Thiolase engineering for enhanced butanol production in Clostridium acetobutylicum


  • Miriam S. Mann,

    1. Abteilung Mikrobiologie, Institut für Biowissenschaften, Universität Rostock, Albert-Einstein-Str. 3, Rostock 18059, Germany; telephone: 49-381-498-6154; fax: 49-381-498-6152
    Search for more papers by this author
  • Tina Lütke-Eversloh

    Corresponding author
    1. Abteilung Mikrobiologie, Institut für Biowissenschaften, Universität Rostock, Albert-Einstein-Str. 3, Rostock 18059, Germany; telephone: 49-381-498-6154; fax: 49-381-498-6152
    • Abteilung Mikrobiologie, Institut für Biowissenschaften, Universität Rostock, Albert-Einstein-Str. 3, Rostock 18059, Germany; telephone: 49-381-498-6154; fax: 49-381-498-6152
    Search for more papers by this author


Biosynthetic thiolases catalyze the condensation of two molecules acetyl-CoA to acetoacetyl-CoA and represent key enzymes for carbon–carbon bond forming metabolic pathways. An important biotechnological example of such a pathway is the clostridial n-butanol production, comprising various natural constraints that limit titer, yield, and productivity. In this study, the thiolase of Clostridium acetobutylicum, the model organism for solventogenic clostridia, was specifically engineered for reduced sensitivity towards its physiological inhibitor coenzyme A (CoA-SH). A high-throughput screening assay in 96-well microtiter plates was developed employing Escherichia coli as host cells for expression of a mutant thiolase gene library. Screening of this library resulted in the identification of a thiolase derivative with significantly increased activity in the presence of free CoA-SH. This optimized thiolase comprised three amino acid substitutions (R133G, H156N, G222V) and its gene was expressed in C. acetobutylicum ATCC 824 to assess the effect of reduced CoA-SH sensitivity on solvent production. In addition to a clearly delayed ethanol and acetone formation, the ethanol and butanol titers were increased by 46% and 18%, respectively, while the final acetone concentrations were similar to the vector control strain. These results demonstrate that thiolase engineering constitutes a suitable methodology applicable to improve clostridial butanol production, but other biosynthetic pathways involving thiolase-mediated carbon flux limitations might also be subjected to this new metabolic engineering approach. Biotechnol. Bioeng. 2013; 110: 887–897. © 2012 Wiley Periodicals, Inc.


The group of thiolases is devided into biosynthetic and degradative enzymes with different substrate specificities. The degradative thiolase (E.C.; acyl-CoA:acetyl-CoA C-acetyltransferase; thiolase I) plays a major role in the fatty acid β-oxidation pathway and is often referred to as 3-ketoacyl-CoA thiolase due to its broad substrate range for long-chain fatty acids (C4–C16). In contrast, the biosynthetic thiolase (E.C.; acetyl-CoA:acetyl-CoA C-acetyltransferase; thiolase II) is specific for the C4 substrate: it catalyzes the reversible reaction of two molecules acetyl coenzyme A (CoA) to acetoacetyl-CoA, that is, both the endergonic condensation and exergonic thiolytic cleavage. As a central metabolic intermediate, acetoacetyl-CoA is involved in various metabolic pathways such as fatty acid metabolism, synthesis and degradation of ketone bodies, mevalonate pathway, and biosynthesis of secondary metabolites.

In general, thiolases share high sequence and structure homologies among eukaryotes and prokaryotes and the Claisen condensation for carbon bond formation was studied in detail (Haapalainen et al., 2006). Crystallographic analyses including various substrate–enzyme complexes, enzyme kinetics, and site-directed mutagenesis provided insights into the catalytic mechanism of the biosynthetic thiolase of Zoogloea ramigera. Naturally, the thiolase of Z. ramigera is involved in the biosynthesis of polyhydroxybutyrate, an abundant type of cytoplasmic energy and carbon storage compound (Lu et al., 2009). The thiolase forms a tetramer and each subunit consists of two core domains and a large loop domain (residues 119–249) which interacts with the CoA moiety (Modis and Wierenga, 1999, 2000). The catalysis proceeds via a two-step ping-pong mechanism: first, cysteine 89 is activated by histidine 348 and then acylated to form a covalently bound acyl-CoA intermediate; second, either acetyl-CoA (condensation reaction) or CoA-SH (thiolysis reaction) are added to the substrate–enzyme complex while cysteine 378 catalyzes the condensation and the product is released. No allosteric domain was found in the thiolase protein, suggesting a competitive inhibition by CoA-SH (Kursula et al., 2002; Meriläinen et al., 2008, 2009).

Escherichia coli harbors two thiolases: thiolase I is encoded by fadA and prefers acyl-CoA substrates with medium chain lengths of up to 16 carbon atoms, whereas the activity with acetoacetyl-CoA as substrate contributes only 5–10% of the activity with 3-ketododecanoyl-CoA, confirming the involvement of thiolase I in fatty acid metabolism. In contrast, thiolase II which is encoded by atoB is specific for acetoacetyl-CoA and does not accept long-chain substrates. Thiolase II is involved in the utilization of butyric acid and also catalyzes the reverse condensation reaction (Feigenbaum and Schulz, 1975; Jenkins and Nunn, 1987a, b; Maloy and Nunn, 1981; Yang et al., 1990). However, E. coli's native thiolase genes are only expressed under specific conditions, for example, when butyrate, acetoacetate, or medium chain length fatty acids such as oleic acid, respectively, are present in the culture medium.

