Discovery and Engineering of Pathways for Production of α-Branched

Article pubs.acs.org/JACS

Cite This: J. Am. Chem. Soc. 2017, 139, 14526-14532

Discovery and Engineering of Pathways for Production of α‑Branched Organic Acids Michael R. Blaisse,† Hongjun Dong,† Beverly Fu,† and Michelle C. Y. Chang*,†,‡ †

Department of Chemistry, University of California, Berkeley, Berkeley, California 94720-1460, United States Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, California 94720-1460, United States

‡

S Supporting Information *

ABSTRACT: Cell-based synthesis offers many opportunities for preparing small molecules from simple renewable carbon sources by telescoping multiple reactions into a single fermentation step. One challenge in this area is the development of enzymatic carbon−carbon bond forming cycles that enable a modular disconnection of a target structure into cellular building blocks. In this regard, synthetic pathways based on thiolase enzymes to catalyze the initial carbon−carbon bond forming step between acyl coenzyme A (CoA) substrates offer a versatile route for biological synthesis, but the substrate diversity of such pathways is currently limited. In this report, we describe the identification and biochemical characterization of a thiolase-ketoreductase pair involved in production of branched acids in the roundworm, Ascaris suum, that demonstrates selectivity for forming products with an αmethyl branch using a propionyl-CoA extender unit. Engineering synthetic pathways for production of α-methyl acids in Escherichia coli using these enzymes allows the construction of microbial strains that produce either chiral 2-methyl-3-hydroxy acids (1.1 ± 0.2 g L−1) or branched enoic acids (1.12 ± 0.06 g L−1) in the presence of a dehydratase at 44% and 87% yield of fed propionate, respectively. In vitro characterization along with in vivo analysis indicates that the ketoreductase is the key driver for selectivity, forming predominantly α-branched products even when paired with a thiolase that highly prefers unbranched linear products. Our results expand the utility of thiolase-based pathways and provide biosynthetic access to α-branched compounds as precursors for polymers and other chemicals.

■

INTRODUCTION Living systems offer unique advantages as multistep catalysts for chemical synthesis. Indeed, cellular chemistry has evolved to take simple carbon sources, such as glucose and carbon dioxide, as universal building blocks to create an enormous array of metabolites and natural products, enabling the industrial-scale production of a broad range of small molecule targets in a single fermentation step from renewable carbon sources.1−7 Surprisingly, cell metabolism depends mainly on just two related carbon−carbon bond forming reactions, the aldol and Claisen condensations. The Claisen condensation catalyzed by thiolase enzymes is particularly versatile, allowing the coupling of two acyl coenzyme A (CoA) thioesters (Figure 1).8,9 This step serves as an initiation point into a reaction cycle of C−C bond formation, followed by ketoreduction, dehydration, and enoyl reduction, which can be used in principle to disconnect any small molecule target. Despite the large number of members in the thiolase superfamily (>50 000), a major challenge in their utilization for this purpose is that the predominant substrate selectivity characterized has been the condensation of simple acetyl-CoA monomers (R = H).10 Broader substrate selectivity has been identified based on the participation of thiolases in diverse metabolic pathways, but the relatively small number of characterized thiolases still limits the construction of synthetic © 2017 American Chemical Society

metabolic pathways. More recently, the promiscuity of thiolases has been used to expand the scope of ω- and αfunctionalization of target compounds.11−13 Notably, the incorporation of propionyl-CoA in a retro-degradative pathway has been demonstrated to produce the α-substituted enoic acid, tiglic acid.11 However, an efficient pathway for 2-methyl-3hydroxy acid production has not yet been reported despite the important role that α-substituted hydroxy acids could serve as new monomers for the production of biodegradable polymers. Polyesters derived from 3-hydroxy acids in the poly(hydroxybutyrate) (PHB) family suffer from problems of processability and instability to thermal decomposition. Incorporation of an α-branched monomer, 2-methyl-3-hydroxybutyrate, into PHB copolymers results in decreased melting enthalpy without significantly lowering melting temperature (Tm) and more facile crystallization related to the decreased cold crystallization temperature (Tcc).14 These effects were not observed in the other PHB-based copolymers tested. Furthermore, a major pathway for thermal degradation of PHB-type polymers proceeds through α-deprotonation and elimination, which is reduced by α-functionalization.15 Received: July 16, 2017 Published: October 9, 2017 14526

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society

Figure 1. Thiolase-based reaction cycle for assembly of small molecule targets. Both decarboxylative and nondecarboxylative Claisen condensations are involved in the biosynthesis of a broad range of small molecule targets, including fatty acids, isoprenoids, and polyketides. However, the predominant substrate selectivity of these enzymes involves utilizing the unfunctionalized monomers (R = H), acetyl-CoA and malonyl-CoA (left). Utilizing the nondecarboxylative Claisen condensation catalyzed by thiolases, a reaction cycle of condensation followed by ketoreduction, dehydration, and reduction could be used in principle to disconnect any small molecule as long as the R groups of interest are accommodated (right).

