Effect of an Oxygen-Tolerant Bifurcating Butyryl Coenzyme A Dehydrogenase/Electron-Transferring Flavoprotein Complex from Clostridium difficile on Butyrate Production in Escherichia coli

Abstract

The butyrogenic genes from Clostridium difficile DSM 1296^T have been cloned and expressed in Escherichia coli. The enzymes acetyl-coenzyme A (CoA) C-acetyltransferase, 3-hydroxybutyryl-CoA dehydrogenase, crotonase, phosphate butyryltransferase, and butyrate kinase and the butyryl-CoA dehydrogenase complex composed of the dehydrogenase and two electron-transferring flavoprotein subunits were individually produced in E. coli and kinetically characterized in vitro. While most of these enzymes were measured using well-established test systems, novel methods to determine butyrate kinase and butyryl-CoA dehydrogenase activities with respect to physiological function were developed. Subsequently, the individual genes were combined to form a single plasmid-encoded operon in a plasmid vector, which was successfully used to confer butyrate-forming capability to the host. In vitro and in vivo studies demonstrated that C. difficile possesses a bifurcating butyryl-CoA dehydrogenase which catalyzes the NADH-dependent reduction of ferredoxin coupled to the reduction of crotonyl-CoA also by NADH. Since the reoxidation of ferredoxin by a membrane-bound ferredoxin:NAD⁺-oxidoreductase enables electron transport phosphorylation, additional ATP is formed. The butyryl-CoA dehydrogenase from C. difficile is oxygen stable and apparently uses oxygen as a co-oxidant of NADH in the presence of air. These properties suggest that this enzyme complex might be well suited to provide butyryl-CoA for solventogenesis in recombinant strains. The central role of bifurcating butyryl-CoA dehydrogenases and membrane-bound ferredoxin:NAD oxidoreductases (Rhodobacter nitrogen fixation [RNF]), which affect the energy yield of butyrate fermentation in the clostridial metabolism, is discussed.

INTRODUCTION

Genome sequencing of organisms provides information regarding the distribution of genes encoding biotechnologically important metabolic pathways. This is true for the clostridial butyrogenic pathway, which converts acetyl-conenzyme A (CoA)—the terminal oxidation product of glucose via glycolysis—to butyrate. Genes encoding enzymes from this pathway are widespread in genome-sequenced clostridia and related species (1–8). In spite of the central importance of butyrate-forming genes in these organisms, only individual enzymes from a comparably small selection of organisms have been purified and carefully studied in the past (9–15). In particular, Clostridium acetobutylicum enzymes were of great interest due to the organism's capability of producing acetone and butanol (16–18). Driven by the urgent need to replace oil-derived fossil fuels, genetic engineering of microbes for production of butan-1-ol, a promising alternative transportation fuel, is proceeding worldwide (16, 18–24).

A long-known metabolic route such as the butyrate pathway (Fig. 1A) can suddenly gain fundamentally new meanings when hitherto-unknown properties of individual enzymes within this pathway become known. While for decades the role of butyrate formation was generally accepted to provide solely an electron sink for glucose oxidation and ATP regeneration in the glycolysis pathway in clostridia, it became recently evident that some species can also use the significant change in Gibbs energy provided by the reduction of crotonyl-CoA by NADH to conserve additional energy by electron transport phosphorylation of ADP. The bifurcating butyryl-CoA dehydrogenases from these organisms couple the NADH-driven (E₀′ = −320 mV) exergonic reduction of crotonyl-CoA to butyryl-CoA (E₀′ = −10 mV) with the endergonic reduction of ferredoxin (E₀′ = −410 mV) by a second molecule of NADH (25, 26). Reoxidation of reduced ferredoxin coupled with the reduction of NAD⁺ by a membrane-bound NADH:ferredoxin oxidoreductase (also involved in Rhodobacter nitrogen fixation [RNF]) is used to create a proton-motive force across the cytoplasmic membrane, which is finally used to conserve energy by ATP synthase (27). The mechanistic coupling of crotonyl-CoA and ferredoxin reduction in bifurcating butyryl-CoA dehydrogenases makes these enzymes interesting targets for biotechnological applications. These enzymes exhibit unique features which can support solventogenesis as well as provide the ability for butyrate-driven hydrogen production in genetically engineered strains (28).