A biotechnologically important pathway involving the thiolase-catalyzed condensation is the acetone–butanol–ethanol (ABE) fermentation of solventogenic clostridia which regained much interest in the past years for microbial biofuel production from renewable resources (Dürre, 2011; Green, 2011; Lee et al., 2008). Nevertheless, the promising potential of the clostridial metabolic capacities have not been explored in much detail and the main regulatory functions are still not understood, because genetic inaccessibility of these bacteria prevented detailed research on the molecular level until recently. Prior to the publication of the genome sequence of Clostridium acetobutylicum, the model organism for solventogenic clostridia, enzymes of the central fermentative metabolism were purified and biochemically characterized (Bennett and Rudolph, 1995; Dürre, 1998; Nölling et al., 2001). The first enzyme of the butyrate/butanol (C4) fermentative pathway is the thiolase which catalyzes the condensation of two molecules acetyl-CoA to acetoacetyl-CoA. In the following three reactions, butyryl-CoA is formed and can be converted to butyrate with concomitant adenosine triphosphate (ATP) formation, or it is utilized by the bifunctional aldehyde/alcohol dehydrogenase resulting in n-butanol production. Thiolase activity is steadily increasing during the course of fermentation, exhibiting its maximum in the early stationary growth phase. C. acetobutylicum possesses two thiolase encoding genes, thlA (CAC2873) and thlB (CAP0078), but only thlA is physiologically relevant whereas thlB exhibited very low expression levels and is only briefly induced during the transition between acidogenic and solventogenic phase (Grimmler et al., 2011; Winzer et al., 2000). Representing a general principle of regulating metabolic pathways, the first enzyme, that is, thiolase, is strictly regulated by the substrate:product ratio in order to control the metabolic flux. Similar to other thiolases, the C. acetobutylicum thiolase is very sensitive to free CoA-SH and significant inhibition occurs at micromolar CoA-SH concentrations, mild inhibitory effects of butyryl-CoA and ATP were also detected (Gheshlaghi et al., 2009; Hartmanis and Gatenbeck, 1984; Wiesenborn et al., 1988).

This study introduces thiolase engineering as a new approach to increase fluxes through acetoacetyl-CoA-dependent metabolic pathways: first, a suitable high-throughput screening method was developed; second, the thiolase of C. acetobutylicum was specifically engineered for reduced CoA-SH sensitivity; and third, an optimized thiolase variant was applied to enhance n-butanol production in C. acetobutylicum.

Materials and Methods

Bacterial Strains and General Cultivation Conditions

All strains and plasmids used in this study are listed in Table I. E. coli was cultivated in LB medium comprising per liter 5 g yeast extract, 10 g tryptone, and 10 g NaCl, antibiotics for plasmid maintenance were added as required (Sambrock and Russell, 2001). C. acetobutylicum was cultivated anaerobically at 37°C without shaking in Hungate tubes (Ochs GmbH, Bovenden, Germany) or serum bottles (Müller & Krempel AG, Bülach, Switzerland). Resazurin (7-hydroxy-10-oxidophenoxazin-10-ium-3-one) was added as a redox indicator for anaerobiosis at a concentration of 1 mg/L and residual oxygen was removed by addition of 50–100 µL titanium (III) nitrilotriacetic acid (NTA) solution (1.3 M NaOH, 0.16 M NTA, 0.27 M Na2CO3, and 1.3% TiCl3). Reinforced clostridial agar (RCA; Oxoid Deutschland GmbH, Wesel, Germany) was used as solid medium, 40 µg/mL erythromycin was added for plasmid-containing strains of C. acetobutylicum. Procedures requiring strictly anaerobic conditions were conducted in an anaerobic chamber with 90% N2 and 10% H2 (MG1000; Meintrup DWS, Lähden-Holte, Germany).

Table I. Bacterial strains, plasmids, and oligonucleotides used in this study.
Strain, plasmid, or oligonucleotideRelevant characteristics or sequenceRefs.
 C. acetobutylicum ATCC 824Wild typeAmercian Type Culture Collection
 E. coli BL21 (DE3)F, ompT, hsdSB(rmath image, mmath image), gal, dcm (DE3)

Studier and Moffatt (1986)

 E. coli DH5αFmath image, ϕ80lacZΔM15, Δ(lacZYA), recA1, endA1, hsdR17 (rmath image, mmath image), phoA, supE44 thi1, gyrA96, relA1, λmath image

Grant et al. (1990)

 E. coli ER2275mcrA, ΔmcrBC, hsdR, recA1

Mermelstein and Papoutsakis (1993)

 E. coli Tuner (DE3) pET15b::phaB1Expression and purification of PhaB1 from Ralstonia eutropha with N-terminal His-Tag, ApR

Budde et al. (2010)

 pAN2p15A, ϕ3TI, TetR

Heap et al. (2007)

 pASK-IBA3F1 ori, P(tetA) ATG, tetR, Strep-tagII (C-terminal), AmpRIBA GmbH, Göttingen (Germany)
 pASK-IBA3::thlAWTF1 ori, P(tetA) ATG, tetR, Strep-tagII (C-terminal), AmpR, thlA wild-type gene of C. acetobutylicum (CAC2873)This study
 pASK-IBA3::thlAMUTF1 ori, P(tetA) ATG, tetR, Strep-tagII (C-terminal), AmpR, thlA mutant libraryThis study
 pASK-IBA3::thlAOPTF1 ori, P(tetA) ATG, tetR, Strep-tagII (C-terminal), AmpR, thlA mutant (R133G, H156N, G222V)This study
 pTColE1, repL, EryR, AmpR, P(thlA)