Figure 2. Characterization of thiolases from A. suum. (A) Activity screen of Acat1−5 thiolases and a canonical acetyl-CoA selective PhaA from C. necator. Activity assayed by CoA release in the presence of either 1 mM acetyl- or propionyl-CoA using DTNB. Data are mean ± s.e. (n = 3). (B) Structures of 3-oxoacyl-CoA substrates. (C) Steady-state kinetic characterization of Acat2,3,5 thiolases in the thiolysis or degradative direction monitoring formation of the thioester bond in the presence of 200 μM CoA. In cases where saturation was not reached, the kcat/KM parameter was fit directly and a lower bound is provided for individual kcat and KM values. Data are mean ± s.e. (n = 3).

accumulate short branched organic acids as part of an anaerobic respiration pathway, providing an unusual example of a biosynthetic thiolase-based pathway that condenses both acetate and propionate units.16,17 We have identified thiolase, ketoreductase, and dehydratase candidates, finding two thiolases that prefer propionyl-CoA as well as a ketoreductase

In this context, we present the discovery and characterization of a new thiolase and ketoreductase pair from the roundworm, Ascaris suum that is selective for biosynthesis of α-branched acids. We also report the use of these enzymes to engineer pathways for the production of α-methyl branched 3-hydroxy and enoic acids in Escherichia coli. A. suum has been reported to 14527

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society

Figure 3. Characterization of the AsHadh ketoreductase. (A) Steady-state kinetic characterization of AsHadh reduction of various 3-oxoacyl-CoA substrates as well as the oxidation of 3S- and 3R-hydroxybutyryl-CoA produced in situ. Data are mean ± s.e. (n = 3). (B) Assay for examining the stereochemical preference of AsHadh.

whose preference for an α-methyl substituent serves as an important selectivity filter. We have used these enzymes to produce α-methyl acids in E. coli at 87% and 44% conversion of propionate for the enoic and 3-hydroxy acids, respectively, to provide a useful route to chemical precursors and chiral building blocks.18

thiolytic directions. Its function remains unknown, but a Ser residue is found in place of the essential Cys that is conserved in all other members of the thiolase superfamily26 (Supporting Information Figure S1). These enzymes were also screened against the four-carbon butyryl-CoA, but no activity was detected with the longer substrate with any of the Acat thiolases. With this simple screen of Acat1−5 activity in hand, we decided to carry out a more detailed characterization of the substrate selectivity of Acat2 and 3. To carry out this experiment, we chemically synthesized a set of 3-oxoacyl-CoA substrates to examine thiolysis since the CoA release assay is not adequate for characterizing mixed substrate condensation in the synthetic direction (Figure 2B, Supporting Information Figures S4−S11). The catalytic efficiencies of Acat2 and Acat3 with all substrates formed from combinations of 2- or 3-carbon units were of similar magnitude, but individual kcat and KM values were both higher for α-methyl substrates compared to their linear counterpart (Figure 2C). This behavior is similar to that observed for human T2 thiolase, although the observed KM values for Acat 2 and 3 were significantly higher than that observed in T2.21 These higher KM values may reflect a preference to release 3-oxoacyl-CoA product since biosynthesis is expected to be the physiologically relevant direction in A. suum mitochondria. Some substrates showed substrate inhibition behavior at higher concentrations. Initial rate data for these substrates fit poorly to the standard Michaelis−Menten equation but fit well to the modified form (Supporting Information Figure S12). Substrate inhibition was most prominent with the six-carbon chain substrates, 3-oxohexanoyl-CoA and 2-methyl-3-oxohexanoyl-CoA, with which Acat2, 3, and 5 also demonstrated lower catalytic efficiencies. Acat5 was characterized to compare a canonical thiolase from the same host organism. In contrast to Acat2 and Acat3, Acat5 had high catalytic efficiency with linear 3-oxoacyl-CoA substrates but a catalytic defect of 100-fold in kcat/KM upon introduction of an α-methyl substituent to 3-oxobutyryl-CoA, mostly due to a large increase in KM (Figure 2C). Extending the main chain by one carbon (2-methyl-3-oxopentanoyl-CoA) lowered the relative kcat/KM by an additional 30-fold. Notably, when comparing the preference between the linear and corresponding branched substrate, Acat5 displayed a preference of two to three orders of magnitude for the linear analogue. This observation is quite different from Acat2 and 3, which displayed only a 1.5 to 4.8-fold deviation between the kcat/KM values for the linear and branched analogues.