Fig 1.

(A) The butyrate biosynthetic pathway in genetically engineered E. coli. THIA1, thiolase; HBD, 3-hydroxybutyryl-CoA dehydrogenase; CRT2, crotonase; BCD/etfAB2, butyryl-CoA dehydrogenase complex; PBT, phosphate butyryltransferase; BUK, butyrate kinase; FNR, ferredoxin:NADP⁺ oxidoreductase; PntB, NAD(P)⁺ transhydrogenase. (B) The constructed expression plasmid pASG-ButD.wt harboring all eight genes necessary for butyrate production. Ori, ColEl orgin; tetP, Tet promoter; Amp, ampicillin resistance gene.

Based on these recent findings, it was reasonably to conclude that clostridia harboring genes orthologous to rnf and butyryl-CoA dehydrogenase genes from C. kluyveri or C. tetanomorphum may also possess the ability to use this mechanism for improved energy conservation (26). To test this hypothesis, we decided to investigate the butyrate-forming pathway of the human pathogen C. difficile in more detail. The organism possesses the rnf genes as well as the genes required for butyrate formation (6). Thus, we decided to extract the genes necessary for butyrate formation from the genome of C. difficile. Individually cloned genes were expressed in Escherichia coli in order to confirm the functional production of enzymes, which were purified and initially characterized in vitro. Subsequently, the individual genes were assembled in a single plasmid vector into an artificial operon (Fig. 1B), which allowed functional coexpression of the required genes and conferred butyrate-forming capability to the host. The data presented here demonstrate the presence of a bifurcating butyryl-CoA dehydrogenase in C. difficile. Based on the in vivo studies of the recombinant strains, we suggest that implementation of a bifurcating butyryl-CoA dehydrogenase into artificial metabolic pathways may provide significant advantages in biotechnological solvent production.

MATERIALS AND METHODS

Materials.

Acetoacetyl-CoA, crotonyl-CoA, CoA-SH (where SH is a free thiol group of CoA), ferredoxin (Fd), flavin adenine dinucleotide (FAD), and ferrocenium hexafluorophosphate (Fc⁺FP₆⁻) were purchased from Sigma-Aldrich (Taufkirchen, Germany). Unless otherwise stated, all other chemicals were purchased from VWR International GmbH (Darmstadt, Germany) or Roth (Karlsruhe, Germany) and were of the highest available quality.

Synthesis of butyryl-phosphate.

Butyryl-phosphate was prepared according to the method of Uede and Otomo (29) with slight adaptations: to a mixture of 2.4 ml of phosphoric acid (85%) and 18.4 ml of butyric acid, 34.2 ml of butyric anhydride was added dropwise, and the resulting mixture was subjected to stirring overnight at room temperature. Then, 1.7 g of solid LiOH was added at room temperature for an additional 3 h to precipitate the product. The purity of the product was 93.6% as determined enzymatically with recombinant butyrate kinase (BUK).

Synthesis of acetyl-CoA and butyryl-CoA.

These substrates were prepared from the corresponding anhydrides and CoA-SH as previously described (30, 31).

Bacterial strains.

The strains listed in Table S1 in the supplemental material were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ; Braunschweig, Germany), Invitrogen (Darmstadt, Germany), or Merck (Darmstadt, Germany).

Auxiliary enzyme cloning and expression.

Glucokinase (EC 2.7.1.2) from C. difficile (CD2459), glucose-6-phosphate dehydrogenase (EC 1.1.1.49) from Bacteroides fragilis (BF1854), and ferredoxin:NADP⁺ reductase (EC 1.18.1.2) from E. coli (b3924) were cloned, expressed, and recombinantly produced in E. coli in this work.