Mann et al. (2012)

 pT::thlAOPTColE1, repL, EryR, AmpR, P(thlA), thlAOPT (R133G, H156N, G222V)This study

Thiolase Gene Cloning and Expression

General recombinant DNA techniques were conducted according to standard protocols (Sambrock and Russell, 2001). All cloning procedures were validated by DNA sequencing (LGC Genomics GmbH, Berlin, Germany). The thlA gene (CAC2873) was amplified by PCR from chromosomal DNA of C. acetobutylicum ATCC 824 (Fischer et al., 2006) using the primers ThlA_EcoRI_fw and ThlA_KpnI_rv (Table I) and ligated via EcoRI and KpnI restriction sites into the vector pASK-IBA3 (IBA GmbH, Göttingen, Germany), resulting in plasmid pASK-IBA3::thlAWT, which was tranformend into E. coli BL21 (DE3). For protein expression, E. coli BL21 (DE3) pASK-IBA3::thlAWT, and E. coli BL21 (DE3) pASK-IBA3::thlAMUT strains were cultivated in LB medium with 100 µg/mL ampicillin at 37°C until an OD600 of 0.6–0.8. After addition of 20 µg/mL anhydrotetracycline (AHT), the cultures were incubated over night at 20°C and 180 rpm prior to crude extract preparation for enzyme activity assays.

The optimized thlA mutant gene identified in the library screening, thlAOPT, was subcloned via EcoRI/KpnI restriction of pASK-IBA3::thlAOPT, agarose gel purification, T4 DNA polymerase treatment and ligation into the pT vector, yielding plasmid pT::thlAOPT. After in vivo methylation in E. coli ER2275 pAN2 (Heap et al., 2007; Mermelstein and Papoutsakis, 1993), pT::thlAOPT was electroporated into C. acetobutylicum ATCC 824 as described previously (Riebe et al., 2009).

Enzyme Activity Measurements

Cell crude extracts of E. coli were prepared by addition of 5 mL of BugBuster 10× Protein Extraction Reagent (Merck Bioscience, Darmstadt, Germany) which was previously diluted 1:10 with 0.1 M Tris/HCl buffer, pH 6.0, per gram cell wet weight and the suspension was incubated at room temperature for 20 min. To remove the cell debris, the suspension was centrifuged at 4°C and 13,000 rpm for 30 min. The supernatant was used for enzyme activity measurements and the protein content was determined according to Bradford (1976).

For the NADPH-coupled thiolase assay, acetoacetyl-CoA reductase (PhaB) activity was provided by crude extracts of E. coli Tuner (DE3) pET15b::phaB1 (Budde et al., 2010). E. coli Tuner (DE3) pET15b::phaB1 was cultivated in LB containing 100 µg/mL ampicillin and 25 µg/mL chloramphenicol at 37°C to an OD600 of 0.6–0.8, 24 mg/mL isopropyl-β-D-thiogalactopyranoside (IPTG) was added and the culture was further incubated at 30°C for 4 h. Cells were harvested by centrifugation at 4°C and 5,000g for 10 min, and the cell pellets were stored at −20°C when necessary. Cell crude extracts were prepared as described above. The PhaB activity was measured spectrophotometrically: 10–50 µL of crude extract were added to 0.9 mL of potassium phosphate buffer (0.05 mM, pH 6.0) containing 32 nM acetoacetyl-CoA and 100 nM NADPH, and the absorption was recorded at 340 nm (Haywood et al., 1988).

The condensation reaction of thiolase activities were determined in extracts of E. coli BL21 (DE3) pASK-IBA3::thlA in 50 mM potassium phosphate buffer, pH 6.0, comprising 1 mM dithiotreitol (DTT), 0.2 mM NADPH, and 2 U PhaB (in crude extracts of E. coli Tuner (DE3) pET15b::phaB1). The reaction was started by addition of 1 mM acetyl-CoA and monitored at 340 nm on an Ultrospec 3000 photometer (Amersham Buchler GmbH & Co. KG, Braunschweig, Germany). The reverse thiolytic cleavage reaction was measured spectrophotometrically by acetoacetyl-CoA decrease at 303 nm as described (Hartmanis and Gatenbeck, 1984). Cell extract preparation of C. acetobutylicum strains and thiolase activity measurements were conducted as described in detail previously (Hartmanis and Gatenbeck, 1984; Lehmann and Lütke-Eversloh, 2011).

Thiolase Library Construction

To construct the thlA mutant library (designated as pASK-IBA3::thlAMUT), synthetic oligonucleotides building the degenerated backbone of the library were assembled without any amplification reaction involved (non-amplified library), and the diversity of the library was directly correlated to the amount of DNA molecules produced, that is, 67 fmol/µL corresponded to 4 × 1010 molecules (Geneart AG, Regensburg, Germany). To generate the amplified library, 4 µL (268 fmol or 1.6 × 1011 molecules) of the non-amplified library were amplified with the primers T7_fw and M13_rv, full-length fragments were gel-purified and resuspended in 100 µL Tris/EDTA buffer. The concentration of the amplified library was determined by UV spectroscopy as 400 ng/µL. The amplified library was digested with EcoRI and KpnI and ligated into the EcoRI/KpnI-restricted vector pASK-IBA3. Ligation reactions were transformed into E. coli BL21 (DE3) and the transformation rate was determined by plating of dilution series. The total number of transformants was 5.7 × 104 colony forming units (CFU). Total cells from the transformation plates were harvested for plasmid preparation. The concentration of the cloned library was determined by UV spectroscopy as 0.46 µg/µL. The plasmid preparations were sequenced with primers pASK_fw, pASK_rv, and 1107435.S1_rv (Geneart AG). Total cells from the transformation plates were harvested, resupended in 50 vol% glycerol and aliquots of the cell suspension comprising 3.9 × 1010 cells/mL were stored at −70°C.