■

RESULTS AND DISCUSSION Identification and Characterization of Thiolase Enzymes from A. suum. We first set out to identify gene candidates responsible for the branched acid production reported in A. suum from the available Transcriptome Shotgun Assembly database in GenBank.19,20 On the basis of the report that these short-chain branched acids were produced in the mitochondria of A. suum muscle tissue, we used the human mitochondrial thiolase T2 responsible for branched amino acid degradation (UniProtKB P24752)21 as a search query and obtained seven thiolase candidates, six of which had sequence identity >50% (Supporting Information Figure S1). Five of these thiolases (Acat1−5) were predicted to be targeted to mitochondria22,23 and were selected for further biochemical characterization. The five synthetic genes were then assembled using a truncated sequence for the predicted mature polypetide using standard codon optimization for heterologous expression in E. coli (Supporting Information Table S1). The Acat1−5 proteins were purified by Ni affinity chromatography and then cleaved with Tobacco Etch Virus (TEV) protease to remove the N-terminal His6-tag (Supporting Information Table S1, Figure S2). All proteins were further purified by size-exclusion chromatography and appeared to elute as active tetramers (Supporting Information Figure S3), as is typically found for biosynthetic thiolases in solution.21 Purified Acat1−5 thiolases were then screened for activity with the two- and three-carbon acyl-coenzyme A (CoA) substrates, acetyl-CoA and propionyl-CoA. Activity was monitored by CoA release using Ellman’s reagent (Dithiobis(2-nitrobenzoic acid), DTNB) and compared to the canonical thiolase (PhaA) from Cupriavidus necator, which selectively assembles two acetyl-CoA monomers.24,25 From this initial screen, Acat5 resembles the canonical PhaA thiolase involved in poly(hydroxyl)alkanoate (PHA) synthesis in bacteria, with high activity for acetyl-CoA and no detectable activity for propionyl-CoA (Figure 2A). Interestingly, Acat2 and Acat3 showed higher activity with respect to propionylCoA compared to acetyl-CoA. Acat1 also displayed activity for propionyl-CoA condensation, but to a lesser degree. Acat4 demonstrated no activity as a thiolase in the synthetic or 14528

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society Identification and Characterization of Ketoreductase AsHadh. In addition to the thiolase, we also sought to identify a ketoreductase capable of accepting the α-branched substrates. Searching the RNASeq database with human mitochondrial short chain hydroxyacyl-CoA dehydrogenase HADH227 as query, we identified just one ketoreductase candidate in A. suum, AsHadh (ADY43626.1). His10-AsHadh was expressed in E. coli from a synthetic gene and purified (Supporting Information Table S1 and Figure S13). Steady-state kinetic characterization of the enzyme revealed a ∼20-fold preference for α-methyl substrates, 2-methyl-3-oxobutyryl-CoA and 2methyl-3-oxopentanoyl-CoA, compared to the linear substrate 3-oxobutyryl-CoA (Figure 3A). This selectivity arose mostly from significantly lower KM values for the α-methyl substrates. Interestingly, the α-ethyl substituted substrate, 2-ethyl-3oxohexanoyl-CoA, is also accepted by this enzyme with similar efficiency as the linear substrate. Potent substrate inhibition was observed for AsHadh (Supporting Information Figure S14), which is common in NAD(P)H oxidoreductases28 based on their ordered-binding reaction pathway.29 We next investigated the stereoselectivity of AsHadh with respect to the 3-position in a coupled assay testing its ability to oxidize the two enantiomers of 3-hydroxybutyryl-CoA. Starting with crotonyl-CoA, either 3S-hydroxybutyryl-CoA or 3Rhydroxybutyryl-CoA can be formed by stereospecific hydration with the crotonase or PhaJ hydratases, respectively (Figure 3B).4 We then monitored the ability of AsHadh to oxidize each enantiomer of in situ-generated 3-hydroxybutyryl-CoA. In these experiments, we found that AsHadh had 40-fold higher catalytic efficiency when coupled to crotonase compared to PhaJ, corresponding to a preference for 3S- over 3R-hydroxybutyrylCoA (Figure 3A). Although the stereochemical preference of AsHadh at the α-branch site remains cryptic, we would predict a (2S,3S) configuration based on the behavior of murine HADH230 and the expectation that the subsequent dehydration proceeds by a syn-elimination mechanism.31 Production of Branched Enoic Acids in E. coli. We next set out to use A. suum enzymes to produce branched organic acids through construction of a synthetic pathway in E. coli (Figure 4A, Supporting Information Figure S15). To produce enoic acids, we identified a gene from A. suum encoding dehydratase AsEch based on its similarity to the human mitochondrial hydratase ECHS132 and predicted mitochondrial localization. Biochemical characterization of His6-AsEch purified from heterologous expression in E. coli showed that it prefers linear substrates but is permissive of the α-methyl substrate (Supporting Information Figures S16−S17A,B). Plasmids containing a thiolase (Acat2, 3, or 5), AsHadh, and AsEch were then constructed from the pTrc99a base vector. A second set of pTrc33-based plasmids was also constructed with a panel of four thioesterases from E. coli (TesA, TesB, YciA, or YdiI)33 to release free organic acids from the corresponding acyl-CoA intermediates. Dual-plasmid systems were transformed into E. coli BAP1, a modified BL21(de3) strain engineered to increase the intracellular concentration of propionyl-CoA.34 E. coli strains 1−16 were cultured in LB media supplemented with 2 g L−1 propionate and tested for their ability to produce different enoic acids (Figure 4B). Product titer was observed to be highly dependent on the thioesterase, with YciA and YdiI both being competent for enoic acid production. The highest titers were obtained using YdiI, which is in agreement with previous studies.33 Strain 12 containing Acat3, AsHadh, and AsEch in combination with YdiI