DNA manipulations.

Synthetic oligonucleotides were designed based on the genome-derived sequences of individual genes according to the instructions of the Stargate supplier (IBA, Göttingen, Germany) (32) and custom synthesized by Eurofins-MWG-Operon (Ebersberg, Germany; primers are listed in Table S2 in the supplemental material). Genomic DNA from C. difficile (DSM 1296^T; ATCC 9689) was used as a template. Target gene amplification was performed using Phusion DNA polymerase (Fisher Scientific, Schwerte, Germany). The PCR products were purified using a PCR purification kit (GeneJET kit; Fisher Scientific, Schwerte, Germany) and analyzed by chip electrophoresis (MCE-202 MultiNA; Shimadzu Deutschland, Duisburg, Germany). Subsequently, the amplificates were cloned into an Entry vector and further subcloned into dedicated Stargate vectors (see Table S3 in the supplemental material) according to standard protocols and amplified in E. coli DH5α cells (Invitrogen, Darmstadt, Germany). Individual plasmids (see Table S4 in the supplemental material) were isolated using a GeneJET plasmid purification kit (Fisher Scientific, Schwerte, Germany) and custom sequenced by Eurofins-MWG-operon (Ebersberg, Germany).

Gene expression and biomass production.

Gene expression was performed in E. coli Rosetta pLysE cells (Merck, Darmstadt, Germany). Freshly transformed cells were plated, and individual colonies were expanded stepwise to yield final culture volumes of 600 ml in LB medium (carbenicillin, 50 μg/ml, and chloramphenicol, 34 μg/ml). Gene expression was induced with anhydrotetracycline (200 ng/ml) at an optical density at 578 nm (OD₅₇₈) of about 0.5. Cells were harvested by centrifugation 2 h postinduction, washed with Tris-buffered saline (10 mM Tris-HCl [pH 7.5], 150 mM NaCl), and stored at −80°C until used.

Protein purification.

All enzyme purifications were performed using StrepTactin-Macroprep columns (IBA, Göttingen, Germany) (5 ml). Depending on the individual enzymes, different buffers (Table 1), supplemented with 350 mM NaCl, were used for purification.

Table 1.

Buffer composition for individual enzyme purifications

Enzymea	Buffer composition
THIA1	20 mM Tris-HCl (pH 8.0), 1 mM DTT, 1 mM EDTA
HBD	20 mM phosphate (pH 7.0), 1 mM DTT, 1 mM EDTA
CRT2	20 mM phosphate (pH 7.0), 1 mM DTT, 1 mM EDTA
BCD	20 mM phosphate (pH 7.5), 1 mM DTT, 5 mM MgCl2
PBT	20 mM phosphate (pH 7.5), 1 mM DTT, 1 mM EDTA
BUK	20 mM phosphate (pH 7.5), 1 mM DTT, 5 mM MgCl2

^aTHIA1, thiolase; HBD, 3-hydroxybutyryl-CoA dehydrogenase; CRT2, crotonase; BCD, butyryl-CoA dehydrogenase complex; PBT, phosphate butyryl-transferase; BUK, butyrate kinase.

Protein purification was started with about 1 g of wet packed cells. The cells were suspended in 15 ml buffer supplemented with lysozyme (50 μg/ml), avidin (5 μg/ml), and RNase and DNase (2.5 μg/ml each). The suspension was incubated at 37 °C with shaking for 15 min. Afterwards, the cells were homogenized by sonication using a Branson Sonifier with a duty cycle of 50% at 60% power for 10 min in an ice-water bath. Cell debris was removed by ultracentrifugation at 160,000 × g for 30 min. Clear supernatants were loaded on equilibrated StrepTactin-Macroprep columns, and the purification steps were done according to the manufacturer's instructions. Protein concentrations were determined in Bradford assays using bovine serum albumin (BSA) as the standard (33). Protein purities and molecular masses were confirmed by SDS-PAGE and Coomassie staining of the gels (34).