Microtiter Scale Cultivation and Screening Procedure

The E. coli pASK-IBA3::thlAMUT library was plated on LB agar plates plus 100 µg/mL ampicillin (LB + Ap) and incubated over night at 37°C. Colonies were transferred with sterile toothpicks into 96-well microtiter plates containing 200 µL LB + Ap and grown over night at 37°C. Each microtiter plate comprised two samples each of E. coli pASK-IBA3::thlAWT and E. coli pASK-IBA3 (empty vector) as controls. Ten microliter aliquots of the cultures were used to inoculate fresh LB + Ap and the grown microtiter plates were stored at −70°C after addition of 10 vol% glycerol. The inoculated microtiter plates were incubated at 37°C until the cultures exhibited an OD600 of 0.5–0.7. For induction of gene expression, 20 µg/mL AHT was added, and the cultures were incubated over night at 20°C. Subsequently, cell crude extracts were prepared using BugBuster 10× Protein Extraction Reagent (Merck Bioscience, Darmstadt, Germany) as described above to determine the thiolase activities of the E. coli pASK-IBA3::thlAMUT library. For this, 150 µL potassium phosphate buffer (50 mM, pH 6.0) comprising 1 mM DTT and 0.2 mM NADPH were pipetted into the wells of fresh microtiter plates. When analyzing the CoA-SH sensitivity, the buffer contained 10 µM free coenzyme A (Sigma–Aldrich Chemie, Deishofen, Germany). After addition of 20 µL PhaB extract (2 U) and 10 µL cell crude extracts, the reaction was started by addition of 20 µL acetyl-CoA solution (20 mM). The absorption at 340 nm was recorded for 3 min in 3 s intervals on a SpectraMax M2e Multi-Mode Microplate Reader and the slope values were automatically calculated using the SoftMax Pro Software 5.4.1 (Molecular Devices GmbH, Biberach an der Riss, Germany). All enzyme activity measurements were normalized according to the individual protein concentrations to counterbalance varying cell densities.

Fermentation Experiments for Solvent Production

Phenotypic characterization of recombinant C. acetobutylicum strains was carried out by cultivation experiments in 200 mL MS-MES comprising 40 µg/mL erythromycin. Medium composition, cultivation conditions, and analytical procedures were described in detail previously (Lehmann et al., 2012a, b; Lehmann and Lütke-Eversloh, 2011).


Development of a Microtiter Scale Thiolase Screening Platform

Biosynthetic thiolases are important branchpoint enzymes of various metabolic pathways and are tightly regulated by product-driven feedback inhibition. In order to increase the carbon fluxes through acetoacetyl-CoA-dependent pathways, we sought to generate an optimized thiolase enzyme being less sensitive to the inhibition by free CoA-SH. In general, high-throughput screenings constitute a compromise of accuracy and technical feasibility of analyzing large populations, provided that a particular feature of a cell or a protein can actually be screened (Dietrich et al., 2010; Sui and Wu, 2007). With respect to deregulated metabolic enzymes, antimetabolites have been employed as a screening parameter for positive selection, for example, to identify amino acid overproducing bacteria (Ikeda, 2003; Lütke-Eversloh and Stephanopoulos, 2005). The underlying mechanism of this traditional approach is the selection of feedback inhibition-resistant mutants, which is also a desired phenotype of the thiolase described in this study. However, suitable CoA analogues are commercially not available and chemical synthesis was shown to be complicated and exhibited only very low yields (Chase et al., 1966; Stewart et al., 1968; Strauss and Begley, 2002). Other strategies allowing positive selection and thus, a fast throughput of large populations were evaluated. On the one hand, butyric acid utilization of an E. coli host strain lacking native thiolase genes, that is, knockout of atoB and fadA, can be mediated by a recombinant thiolase enabling growth on agar plates containing butyric acid as sole carbon source (Jenkins and Nunn, 1987a). On the other hand, thiolase complementation of recombinant polyhydroxybutyrate biosynthesis in E. coli can be visualized by nile red staining and subsequent fluorescence measurements either on agar plates or in microtiter plates in a high-throughput manner (Hoover et al., 2012; Rhie and Dennis, 1995; Slater et al., 1992; Spiekermann et al., 1999). However, these phenotypic selection strategies were considered as not appropriate because of inaccuracy, low assay sensitivity, and non-quantifiability. Therefore, a direct approach to measure thiolase activity and its CoA-SH sensitivity was preferred to select a desired enzyme derivative. Employing E. coli as host strain and well-established technical protocols, a microtiter scale assay was developed. The thiolase gene thlA of C. acetobutylicum ATCC 824 (CAC2873) was cloned into the expression vector pASK-IBA3, exhibiting high activities in E. coli BL21 (DE3) of both thiolytic cleavage and condensation reactions (Table II). Since the physiological direction of the thiolase activity, that is, the condensation reaction, was subjected to optimization, a coupled spectrophotometric thiolase assay was required. Therefore, acetoacetyl-CoA consumption was established by heterologous phaB1 expression from Ralstonia eutropha (Budde et al., 2010), resulting in 34.3 ± 1.4 U/mg activities of the NADPH-dependent acetoacetyl-CoA reductase in crude extracts of E. coli Tuner (DE3) pET15b::phaB1. As shown in Table II, the E. coli host strains did neither exhibit native thiolase nor acetoacetyl-CoA reductase activity and thus, provided a suitable basis for a recombinant thiolase activity screening.