Figure 4. Production of enoic acids in E. coli BAP1 strains. (A) A twoplasmid system was constructed for production of enoic acids. The first plasmid consists of an operon containing a thiolase (Acat2, 3, or 5; A2, 3, or 5), a ketoreductase (AsHadh), and a dehydratase (AsEch) driven by a single promoter. The second plasmid contains different thioesterases (TesA, TesB, YciA, or YdiI). (B) Production of enoic acids in select strains. E. coli BAP1 were cotransformed with both plasmids. Strains were cultured for 2 d at 30 °C in LB broth containing 1% (w/v) glucose and 0.2% (w/v) sodium propionate before collecting samples for analysis. Data are mean ± s.d. of biological replicates (n = 3). Data are not displayed for Strains 1−4 (*) as no product was detected.

produced a total of 1120 ± 60 mg L−1 branched enoic acid products, tiglic acid and 2-methylpentenoic acid, resulting in the incorporation of 87% of supplemented propionate into these products. The dominant product was 2-methylpentenoic acid resulting from the condensation of 2 equiv of propionylCoA. Interestingly, branched enoic acids were produced almost exclusively in all strains, including those containing Acat5 (15 and 16), a thiolase that had poor activity with branched substrates in vitro. This product distribution may reflect the importance of the preference of the AsHadh ketoreductase for the α-methyl branch in driving the pathway equilibrium forward given the reversibility of the thiolase reaction. Another factor that may affect enoic acid product distribution, however, is the thermodynamic difference in dehydration of linear or branched hydroxyacyl-CoA substrates. Using component contribution methods to estimate the pathway thermodynamics with eQuilibrator,35 we noted that pathway energetics diverged greatly for the dehydration of linear and branched intermediates (Supporting Information Figure S17C,D). As an experimental measure, we then determined Keq values for the hydration of the linear crotonyl-CoA and corresponding branched tigloyl-CoA to be 4.68 ± 0.06 and 0.164 ± 0.001, respectively. The unfavorable hydration of α-branched 14529