Enzyme activity measurements.

Unless otherwise stated, all enzyme measurements were carried out in 1-cm-path-length half-microcuvettes at room temperature (23 to 25°C) using a final volume of 1 ml and started by the addition of enzyme.

Thiolase (acetyl-CoA C-acetyltransferase [THIA1]) was determined in 100 mM Tris-HCl (pH 8.0) supplemented with 10 mM MgCl₂, 1 mM dithiothreitol (DTT), 50 μM acetoacetyl-CoA, and 50 μM CoA-SH as standard conditions. The disappearance of acetoacetyl-CoA was monitored at 303 nm in quartz cuvettes, and ε_{303 nm} = 14 mM⁻¹ cm⁻¹ was used to calculate enzyme activities (35).

3-Hydroxybutyryl-CoA dehydrogenase activity was determined with 50 μM acetoacetyl-CoA and 200 μM NADH in 50 mM potassium phosphate buffer (pH 6.0) supplemented with 1 mM DTT. The decrease in absorbance at 340 nm was monitored, and ε_{340 nm} = 6.3 mM⁻¹ cm⁻¹ was used to calculate activities (36).

Crotonase (enoyl-CoA hydratase) was measured in 100 mM Tris-HCl (pH 7.5) supplemented with 5 mM MgCl₂, 1 mM EDTA, and 1 mM DTT. Hydration of 50 μM crotonyl-CoA was followed photometrically at 263 nm in quartz cuvettes using ε_{263 nm} = 6.7 mM⁻¹ cm⁻¹ between crotonyl-CoA and 3-hydroxybutyryl-CoA to calculate activities (12). Dilution of the enzyme was performed in Tris-buffered saline (10 mM Tris-HCl [pH 7.5], 150 mM NaCl) supplemented with BSA (1 mg/ml).

All aerobic measurements of butyryl-CoA dehydrogenase were performed in 50 mM potassium phosphate (pH 7.5) supplemented with 50 μM butyryl-CoA, 5 μM FAD, and 100 μM ferricenium hexafluorophosphate (Fc⁺FP₆⁻) in quartz cuvettes. The reduction of ferricenium was followed at 300 nm, and ε_{300 nm} = 2 × 4.3 mM⁻¹ cm⁻¹ was used to calculate activities (37), considering that 2 mol Fc⁺FP₆⁻ is reduced per mol butyryl-CoA. The physiological crotonyl-CoA reduction was measured in 50 mM Tris-HCl (pH 7.5) supplemented with 10 mM MgCl₂, 50 mM NaCl, 5 μM FAD, 150 μM NADH, and 50 μM crotonyl-CoA. Diaphorase activity was obtained in the buffer described above without crotonyl-CoA. Enzyme activity was calculated based on the oxidation of NADH and corrected for the diaphorase activity. Under anaerobic conditions, only the physiological crotonyl-CoA-mediated oxidation of NADH was followed. To establish further requirements of the enzyme, ferredoxin from C. pasteurianum (Sigma-Aldrich, Taufkirchen, Germany) was used at concentrations ranging from 1 to 10 μM and reoxidized by recombinant ferredoxin:NADP⁺ oxidoreductase (FNR; 0.3 U/ml) in the presence of 100 μM NADP⁺.

Butyrate kinase was assayed in 50 mM potassium phosphate (pH 7.5) supplemented with 5 mM MgCl₂, 4 mM glucose, 1 mM DTT, 200 μM NADP⁺, 100 μM ADP, and 1 mM butyryl-phosphate. The butyrate kinase-dependent formation of ATP was coupled with recombinant glucokinase from C. difficile (1 U/ml) and glucose 6-phosphate dehydrogenase from B. fragilis (1 U/ml) to the reduction of NADP⁺ against a control, in which butyryl-phosphate was omitted.