Table II. Enzyme activities in cell extracts of E. coli hosts.
StrainAcetoacetyl-CoA reductase (PhaB)Thiolase (ThlA)
NADHNADPHThiolytic cleavageCondensation
E. coli BL21 (DE3) pASK-IBA3 (control)0.03 ± 0.020.02 ± 0.010.08 ± 0.040.2 ± 0.09
E. coli BL21 (DE3) pASK-IBA3::thlAWT0.05 ± 0.030.73 ± 0.0965.94 ± 4.6168.96 ± 1.54
E. coli Tuner (DE3) pET15b::phaB10.15 ± 0.0234.27 ± 1.430.02 ± 0.0030.3 ± 0.11

Using the thlA wild-type gene of C. acetobutylicum, the screening assay was simulated in order to determine the range of thiolase activity data, which is summarized in Figure 1. Microtiter cultures of recombinant E. coli strains were used to prepare cell-free extracts which were subjected to spectrophotometric thiolase activity measurements according to NADPH decrease at 340 nm. It is important to note that the standard thiolase assay was performed in the PhaB-coupled forward direction. Since PhaB activity was provided in cell crude extracts, the saturation of the coupling enzyme remained unknown, thus, the measured data represent relative values determined as units thiolase activity per milligram cell protein rather than true units per milligram enzyme. Due to significant evaporation, the proximal wells of the 96-well plates were not used for cultivation and enzyme assays. The average thiolase activity of E. coli pASK-IBA3::thlAWT was 71 ± 4 U/mg exhibting a standard deviation of only 5.2% (n = 144). In order to screen for ThlA mutants with reduced sensitivity towards free CoA-SH as the physiological inhibitor, the assay was repeated under the same conditions with the addition of 10 µM CoA-SH each, revealing 27 ± 1 U/mg of the wild-type ThlA activity. The CoA-SH concentration of 10 µM was chosen as the standard concentration for the screeening because it was optimum for a good measurable range while clearly exhibiting inhibitory effects. Hence, C. acetobutylicum wild-type thiolase activities in cell-free extracts of recombinant E. coli constituted the reference values for the subsequent library screening. However, other thiolase encoding genes expressed in E. coli are similarly suitable for this assay as a target enzyme to be engineered for enhanced activity and/or for reduced or alleviated sensitivity to inhibition.

Figure 1.

Thiolase assay development in 96-well microtiter plates. a: Time course of absorbance at 340 mM (A340 nm) of the coupled ThlA/PhaB assay in crude extracts of E. coli BL21 (DE3) pASK-IBA3::thlAWT. b: Reproducibility of the coupled ThlA/PhaB assay in crude extracts of E. coli BL21 (DE3) pASK-IBA3::thlAWT. Distribution of 144 independent thiolase activity measurements; the gray dashed line indicates the mean value.

Library Construction and Screening for Optimized Thiolase Derivatives

According to the sequence homology of the Z. ramigera thiolase, which was subjected to detailed biochemical and crystallographic analyses, the amino acid residues 119–249 form a large loop domain which interact with the CoA moiety of the substrate (Meriläinen et al., 2008, 2009). In order to minimize the chance of generating inactive mutants and maximize the number of hits, only the loop region was subjected to mutagenesis, that is, the catalytic residues Cys89, His348, and Cys378 were not involved. The random mutant library of ThlA comprised an average number of 2.5 amino acid substitutions in the region of residues 119–254 (Fig. 2), and the correctness of the mutant library was validated by DNA sequencing of 48 randomly selected clones.

Figure 2.

Amino acid sequence alignment of thiolases from C. acetobutylicum and Z. ramigera. Sequence alignment was performed using ClustalW 2.1; “*” represents identical, “:” highly similar, and “.” similar amino acid residues. CAC2873, thiolase of C. acetobutylicum; Z.ram., thiolase of Z. ramigera. The catalytic histidine and cysteine residues are underlined, the residues of the loop domain involved in the CoA binding are shaded.

A total of 1620 clones of the E. coli BL21 (DE3) pASK-IBA3::thlAMUT library were analyzed for thiolase activities and sensitivity towards free CoA-SH. As illustrated in Figure 3, mutant thiolase activities exhibited a wide range of both increased and reduced values as compared to the wild-type ThlA activity (Fig. 3a and c), and only 56 clones (<4%) revealed no thiolase activity. Interestingly, such wide-spread values were not observed when the same clones were subjected to thiolase activity measurements in the presence of 10 µM CoA-SH, that is, 80% of the mutants showed activities within ±20% of the wild-type ThlA activity (Fig. 3b and d). Finally, 14 ThlMUT variants were identified by significantly increased activity, that is, >150% of the ThlWT activity in the presence of free CoA-SH. It is noteworthy that increased activity values of both assays detected during the screening did not correlate to each other.

Figure 3.