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society

lower than for enoic acids. We thus set out to optimize flux through branched hydroxy acid synthesis by modulating pathway enzyme expression levels with different gene promoters and ribosome-binding site (RBS) sequences. Improvements in titers were achieved by using a TesB expression construct with an alternative RBS (pTrc33-TesB2, strains 29 and 30). Although pTrc33-TesB2 was predicted to have ∼11-fold higher expression level (110 K) than the initial RBS (10 K),36 lower level of soluble TesB was observed (Supporting Information Figure S19). Titration of upstream enzymes Acat3 and AsHadh (Supporting Information Figure S20) gave only slight improvement in selective production of αmethyl products (strain 38). A titration of Acat3 in conjunction with CnHadh (WP_011614598), a homologue of AsHadh (52% sequence identity) from C. necator, showed these strains were capable of producing higher titers of branched hydroxy acids (up to 140 ± 20 mg L−1, strain 46). However, this improved productivity was accompanied by a loss of selectivity with an additional 500 ± 100 mg L−1 of 3-hydroxybutyrate and 3-hydroxypentanoate products being formed. In vitro characterization of CnHadh showed that it did indeed accept α-methyl substituted substrates with high catalytic efficiency, although with 5-fold lower preference for the branched substrate compared to AsHadh (Supporting Information Figure S21). In vitro assays for 3-hydroxy acid production comparing AsHadh and CnHadh to two well-characterized ketoreductases PhaB and Hbd with strong linear substrate preference reflect the product distribution observed in vivo (Supporting Information Figure S22). These data indicate that the ketoreductase is indeed a strong determinant of the product distribution between branched and linear 3-hydroxy acids. The improved titers seen in strains 29 and 30 compared to 22 and 26 suggested a critical role for TesB in controlling pathway flux, suggesting that overly high TesB levels can derail successful hydroxy acid production (Supporting Information Figure S19). To further explore the origin of pathway sensitivity to TesB, we purified TesB and tested its hydrolytic activity against various acyl-CoA substrates and found that TesB accepted propionyl-CoA and 2-methyl-3S-hydroxybutyryl-CoA with similar efficiency (Supporting Information Figure S23). Taken together, these data are consistent with a model that the pathway is limited by hydrolysis of the propionyl-CoA building block by the thioesterase. Given the importance of thioesterase selectivity in controlling pathway yield, we screened a panel of hydrolases that could have improved performance with 3-hydroxyacyl thioesters. These candidates included putative 3-hydroxyisobutyrylCoA37−39 and hydroxybenzoyl-CoA hydrolases,40−42 as well as the thioesterase domain of the 6-deoxyethronolide polyketide synthase (Supporting Information Figure S24).43 The most promising candidate of those tested was the 3hydroxyisobutyryl-CoA hydrolase from Bacillus cereus, BcBch. The resulting strains 56 containing BcBch thioesterase outperformed TesB-expressing strain 38 and allowed branched hydroxy acid titers of440 ± 70 mg L−1 in TB media (Figure 5). In vitro characterization of purified BcBch showed low catalytic efficiency on both propionyl-CoA and 2-methyl-3-oxobutyrylCoA (Supporting Information Figure S25), so we reasoned that high BcBch expression could further improve hydroxy acid production. Strain 57 expressing BcBch from a T7 promoter did indeed achieve a titer of 600 ± 100 mg L−1 (Figure 5), although SDS-PAGE did not indicate an obvious difference in

substrates and their corresponding stability in the enoic form is likely related to the chemical stability of the trisubstituted alkene compared to a disubstituted alkene and may also contribute to the observed product distribution. Production of Branched Hydroxy Acids in E. coli. While the production of branched enoic acids has been previously engineered,11 robust pathways for the production of their corresponding 3-hydroxy acids have yet to be reported. In order to construct a pathway for the production of hydroxy acids, the same panel of thiolases and thioesterases were assembled with AsHadh in the absence of AsEch and tested in E. coli BAP1 cultured in LB media supplemented with 2 g L−1 propionate (Figure 5, Supporting Information Figure S18, strains 17−28). Even in strains containing the thioesterase TesB, which has been shown to be effective for producing short linear hydroxy acids in E. coli,12 productivity was considerably

Figure 5. Production of 3-hydroxy acids acids in E. coli BAP1 strains. (A) A two-plasmid system was constructed for production of 3hydroxy acids. The first plasmid consists of an operon containing a thiolase (Acat2, 3, or 5; A2, 3, or 5) and a ketoreductase (AsHadh or CnHadh) driven by a single promoter. The second plasmid contains different thioesterases (TesA, TesB, YciA, YdiI, or BcBch). pTrc33TesB and pTrc33-TesB2 contain different RBS sequences, with reduced TesB expression from pTrc33-TesB2 leading to increased product titer. Strain 58 also contains a propionyl-CoA transferase (Pct) in an operon with the thioesterase. (B) Production of 3-hydroxy acids in select strains. E. coli BAP1 were cotransformed with both plasmids. Strains were cultured for 2 d at 30 °C in LB broth containing 1% (w/v) glucose or TB broth containing 2.5% (w/v) glucose. Media was supplemented with 0.2% (w/v) sodium propionate. Data are mean ± s.d. of biological replicates (n = 3). Data are not displayed for Strains 17−25 (*) as no product was detected. 14530