Phosphate butyryltransferase was measured as described for butyrate kinase, replacing butyryl-phosphate with 50 μM butyryl-CoA and providing butyrate kinase (1 U/ml) in excess.

Enzyme kinetics.

Kinetic constants for individual enzymes were determined using the conditions essentially described in the activity measurement section. The kinetic constants for thiolase were determined with respect to thiolytic function using various amounts of acetoacetyl-CoA (5 to 100 μM) in the presence of fixed concentrations of CoA-SH (1.5 to 100 μM) and 21 ng pure enzyme per ml in the assay. Kinetic properties of 3-hydroxybutyryl-CoA dehydrogenase (HBD) were analyzed with respect to physiological function using various amounts of acetoacetyl-CoA (3 to 100 μM) in the presence of fixed concentrations of NADH (25 to 400 μM) using an enzyme concentration of 900 ng per ml in the assay. V_max and K_m for crotonase were determined with crotonyl-CoA (3 to 150 μM) as the substrate and 1.7 ng enzyme. Likewise, kinetic constants of phosphate butyryltransferase (PBT) were determined using 50 ng pure enzyme per ml at different butyryl-CoA concentrations (1.5 to 200 μM). Butyrate kinase kinetic constants were analyzed with respect to butyrate-forming function with butyryl-phosphate (0.09 to 5.75 mM) in the presence of fixed ADP concentrations (25 to 250 μM) and 43 ng enzyme/ml in the assay. The least-squares method was used for curve-fitting routines.

In vivo reconstitution of the butyrate pathway in E. coli.

In order to reconstitute the butyrate pathway of Clostridium difficile in E. coli batch cultures, the plasmid pASG-ButD.wt (Fig. 1) was synthesized and introduced in Rosetta pLysE cells. These cells were grown in M₉ liquid medium (34 mM Na₂HPO₄, 22 mM KH₂PO₄, 19 mM NH₄Cl, 9 mM NaCl, 20 mM glucose, 1 mM thiamine-HCl, 1 mM MgSO₄, 0.1 mM CaCl₂, 50 μM FeSO₄) supplemented with carbenicillin (50 μg/ml) and chloramphenicol (34 μg/ml). The cultures were incubated at 37°C with shaking and expanded in a stepwise manner, maintaining exponential growth to a test volume of 500 ml. The foreign genes were induced with anhydrotetracycline (200 ng/ml) when an OD₅₇₈ of 0.5 was reached. The cultures were sampled at the time intervals shown in Results and centrifuged in order to separate cells from medium.

The harvested cells were suspended in 10 mM Tris-HCl (pH 7.5) supplemented with 150 mM NaCl and used for OD₅₇₈ measurements. The cells were then opened to determine enzyme activity. The medium was analyzed for glucose consumption and fatty acid formation. Glucose concentrations were determined in a coupled enzymatic test using glucokinase from C. difficile and glucose-6-phosphate dehydrogenase from B. fragilis (38). The assay was performed in 50 mM Tris-HCl (pH 7.5) containing 5 mM MgCl₂, 1 mM DTT, 1 mM ATP, and 200 μM NADP⁺ in the presence of glucokinase (0.3 U/ml) and glucose-6-phosphate dehydrogenase (0.14 U/ml). A standard curve with glucose was used as the reference. Short-chain fatty acids were analyzed by high-performance liquid chromatography (HPLC) of p-nitrophenylesters (39).

RESULTS

Cloning and expression of butyrogenic genes.