Screening of the E. coli pASK-IBA3::thlAMUT library. Thiolase activities were measured in cell crude extracts without (a) and in the presence of 10 µM free CoA-SH (b). The relative distribution as percent of wild-type thiolase activity without (c) and in the presence of 10 µM free CoA-SH (d) are shown. The arrow in (d) indicates the 14 positive clones selected for further characterization.

Characterization of Positive Clones

In order to validate the microtiter screening results, the 14 putative positive clones were cultivated at a larger scale and thiolase activities were spectrophotometrically determined in the cuvette format (Table III). None of the ThlAMUT variants exhibited an increased overall thiolase activity, but differences to ThlAWT were detected in the presence of 10–50 µM CoA-SH. Whereas eight clones showed similar activities at various CoA-SH concentrations as the wild-type ThlA, the clones 9G4, 13D7, 19C10, 21B8, 25C8, and 28C9 revealed significantly increased thiolase activities when CoA-SH was added to the assay (Table III). Subsequent DNA sequencing of the respective plasmids exhibited that all six positive clones had exactly the same genotype, indicating that in fact only one novel ThlAMUT derivative was isolated. Basically, the six independently identified hits confirmed the reliability of the screening procedure to find optimized thiolase mutants. Thus, clone 9G4 was chosen as a representative for subsequent experiments and was referred to as E. coli BL21 (DE3) pASK-IBA3::thlAOPT. The clearly reduced sensitivity of ThlAOPT towards free CoA-SH in comparison to ThlAWT is illustrated in Figure 4. Remarkably, at a concentration of 50 µM CoA-SH, ThlAOPT showed a more than 10-fold higher thiolase activity than ThlAWT (Fig. 4).

Table III. Thiolase activities of putative positive clones identified in the E. coli BL21 (DE3) pASK-IBA3::thlAMUT library screening.
ThlA variantThiolase activity in the presence of free CoA-SH
0 µM10 µM20 µM30 µM40 µM50 µM
Wild type68.96 ± 1.5425.60 ± 0.848.44 ± 0.671.30 ± 0.030.28 ± 0.030.11 ± 0.03
7D569.80 ± 3.4126.23 ± 1.588.54 ± 0.131.33 ± 0.080.29 ± 0.040.13 ± 0.01
9G469.19 ± 0.5534.60 ± 1.6717.23 ± 1.069.83 ± 0.252.06 ± 0.101.18 ± 0.08
11B969.39 ± 2.0124.81 ± 0.588.38 ± 0.061.22 ± 0.040.30 ± 0.050.10 ± 0.03
12B869.50 ± 1.3725.23 ± 0.238.52 ± 0.091.30 ± 0.120.27 ± 0.030.11 ± 0.03
13D768.94 ± 0.8936.45 ± 0.6817.28 ± 0.5711.87 ± 0.081.98 ± 0.171.17 ± 0.07
17C570.11 ± 1.2724.58 ± 0.778.30 ± 0.411.06 ± 0.040.32 ± 0.020.13 ± 0.02
19C1069.76 ± 0.9037.89 ± 0.5318.57 ± 0.5810.52 ± 0.281.86 ± 0.761.43 ± 0.09
21B868.72 ± 0.1234.13 ± 1.0616.79 ± 0.4311.32 ± 0.202.37 ± 0.091.38 ± 0.15
22B769.85 ± 1.1924.85 ± 0.988.47 ± 0.093.76 ± 4.830.32 ± 0.050.13 ± 0.02
25C870.33 ± 0.5335.92 ± 0.4717.33 ± 0.539.95 ± 0.772.12 ± 0.181.34 ± 0.19
28B968.61 ± 0.5424.80 ± 0.538.90 ± 0.731.02 ± 0.080.37 ± 0.020.15 ± 0.01
28C469.88 ± 1.0024.61 ± 0.818.42 ± 0.301.12 ± 0.030.31 ± 0.050.14 ± 0.00
28C969.51 ± 1.0037.89 ± 0.5320.73 ± 4.519.83 ± 0.332.00 ± 0.171.23 ± 0.01
30F371.03 ± 0.7924.56 ± 0.658.41 ± 0.011.32 ± 0.200.29 ± 0.040.12 ± 0.01
Figure 4.

Thiolase inhibition by free CoA-SH. Open squares and dashed line, ThlAWT; closed circles and solid line, ThlAOPT. Mean values of six independent measurements are shown.

Genotypic characterization of ThlAOPT elucidated the molecular changes which led to the reduced sensitivity towards CoA-SH, revealing three distinct amino acid substitutions in the CoA-binding loop region of the enzyme. First, arginine 133 was replaced by a glycine residue (R133G); second, the conserved histidine 156 was substituted by asparagine (H156N), and third, glycine 222 was exchanged to a valine residue (G222V). These modifications obviously altered the binding characteristics of CoA-SH without affecting the specific activity in the absence of the inhibitor.

Application of an Optimized Thiolase for Butanol Production

Thiolase engineering was performed in an E. coli host strain which did not physiologically benefit from an optimized enzyme derivative. Therefore, the thlAOPT gene was subcloned into an E. coliC. acetobutylicum shuttle vector under the control of its own promotor, resulting in plasmid pT::thlAOPT. After transformation into C. acetobutylicum ATCC 824, the recombinant clostridial strain exhibited a thiolase activity of 53.03 ± 0.63 U/mg, whereas the vector control strain C. acetobutylicum pT revealed a specific actitivity of 28.22 ± 0.68 U/mg. To investigate the metabolic impact of ThlAOPT, cultivation experiments in mineral salts medium comprising glucose as carbon source were performed to assess the reduced CoA-SH sensitivity of the new thiolase variant on solvent formation. C. acetobutylicum pT::thlAOPT produced similar amounts of acetone but higher levels of ethanol and butanol as compared to the vector control strain (Fig. 5). Most interestingly, acetone and ethanol formation was clearly delayed whereas butanol synthesis began at the same time as in the control culture. Moreover, expression of thlAOPT resulted in a significantly increased butanol production during the stationary growth phase, elevating the final butanol titer from 142 mM (10.5 g/L) to 168 mM (12.4 g/L).