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society

It is also important to note that the thiolase selectivity does appear to greatly affect substitution at the ω-position, as the pathways with Acat2 and Acat3 preferentially use propionylCoA as a starter to produce the 6-carbon product, 2-methylpentenoic acid, rather than the 5-carbon product, tiglic acid, as reported for the FadA thiolase.11 As previously observed, thioesterase selectivity plays an important role in pathway yield as nonspecific hydrolysis of the precursor acyl-CoAs limits carbon flux into the pathway.33 In this study, we find that enoic acids are generated at high yield due to the high selectivity of the thioesterase for the enoyl-CoA functional group. In comparison, entry into the 3-hydroxy acids appears to be more limited, which we attribute to the hydrolysis of the propionyl-CoA starter but can be improved by addition of a propionyl-CoA transferase. This large fraction of unconsumed propionate lowers yields of the 3-hydroxy acids to 44% of the total propionate added despite being produced with high selectivity (>90% of consumed propionate). Taken together, these studies provide insight into the use of thiolase-based pathways for developing carbon−carbon bond forming cycles using enzymes. In addition, the enzymes identified in this study can be used to construct high-yielding pathways for the production of α-substituted compounds, particularly 2-methyl3-hydroxy acids that may be used as monomers for controlling the material properties of plastics.46

soluble BcBch expression compared to strain 56 (Supporting Information Figure S26). Unconsumed propionate remaining in spent media could account for nearly all of the propionate not incorporated into product for strain 57, resulting in a 97% yield of consumed propionate. Finally, the additional overexpression of propionyl-CoA transferase Pct from Clostridium propionicum44 afforded 1.1 ± 0.2 g L−1 of 2-methyl-3oxobutyrate in strain 58, corresponding to 44% conversion of supplemented propionate (Supporting Information Figure S27).

■

CONCLUSION The noncanonical branched acid fermentation of A. suum serves as a starting point to identify new enzymes involved in physiological thiolase-based pathways. Characterization of this pathway also yields new genetic components for engineering synthetic metabolic pathways for production of α-branched acids. Of the five thiolases expressed in A. suum, two (Acat2 and Acat3) were found to demonstrate a preference for propionylCoA over acetyl-CoA. Upon further characterization in the C− C bond cleaving or degradative direction, we observed that Acat2 and Acat3 demonstrated up to 640-fold increase in accommodation of branched substrates in comparison to the control thiolase, Acat5. Interestingly, sequence alignments show that Acat5 contains a Phe residue that was reported to serve as the gatekeeper residue to allow for substrate branching,21 highlighting the need to identify and characterize more thiolase enzymes to increase our understanding of sequence-substrate selectivity relationships. We have also examined a ketoreductase, AsHadh, which we assign to this pathway based on its preference for branched substrates. Further stereochemical characterization indicates that AsHadh prefers the 3S- rather than the 3R-hydroxy enantiomer, despite falling in the same superfamily of ketoacyl reductases as the 3R-hydroxy acidforming ketoreductases exemplified by PhaB in the PHA pathway.45 Characterization of the dehydratase, AsEch, shows that it can accommodate branched substrates with a 15-fold preference for the linear congener. These enzymes have been used to demonstrate a cellular route for the production of branched organic acids starting from glucose and propionate, including α-methyl enoic acids (1120 ± 60 mg L−1) and chiral 2-methyl-3-hydroxy acids (1100 ± 200 mg L−1) at 87% and 44% yield from fed propionate, respectively. Based on biochemical and thermodynamic analysis, we propose that the product distribution of these pathways is controlled both by the ketoreductase selectivity as well as difference in reaction equilibrium for dehydration of linear and branched 3-hydroxy acids. Given the reversibility of the thiolase-catalyzed condensation, our data suggest that the 20-fold preference of AsHadh for branched substrates leads to the dominant production of branched products even when the thiolase demonstrates a strong preference against these pathways (Acat5). This model is consistent with the observation that the selectivity for branched 3-hydroxy acids is significantly diminished with CnHadh, which displays only a 4-fold branched substrate preference in vitro. However, even with a 1000-fold selectivity difference, the Acat5 thiolase retains a catalytically competent kcat/KM (∼104) for branched substrates. This effect appears to be further amplified in the case of enoic acids, possibly because of a change of equilibrium of 3-hydroxy acid dehydration conferred by the α-methyl substituent to favor the enoic acid product.