The six genes encoding the conversion of acetyl-CoA via acetoacetyl-CoA, 3-hydroxybutyryl-CoA, and crotonyl-CoA to butyryl-CoA as well as the two genes encoding the production of butyrate from butyryl-CoA are clustered in the genome of C. difficile strain 630. These genes encode the butyryl-CoA dehydrogenase complex (BCD; EC 1.3.99.2), comprising the dehydrogenase subunit (CD1054, bcd2), electron transfer flavoprotein subunit beta (CD1055, etfB2), electron transfer flavoprotein subunit alpha (CD1056, etfA2), crotonase (EC 4.2.1.17) (CD1057, crt2), 3-hydroxybutyryl-CoA dehydrogenase (EC 1.1.1.157) (CD1058, hbd), and acetyl-CoA C-acetyltransferase (EC 2.3.1.9, thiolase) (CD1059, thiA1). Likewise, the two genes encoding enzymes for butyrate release and energy conservation, phosphate butyryltransferase (EC 2.3.1.19) (CD0112, pbt) and butyrate kinase (EC 2.7.2.7) (CD0113, buk), are joined. This genotype is also found in other sequenced strains of C. difficile and many other clostridia (see Fig. S1 in the supplemental material).

The individual genes were cloned and sequenced from the DSM 1296^T strain. The genes were almost identical (>99% identity) to those referenced in the genome-derived data of the 630 strain (Table 2). The observed deviations were predominantly located on wobble bases and did not change the amino acid sequences of the encoded proteins. The single amino acid exchange in crotonase (I₁₈₄ to V) was found also in PCR products with genomic DNA of the 1296^T strain, suggesting that this exchange results in an amino acid polymorphism between the encountered strains.

Table 2.

Nucleotide and amino acid substitutions observed between Clostridium difficile strain 1296^T and the genome sequence of strain 630^a

Encoded protein	EC no.	Gene name	Substitution
			Nucleotide	Amino acid
Butyryl-CoA dehydrogenase	1.3.99.2	bcd2	None	None
Electron transfer flavoprotein subunit beta		etfB2	None	None
Electron transfer flavoprotein subunit alpha		etfA2	T315→G	None
			A414→C	None
Enoyl-CoA dehydratase (crotonase)	4.2.1.17	crt2	A459→G	None
			A550→G	I184→V
3-hydroxybutyryl-CoA dehydrogenase	1.1.1.157	hbd	T696→C	None
Acetyl-CoA C-acetyl-transferase (thiolase)	2.3.1.9	thiA1	C766→T	None
			T1140→C	None
Phosphate butyryl-transferase	2.3.1.19	pbt	A633→G	None
			A706→C	None
Butyrate kinase	2.7.2.7	buk	T555→C	None
			T748→C	None

^aThe mutation in the pbt gene (indicated in bold) was intentionally introduced in order to remove a natural LguI recognition site by a silent mutation.

The feature of stepwise assembly of the artificial operon which is offered by the Stargate cloning system (32) was used to create fusion plasmids for the recombinant butyryl-CoA dehydrogenase complex composed of the butyryl-CoA dehydrogenase subunit and the two electron-transferring flavoprotein subunits and, later, also to create a fusion plasmid encoding all eight genes required for butyrate formation. To create fusion plasmids for butyryl-CoA dehydrogenase, the three genes were shuffled to obtain the orders ABC, BCA, and CAB, respectively (see Fig. S2 in the supplemental material).

Afterwards, single as well as assembled genes were subcloned into the tetR-regulated expression vectors pASG-IBA3 and pASG-IBA5 in order to allow the recombinant production of C-terminal and N-terminal StrepII-tagged variants, respectively (see Table S4 in the supplemental material). Transfer of the preassembled butyryl-CoA dehydrogenase genes into these vectors allowed production of six variants of the recombinant enzyme, each of which was individually tagged at one site of the particular subunit (see Fig. S2 in the supplemental material). Likewise, the entry vector harboring all eight genes was used to create an expression plasmid for the butyrogenic pathway in pASG-wt1 (Fig. 1B). The resulting expression plasmids were used for recombinant enzyme production in Rosetta-pLysE E. coli cells (Novagen).

Enzyme purification and characterization.

As shown in Fig. 2, each enzyme was obtained at high purity, and the kinetic parameters were established (Table 3).

Fig 2.

