.7 ± 0.6 (9)	40.7 ± 0.5 (10 & 13)	19.6 ± 0.2 (11)

To investigate the role of Tyr1856, we generated a Phe substitution. Tyr is conserved at this position in most, but not all, PKS DHs. However, uniquely in GphF DH1, the side chain hydroxy is hydrogen bonded to the catalytic Asp1898 carboxylate (Figures 2C, 2E, S9, Table 3)^{4, 11, 12, 27}. GphF DH1/Y1856F catalyzed dehydration at a comparable rate to the wild type but lacked any detectable isomerase activity with the 2-methyl substrates (2, 3, 6, rac-7, 12; Figure S12; Table 3). However, with substrates lacking a 2-methyl substituent (9, 10, 11, 13), GphF DH1/Y1856F retained isomerase activity. Thus, while Asp1898 is critical to both dehydratase and isomerase activity, Tyr1856 plays an essential role in isomerization of 2-methyl substrates.

Multifunctional PKS Dehydratases

Here, we describe the formation of an odd-to-even double bond by the GphF DH1 through dual dehydratase and isomerase activities. The (S)-2,4-dimethyl-3-pentenoyl thioester product 14 is consistent with the structure of gephyronic acid 1, the natural product of the Gph PKS pathway³⁰. GphF DH1 had dehydratase activity only with 2-hydroxy compounds capable of syn-elimination to yield an (E)-2-enoyl product (2, 3 and 9), in agreement with previously characterized DHs. DH1 exhibited relaxed substrate selectivity for both the dehydration and isomerization reactions. It dehydrated both 2-methyl-3-hydroxy enantiomers 2 and 3 as well as the non-methylated 9. DH1 was an isomerase for substrates with and without a 2-methyl substituent and even acted on (Z)-2-enoyl substrates. The FabA of bacterial fatty acid synthesis also catalyzes sequential dehydration and isomerization reactions, but it is specific for a 3-hydroxydecanoyl substrate and forms a (Z)-3-decenoyl product⁷.

The crystal structure of GphF DH1, the first for an isomerizing PKS DH, offers clues to the basis for the dual catalytic activities. The active site His is assumed to be the sole catalytic base in all DH family members (FabA dual dehydratase/isomerase, PksEI14 and non-isomerizing DHs). As in other DHs, His1735 and Asp1898 together catalyze syn-dehydration and therefore must be on the same face of the natural (2R,3R)-2-methyl-3-hydroxy substrate 2. The His-Asp catalytic dyad and Tyr1856 are all required for isomerization of the (E)-2-methyl-2-enoyl dehydration product 6. Thus, we propose that Asp1898 activates Tyr1856 to remove a C4 proton, creating an dienolate intermediate, and that His1735 re-protonates C2 to form the (S)-2-methyl-3-enoyl product (enantiomer of 8) (Scheme 1). Tyr1856 is conserved in most PKS DHs (Figure S9). Among those of known structure, only GphF DH1 Tyr1856 forms a hydrogen bond with catalytic Asp1898 and only GphF DH1 catalyzes isomerization^{4, 11–14}

Scheme 1:

Proposed mechanism for GphFDH1 activity on substrates containing a 2-methyl group.

After dehydration, Asp1898 activates Tyr1856, which abstracts a proton from C4, generating a dienolate intermediate. His1735 then reprotonates C2, setting the methyl in the (S) configuration.

Despite the relaxed substrate selectivity of both the dehydratase and isomerase reactions, the mutagenesis data revealed a striking divergence of isomerase mechanism between substrates with a 2-methyl substituent (2, 3, 6, and rac-7) and those without (9, 10/11). Whereas both the branched side chain of Leu1744 and the hydroxy group of Tyr1856 were essential to isomerase activity with any 2-methyl substrate, neither the branched side chain nor the Tyr hydroxy was essential to isomerization of substrates lacking a 2-methyl. The Leu1744 side chain protrudes into the active site further than does a Pro side chain (Figures 2B, 2C), and thus imposes greater conformational and positional constraints on binding of (E)-2-methyl-2-enoyl 6 than on the binding of (E)-2-enoyl 10. With the 2-methyl substrate, these constraints are evidently needed to position the C4 atom appropriately for abstraction of a proton during the isomerization reaction. In contrast, the isomerase activity of GphF DH1/Y1856F with non-methylated 10 may be due to C4 deprotonation by Asp1898, directly or through a water molecule.

Epimerization at C2

We made the unexpected discovery that GphF DH1 acts as a 2-methyl epimerase through sequential dehydration and isomerization reactions. The preferred (and anticipated natural) substrate was (2R,3R)-2-methyl-3-hydroxy 2, although both 1 and the odd-to-even olefin product 14 have an (S)-methyl at this position. Epimerization requires removal of the (2S)-proton in the dehydration step and addition of a (2R)-proton in the isomerization step, implying substantial re-positioning of the substrate in the active site. While GphF DH1 epimerizes the C2 position of its substrate, FabA catalysis involves only the pro 2S proton³³. Despite the similar overall structures, many details of the GphF DH1 and FabA active sites differ, which may explain the differences in the stereochemical course of reaction catalyzed by the two DHs. Notably, the analog of Tyr1856 is Pro in FabA, as is the analog of Leu1744.