Figure 5.

Fermentation profiles of recombinant C. acetobutylicum strains. Closed circles, C. acetobutylicum pT::thlAOPT; open diamonds, C. acetobutylicum pT (vector control). Average data of three independent replicates for each strain are shown.


A general biological principle of regulating biosynthetic pathways is the product-driven enzyme feedback inhibition, which typically applies to various branch point enzymes. Alleviating such inhibitory mechanisms can be regarded as traditional successful metabolic engineering strategies to generate microbial production strains, amino acid producers being the most prominent examples (Ikeda, 2003). Biosynthetic thiolases are important enzymes for C–C building block pathways such as fatty acid metabolism, mevalonate, and polyhydroxybutyrate biosynthesis as well as several secondary metabolite pathways. In addition, the thiolase constitutes a key enzyme in clostridial n-butanol formation, an important source for advanced biofuels, because it catalyzes the condensation of two molecules acetyl-CoA as the initial step of the C4 biosynthetic pathway (Gheshlaghi et al., 2009).

Considering the recent development of new analytical and metabolic engineering tools for members of the genus Clostridium, several rational approaches to optimize butanol production were conducted. These are, however, limited to the manipulation of gene expression patterns such as gene knockout or downregulation, respectively, and/or gene overexpression to manipulate metabolic fluxes towards the desired product (Lee et al., 2008; Lütke-Eversloh and Bahl, 2011; Papoutsakis, 2008). In an attempt to direct the flux from acetate towards butanol formation, the aldehyde/alcohol dehydrogenase gene adhE1 (aad) and the thiolase gene thlA (thl) were homologously overexpressed in the solvent-negative degenerated strain C. acetobutylicum M5. Whereas adhE1 overexpression resulted in high alcohol titers, additional thlA overexpression could not alter the phenotype of acetate being the main fermentation product in the recombinant C. acetobutylicum M5 strain. Moreover, butanol production of a C. acetobutylicum ATCC 824 strain comprising a downregulated acetone pathway and adhE1 overexpression could not be increased by simultaneous thlA overexpression (Sillers et al., 2008, 2009). Interestingly, thlA overexpression alone led to clearly reduced solvent production and the previously excreted acids were not re-assimilated by C. acetobutylicum 824(pTHL) (Sillers et al., 2009). These findings indicated that the thiolase is regulated on the enzymatic rather than on the genetic level, and that only increased amounts of thlA transcripts do not necessarily support an increased flux through the C4 pathway. Thermodynamically, the thiolytic cleavage of acetoacetyl-CoA is the preferred reaction, but the physiological condensation reaction is driven by the subsequent reduction of acetoacetyl-CoA to 3-hydroxybutyryl-CoA. Not only the cellular acetyl-CoA and acetoacetyl-CoA pools are important parameters determining the thiolase reaction, the CoA-SH concentrations tightly regulate thiolase activities in vivo (Boynton et al., 1994; Wiesenborn et al., 1988).

Therefore, we sought to specifically engineer the key branch point enzyme of butanol production for improved activity in the presence of CoA-SH, a potent inhibitor of the thiolase in C. acetobutylicum (Wiesenborn et al., 1988). To our best knowledge, no clostridial enzyme involved in biofuel production has been subjected to directed evolution approaches thus far.

With respect to convenient cultivation techniques and handling protocols, E. coli was chosen as a host to establish the thiolase activity screening platform. Activities were measured in the physiological direction, that is, the endergonic condensation reaction in a coupled assay with crude extracts of E. coli Tuner (DE3) pET15b::phaB1 to provide NADPH-dependent acetoacetyl-CoA reductase (PhaB) activity (Budde et al., 2010). In order to establish a fast and reproducible assay, the fixed amount of 2 U PhaB in cell-free extracts was employed which might not represent the saturation of the coupling enzyme. Using 96-well microtiter plates, thiolase activities were measured twice in the absence and presence of free CoA-SH and the recombinant wild-type thiolase activities of E. coli pASK-IBA3::thlAWT were well reproducible (Fig. 1). Hence, this screening platform was used to search a mutagenized thiolase population for improved activities. According to the structural information of the homologous thiolase protein of Z. ramigera, only the CoA-binding loop region was subjected to random mutagenesis to avoid substitutions of the catalytic residues and a large number of ineffective amino acid exchanges (Fig. 2). From 1620 analyzed clones, 14 mutants revealed highly increased thiolase activities in the presence of 10 µM CoA-SH. These putative positive ThlAMUT clones were characterized in a larger scale, and six clones were validated for reduced CoA-SH sensitivity (Table III). As determined by DNA sequencing, all six thlAMUT variants were identical, clone no. 9G4 was selected as a representative for further characterization and referred to as ThlAOPT. As shown in Figure 4, ThlAOPT was significantly less sensitive to free CoA-SH as compared to ThlAWT, whereas the thiolase activities without CoA-SH addition were similar.