Purified recombinant enzymes. Coomassie-stained gels with individually purified proteins are shown. C-terminally StrepII-tagged thiolase (45 kDa, lane 1), 3-hydroxybutyryl-CoA dehydrogenase (34 kDa, lane 2), crotonase (30 kDa, line 3), C-terminally StrepII-tagged phosphate butyryltransferase (35 kDa, lane 4), and butyrate kinase (43 kDa, lane 5) (5 to 10 μg each) were separated by 12% SDS-PAGE. Lane 6, 15 μg of the BCD/Etf complex, which is composed of the dehydrogenase (42 kDa) and Etf subunits A (39 kDa) and B (29 kDa), with a C-terminal StrepII tag attached to EtfA, was separated on 15% SDS-PAGE. The sizes of relevant proteins of PageRuler (Fermentas) molecular mass standards (M) are indicated for comparison.

Table 3.

Kinetic constants of butyrogenic enzymes from Clostridium difficile

Enzymea	Substrate	Vmax (U/mg)	Km (μM)	kcat (s−1)	kcat/Km (μM−1 s−1)	Mol mass (kDa)
THIA1	Acetoacetyl-CoA	625	33	469	14	45
HBD	CoA-SH		27
	Acetoacetyl-CoA	54	17	31	1.8	34
	NADH		210
CRT2	Crotonoyl-CoA	22.5 × 103	52	11.2 × 103	215	30
BCD	Butyryl-CoA	15.3	10	28	2.8	110
	Crotonyl-CoA	10.2	2.5	19	7.6
	Ferredoxin	20.2	2.0	37	18.5
	NADH	19.4	145	36	0.25
PBT	Butyryl-CoA	257	72	151	2.1	35
BUK	Butyryl-P	345	470	245	0.5	43
	ADP		70

^aTHIA1, thiolase; HBD, 3-hydroxybutyryl-CoA dehydrogenase; CRT2, crotonase; BCD, butyryl-CoA dehydrogenase complex; PBT, phosphate butyryl-transferase; BUK, butyrate kinase.

The kinetic constants for thiolase are presented in Fig. S3a, b, and c in the supplemental material. The parallel straight lines in the Lineweaver-Burk plots indicate that the kinetic mechanism for thiolase follows a ping-pong pattern involving an acetyl-enzyme covalent intermediate (see Fig. S3a). Substrate inhibition by CoA-SH was observed at low concentrations of acetoacetyl-CoA.

Kinetics properties of 3-hydroxybutyryl-CoA dehydrogenase are shown in Fig. S4a, b, and c in the supplemental material. The enzyme exhibited a pronounced substrate inhibition by acetoacetyl-CoA in particular at low NADH concentrations (see Fig. S4b and d).

V_max and K_m data for crotonase are displayed in Fig. S5 in the supplemental material. Likewise, kinetics constants of phosphate butyryltransferase are presented in Fig. S6 in the supplemental material.

The straight, intersecting lines observed in the Lineweaver-Burk plots for butyrate kinase suggest a random binding mode for the substrates (see Fig. S7b in the supplemental material). To our knowledge, this is the first measurement of butyrate kinase with respect to physiological function.

Recombinant production of the six affinity-tagged variants of the butyryl-CoA dehydrogenase complex in cell extracts revealed no remarkable differences. For initial screening and purification, the enzyme variants were measured with butyryl-CoA and ferricenium hexafluorophosphate (Fc⁺FP₆⁻) with respect to oxidative function. The enzyme was readily detectable in cell extracts of recombinant cells harboring individual enzyme variants, with specific activities ranging from 100 to 250 mU/mg protein. Two variants (C-terminally tagged dehydrogenase subunit and N-terminally tagged EtfB), however, yielded significantly lower activities (20 and 40 mU/mg, respectively). These values were just above the value corresponding to the endogenous activity of the host strain (<10 mU/mg) in this assay. The purified enzyme variants, however, yielded highly active preparations as well as completely inactive ones. Only three of the six variants, namely, the variants with the tag attached to the C terminus of the EtfA subunit or to the N terminus of either the EtfB or the EtfA subunit, were composed of all three subunits present in approximately equal amounts following purification (Fig. 3, lanes 1, 4, and 6). Notably, the specific activities of these active variants differed significantly (12.6, 3.4, and 1.4 U/mg, respectively). Other variants were composed of individual subunits (Fig. 3, lanes 2, 3, and 5) which were inactive after purification. Thus, all of the following data were collected using the most active enzyme variant with the StrepII tag attached to the C terminus of EtfA.