We note that several non-isomerizing PKS DH homologs can function as C2 epimerases for 2-methyl-3-keto compounds, which are not dehydration substrates³⁶. These enzymes and GphF DH1 have the common problem of deprotonating and reprotonating opposite faces of the substrate C2 position. In all cases where epimerase activity has been reported, the DHs also have a Tyr at the position corresponding to Tyr1856 in GphF DH1. While these DH homologs do not catalyze isomerization, they may use both His and Tyr for epimerization. In contrast, for the PksEI14, where the Asp1898 analog is Asn and the Tyr1856 analog is Trp, the acidity of the C4 proton is enhanced by a double bond at C5. No base was identified for deprotonation of the substrate C4 and it was concluded that the catalytic His is responsible for all proton transfers²⁷.

Of the cis-AT DHs predicted to produce odd-to-even unsaturated final products, DH4 from the ambruticin pathway catalyzes sequential dehydration, epimerization and isomerization reactions on a (3S,4S)-3-hydroxy-4-methyl substrate in which epimerization of the 4-methyl occurs after dehydration to a 4-methyl-2-enoyl intermediate and before double bond shift to the 4-methyl-3-enoyl product (Scheme S1)⁸. The Amb DH4 is most similar to non-isomerizing DHs with His, Asp, Pro and Tyr amino acids in the active site (Figure S9). It is tempting to speculate that the His and Tyr may be involved in the epimerization step.

Isomerase activity and pathway throughput

In both GphF DH1 and FabA, the isomerization reaction is thermodynamically unfavorable as it shifts a double bond out of resonance with the thioester carbonyl. In biochemical assays, both enzymes produce low levels of the 3-enoyl isomerization product⁷ (Table 1). While the level of the (Z)-3-decenoyl final product from FabA (3%) appears too low to support the critical biological requirement for unsaturated fatty acid biosynthesis, E. coli drives the equilibrium in the forward direction by co-expression of fabA with fabB, a gene encoding a specialized KS for elongation of the (Z)-3-decenoyl product³⁷. We anticipate that the KS of GphF module 2 drives gephyronic acid biosynthesis in a similar manner. In contrast to GphF DH1 and FabA, the PksEI14 isomerization is more thermodynamically favorable, as the double bond shift forms a conjugated triene with double bonds at C5 and C7.

Rational Engineering in PKS DHs

We propose that GphF DH1 had a non-isomerizing DH ancestor that may have acted on a 2-methyl substrate. Introduction of a branched side chain in the active site led to selective 2-methyl epimerization relative to the (2R,3R)-2-methyl-3-hydroxy product of the module KR. The presence of a branched-chain amino acid in the PKS DH active site is correlated with a capability for double-bond migration²⁷, although the Amb DH4 is an exception⁸.

To investigate the hypothesis that the His-proximal branched-chain amino acid is essential for isomerization activity, we created an analogous P1005V variant of the non-isomerizing CurK DH. CurK DH is a typical non-isomerizing DH, with His, Asp, Pro, and Tyr sidechains in the active site (Fig 2B). The natural CurK DH substrate has a 4-methyl substituent but lacks a 2-methyl, so 9 was used as a surrogate substrate. The wild-type CurK DH readily dehydrated 9, producing only the 2-enoyl species 10. CurK DH/P1005V also dehydrated 9, but produced both the 2-enoyl species 10, and the 3-enoyl 11 (Figure S13, Table 4). This result further emphasizes the importance of the branched sidechain to substrate positioning for isomerization. Similar to FabA and GphF DH1, the isomerization activity was low in the assay with a purified DH in absence of downstream enzymes to drive pathway throughput. Nevertheless, this result may represent a stepping-stone for PKS engineering. Attempts to engineer isomerization activity in bacterial FAS DHs required swapping large domains between FabZ and FabN (a FabZ homolog capable of isomerization)³⁸. In contrast, the PKS DHs evolved a secondary isomerase function in a simpler manner, but at some expense to dehydration efficiency. If the secondary isomerase activity is to be developed for chemoenzymatic synthesis, then a scheme such as a capture enzyme may be needed to drive an unfavorable reaction forward.

Table 4.

CurK DH Activity

	Percent distribution of substrate & products (cmpd)
Starting material	3-hydroxy	2-enoyl	3-enoyl
Wild type
9	49.1 ± 0.2 (9)	50.9 ± 0.2 (10)	< 0.01 (11)
10 & 11 (2:1)	34.8 ± 0.3 (9)	34.9 ± 0.1 (10)	30.3 ± 0.3 (11)
P1005V
9	60.2 ± 0.3 (9)	38.4 ± 0.2 (10)	1.4 ± 0.2 (11)
10 & 11 (2:1)	34.9 ± 0.3 (9)	36.9 ± 0.5 (10)	28.2 ± 0.7 (11)

In summary, GphF DH1 is a novel addition to the versatility of the DH and DH-like domains in PKS systems, expanding the repertoire of reactions catalyzed by these fascinating enzymes. The biochemical characterization and crystal structure together shed light on the molecular requirements for secondary function in PKS DHs. Both the substrate structure and the arrangement of Leu1744 and Tyr1856 adjacent to the His1735/Asp1898 catalytic dyad control isomerization subsequent to dehydration. The results of these experiments facilitated rational engineering of an isomerizing variant of the CurK DH, representing a step towards a more nuanced approach to PKS engineering. As the overall fold and general catalytic machinery of PKS DHs are highly conserved, we anticipate that additional secondary functions will be driven, in an analogous manner to GphF DH1, via subtle alterations to the active site that are specific to the natural DH substrate, as for the dehydratase-cyclase Amb DH3¹⁴.