The molecular mechanism underlying the lower sensitivity of ThlAOPT towards CoA-SH inhibition remains to be elucidated. However, the three amino acid substitutions presumably resulted in a lower affinity of the CoA-binding loop domain without affecting catalysis of the Claisen condensation. Whereas the wild-type C. acetobutylicum enzyme contains an arginine and the Z. ramigera thiolase a lysine residue at position 133, ThlAOPT comprised a non-basic glycine residue instead. Serine 247 and histidine 156 are the only hydrophilic residues of the CoA binding pocket of the Z. ramigera thiolase and were reported to be highly conserved (Modis and Wierenga, 1999). Interestingly, the histidine 156 of ThlAOPT was replaced by asparagine, indicating its important but non-essential role for substrate binding. Finally, glycine 222 was substituted by valine in the optimized thiolase derivative, but the structural effects might not necessarily be relevant since both are small hydrophobic amino acids (Meriläinen et al., 2009, 2008; Modis and Wierenga, 2000). Nevertheless, more detailed analyses such as site-specific mutagenesis for single amino acid substitutions will be required to specify the individual roles of the three amino acids involved in the CoA binding of the C. acetobutylicum thiolase enzyme.

In order to provide a physiological proof of concept, the engineered thiolase gene thlAOPT was subcloned into an E. coliC. acetobutylicum shuttle vector and expressed in C. acetobutylicum ATCC 824. Sillers et al. showed previously that overexpression of the native thlAWT gene did not result in an improved butanol production in C. acetobutylicum, suggesting that enzyme inhibition could not be alleviated by increased gene expression and enzyme levels, respectively (Sillers et al., 2009). In contrast, expression of the thlAOPT gene led in fact to an improved phenotype with clearly delayed acetone and ethanol formation (Fig. 5). The final butanol titer was elevated by 18% and interestingly, the ethanol concentrations were also increased whereas the final acetone levels were similar as compared to the vector control strain. The reason for the higher ethanol production of C. acetobutylicum pT::thlAOPT remains unclear, because thlAOPT expression should in theory shift the carbon flux from acetate and ethanol synthesis towards C4 metabolites, but during the late stationary growth phase, ethanol concentrations exceeded those of the control cultures. We hypothyze that the overall improved solventogenic metabolism of C. acetobutylicum pT::thlAOPT might influence the adhE2-encoded NADH-dependent aldehyde/alcohol dehydrogenase activities related to an alcohologenic phenotype with high ethanol titers as indicated previously (Fontaine et al., 2002; Hönicke et al., 2012). On the other hand, increased ethanol production might also be caused by a depletion of butyryl-CoA, the precursor for butanol synthesis (Sillers et al., 2009). Interestingly, three recently reported knockout mutants of C. acetobutylicum also exhibited an ethanologenic phenotype, that is, these mutants revealed significantly elevated ethanol titers. Ethanol production by a mutant with an inactivated hbd gene (coding for the 3-hydroxybutyryl-CoA dehydrogenase of the common C4 pathway) seems reasonable, because butyrate and butanol biosynthetic pathways were eliminated and ethanol formation was the only fermentative branch allowing NADH consumption (Lehmann and Lütke-Eversloh, 2011). On the other hand, disruption of the butyrate biosynthetic pathway did not increase the metabolic flux towards butanol synthesis, but ethanol was the major fermentation product in pH-controlled batch cultivations (Lehmann et al., 2012b). The redox-sensing regulator Rex was found to be a transcriptional repressor of the C4 biosynthesis operon as well as of the adhE2 gene, and knockout of the rex gene resulted in a phenotype with an earlier solventogenic shift as well as high ethanol production (Wietzke and Bahl, 2012). However, supporting information on why the carbon fluxes were re-directed towards ethanol biosynthesis is currently not available and future research will elucidate the molecular mechanisms which determine the metabolic routes in solventogenic clostridia.

In conclusion, this study provided a novel strategy to overcome limitations of thiolase-dependent anabolic pathways. We developed a simple high-throughput screening method in recombinant E. coli cells to identify thiolase derivatives with reduced sensitivity towards its natural inhibitor CoA-SH, and this assay can easily be adapted to an automated robot system for 96-well microtiter plates. An optimized thiolase derivative, ThlAOPT, was identified and validated for significantly increased activity in the presence of free CoA-SH, whereas the specific activity was similar to the wild-type enzyme in the absence of the inhibitor. ThlAOPT outperformed the wild-type thiolase at ≥20 µM CoA-SH, considering more than fivefold higher physiological CoA-SH concentrations (Boynton et al., 1994; Wiesenborn et al., 1988). Future studies on the purified ThlAOPT will provide more details on the enzymology of the engineered thiolase, linking the altered biochemical properties as determined in cell-free extracts to true feedback inhibition resistence. Finally, the engineered thlAOPT gene was expressed in C. acetobutylicum to investigate its lower inhibitory effects on solvent formation. Interestingly, the butanol titer was significantly increased while acetone and ethanol formation was clearly delayed, confirming the suitability of thiolase optimization for phenotype improvement. However, thiolase engineering is not only a useful tool for clostridial biofuel production, it can be applied to other microbial cell factories as well to enhance carbon fluxes through acetoacetyl-CoA-dependent natural or synthetic pathways.


The authors thank Jingnan Lu, Massachusetts Institute of Technology, Cambridge MA (USA), for kind provision of E. coli Tuner (DE3) pET15b::phaB1.