■

METHODS

■

ASSOCIATED CONTENT

Detailed procedures for plasmid construction, protein purification, acyl-CoA substrate synthesis and characterization, cell culture, biochemical assays, and metabolite analysis are provided in Supporting Information. S Supporting Information *

Supporting Information includes a full description of . The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b07400. Materials and methods; DNA and protein sequences; SDS-PAGE, NMR spectra, HPLC and size-exclusion chromatography traces; biochemical and production data (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Michael R. Blaisse: 0000-0002-2336-0223 Michelle C. Y. Chang: 0000-0003-3747-7630 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Dr. Jeffery Hanson for purification of PhaA, PhaB, PhaJ, Hbd, and Crt enzymes and Tristan de Rond for advice on chemical synthesis. M.R.B. acknowledges the support of an NSF Graduate Research Fellowship. H.D. thanks the U.C. Berkeley Tang Distinguished Scholars Program for a postdoctoral fellowship. This work was funded by the generous support of the Center for Sustainable Polymers, a National Science Foundation-supported center for Chemical Innovation (Grant CHE-1413862). The College of Chemistry NMR 14531

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532

Article

Journal of the American Chemical Society

(31) Willadsen, P.; Eggerer, H. Eur. J. Biochem. 1975, 54, 247. (32) Agnihotri, G.; Liu, H. W. Bioorg. Med. Chem. 2003, 11, 9. (33) McMahon, M. D.; Prather, K. L. J. Appl. Environ. Microbiol. 2014, 80, 1042. (34) Pfeifer, B. A.; Admiraal, S. J.; Gramajo, H.; Cane, D. E.; Khosla, C. Science 2001, 291, 1790. (35) Flamholz, A.; Noor, E.; Bar-Even, A.; Milo, R. Nucleic Acids Res. 2012, 40, D770. (36) Espah Borujeni, A.; Channarasappa, A. S.; Salis, H. M. Nucleic Acids Res. 2014, 42, 2646. (37) Lee, S.-H.; Park, S. J.; Lee, S. Y.; Hong, S. H. Appl. Microbiol. Biotechnol. 2008, 79, 633. (38) Redondo-Nieto, M.; Barret, M.; Morrisey, J. P.; Germaine, K.; Martínez-Granero, F.; Barahona, E.; Navazo, A.; Sánchez-Contreras, M.; Moynihan, J. A.; Giddens, S. R.; Coppoolse, E. R.; Muriel, C.; Stiekema, W. J.; Rainey, P. B.; Dowling, D.; O’Gara, F.; Martín, M.; Rivilla, R. J. Bacteriol. 2012, 194, 1273. (39) Wong, B. J.; Gerlt, J. A. J. Am. Chem. Soc. 2003, 125, 12076. (40) Ismail, W. Arch. Microbiol. 2008, 190, 451. (41) Song, F.; Thoden, J. B.; Zhuang, Z.; Latham, J.; Trujillo, M.; Holden, H. M.; Dunaway-Mariano, D. Biochemistry 2012, 51, 7000. (42) Song, F.; Zhuang, Z.; Dunaway-Mariano, D. Bioorg. Chem. 2007, 35, 1. (43) Yuzawa, S.; Eng, C. H.; Katz, L.; Keasling, J. D. Biochemistry 2013, 52, 3791. (44) Selmer, T.; Willanzheimer, A.; Hetzel, M. Eur. J. Biochem. 2002, 269, 372. (45) Cantu, D. C.; Dai, T.; Beversdorf, Z. S.; Reilly, P. J. Protein Eng., Des. Sel. 2012, 25, 803. (46) Schneiderman, D. K.; Hillmyer, M. A. Macromolecules 2016, 49, 2419.

Facility at U.C. Berkeley is supported in part by the National Institutes of Health (1S10RR023679 and 1S10RR16634-01).

■

REFERENCES

(1) Clomburg, J. M.; Crumbley, A. M.; Gonzalez, R. Science 2017, 355, 355. (2) Liao, J. C.; Mi, L.; Pontrelli, S.; Luo, S. Nat. Rev. Microbiol. 2016, 14, 288. (3) Zhou, K.; Edgar, S.; Stephanopoulos, G. In Methods in Enzymology; Sarah, E. O. C., Ed.; Academic Press: 2016; Vol. 575, p 225. (4) Bond-Watts, B. B.; Bellerose, R. J.; Chang, M. C. Nat. Chem. Biol. 2011, 7, 222. (5) Dellomonaco, C.; Clomburg, J. M.; Miller, E. N.; Gonzalez, R. Nature 2011, 476, 355. (6) Jiang, M.; Zhang, H. Curr. Opin. Biotechnol. 2016, 42, 1. (7) Meadows, A. L.; Hawkins, K. M.; Tsegaye, Y.; Antipov, E.; Kim, Y.; Raetz, L.; Dahl, R. H.; Tai, A.; Mahatdejkul-Meadows, T.; Xu, L.; Zhao, L.; Dasika, M. S.; Murarka, A.; Lenihan, J.; Eng, D.; Leng, J. S.; Liu, C.-L.; Wenger, J. W.; Jiang, H.; Chao, L.; Westfall, P.; Lai, J.; Ganesan, S.; Jackson, P.; Mans, R.; Platt, D.; Reeves, C. D.; Saija, P. R.; Wichmann, G.; Holmes, V. F.; Benjamin, K.; Hill, P. W.; Gardner, T. S.; Tsong, A. E. Nature 2016, 537, 694. (8) Haapalainen, A. M.; Meriläinen, G.; Wierenga, R. K. Trends Biochem. Sci. 2006, 31, 64. (9) Fox, A. R.; Soto, G.; Mozzicafreddo, M.; Garcia, A. N.; Cuccioloni, M.; Angeletti, M.; Salerno, J. C.; Ayub, N. D. Gene 2014, 533, 5. (10) Peretó, J.; López-García, P.; Moreira, D. J. Mol. Evol. 2005, 61, 65. (11) Cheong, S.; Clomburg, J. M.; Gonzalez, R. Nat. Biotechnol. 2016, 34, 556. (12) Martin, C. H.; Dhamankar, H.; Tseng, H.-C.; Sheppard, M. J.; Reisch, C. R.; Prather, K. L. J. Nat. Commun. 2013, 4, 1414. (13) Sheppard, M. J.; Kunjapur, A. M.; Wenck, S. J.; Prather, K. L. J. Nat. Commun. 2014, 5, 5031. (14) Watanabe, Y.; Ishizuka, K.; Furutate, S.; Abe, H.; Tsuge, T. RSC Adv. 2015, 5, 58679. (15) Tijsma, E. J.; Does, L. V. D.; Bantjes, A.; Vulic, I. J. Macromol. Sci., Polym. Rev. 1994, 34, 515. (16) de Mata, Z. S.; Lizardo, R.; Diaz, F.; Saz, H. J. Arch. Biochem. Biophys. 1991, 285, 158. (17) Komuniecki, R.; Campbell, T.; Rubin, N. Mol. Biochem. Parasitol. 1987, 24, 147. (18) Dunn, E. W.; Coates, G. W. J. Am. Chem. Soc. 2010, 132, 11412. (19) Altschul, S. F.; Gish, W.; Miller, W.; Myers, E. W.; Lipman, D. J. J. Mol. Biol. 1990, 215, 403. (20) Benson, D. A.; Clark, K.; Karsch-Mizrachi, I.; Lipman, D. J.; Ostell, J.; Sayers, E. W. Nucleic Acids Res. 2015, 43, D30. (21) Haapalainen, A. M.; Merilainen, G.; Pirila, P. L.; Kondo, N.; Fukao, T.; Wierenga, R. K. Biochemistry 2007, 46, 4305. (22) Claros, M. G.; Vincens, P. Eur. J. Biochem. 1996, 241, 779. (23) Nielsen, H.; Engelbrecht, J.; Brunak, S.; von Heijne, G. Protein Eng., Des. Sel. 1997, 10, 1. (24) Kim, E.-J.; Kim, K.-J. Biochem. Biophys. Res. Commun. 2014, 452, 124. (25) Sato, S.; Maruyama, H.; Fujiki, T.; Matsumoto, K. J. Biosci Bioeng 2015, 120, 246. (26) Jiang, C.; Kim, S. Y.; Suh, D. Y. Mol. Phylogenet. Evol. 2008, 49, 691. (27) Ofman, R.; Ruiter, J. P. N.; Feenstra, M.; Duran, M.; Poll-The, B. T.; Zschocke, J.; Ensenauer, R.; Lehnert, W.; Sass, J. O.; Sperl, W.; Wanders, R. J. A. Am. J. Hum. Genet. 2003, 72, 1300. (28) Soler, J.; de Arriaga, D.; Cadenas, Q.; Cadenas, E. Exp. Mycol. 1981, 5, 357. (29) Cooper, W. C.; Jin, Y.; Penning, T. M. J. Biol. Chem. 2007, 282, 33484. (30) Luo, M. J.; Mao, L. F.; Schulz, H. Arch. Biochem. Biophys. 1995, 321, 214. 14532

DOI: 10.1021/jacs.7b07400 J. Am. Chem. Soc. 2017, 139, 14526−14532