Fig 3.

Composition of purified butyryl-CoA dehydrogenases fused to StrepII tags at different sites of individual subunits. Recombinant enzyme variants (20 μg of total protein) with C-terminal StrepII tags fused to EtfA (lane 1), dehydrogenase (lane 3), and EtfB (lane 5) and N-terminally tagged dehydrogenase (lane 2), EtfB (lane 4), and EtfA (lane 6) were separated on 12% SDS-PAGE and subjected to Coomassie staining. Note that only the enzymes represented in lanes 1, 4, and 6 are composed of all three subunits. The positions of relevant proteins in a Precision (Bio-Rad) molecular mass standard (M) are shown for comparison.

First, the kinetic data for this enzyme were determined with respect to oxidative function with butyryl-CoA and Fc⁺FP₆⁻, which were also used as the substrates throughout purification. Butyryl-CoA dehydrogenase from C. difficile exhibited an apparent K_m = 10 μM for butyryl-CoA and a V_max = 15.3 U/mg in the presence of 100 μM Fc⁺FP₆⁻ and 5 μM FAD (see Fig. S8 in the supplemental material). The turnover number (k_cat) calculated from V_max and the complex mass of 110 kDa was 28 s⁻¹, and the k_cat/K_m was 2.8 μM⁻¹ s⁻¹ for butyryl-CoA (Table 3).

The findings regarding measurements of butyryl-CoA dehydrogenase activity with respect to physiological function using crotonyl-CoA and NADH as the substrates were much more interesting. The diaphorase activity, determined by measuring the reduction of molecular oxygen with NADH (see equation [Eq.] S1 in the supplemental material), was very low when crotonyl-CoA was omitted from the reaction mixture (0.22 U/mg). In the presence of crotonyl-CoA, this activity increased to 1.1 U/mg and was stimulated only slightly (2 U/mg) by addition of oxidized ferredoxin (10 μM; see Fig. S9 in the supplemental material). Notably, butyryl-CoA was readily detected by mass spectrometry in all tests containing crotonyl-CoA and NADH and, therefore, was not strictly dependent on the presence of ferredoxin under aerobic conditions (data not shown). The enzyme is apparently capable of reducing oxygen, which provides the co-oxidant for reduction of crotonyl-CoA in the presence of air. Indeed, we found that the diaphorase activity of the enzyme with oxygen as the electron acceptor was significantly increased by the presence of crotonyl-CoA (see Fig. S9 in the supplemental material).

The picture significantly changed when the enzyme was measured under anaerobic conditions. As expected, NADH oxidation essentially relied on the presence of crotonyl-CoA. In the absence of ferredoxin, a remarkable stimulation of crotonyl-CoA reduction by FAD was observed, but the effect turned out to be limited by the available flavin concentration. The initially observed enzyme activity was increased from 0.3 U/mg to 2.6 U/mg in the absence and presence of 10 μM FAD, respectively (Table 4). The addition of oxidized ferredoxin from C. pasteurianum at low concentrations (1 to 5 μM) affected the enzyme activity only slightly. A pronounced effect, however, was seen after the establishment of a functional ferredoxin reoxidizing system (see Eq. S2 in the supplemental material) provided by recombinant ferredoxin:NADP⁺ oxidoreductase (FNR; EC 1.18.1.2) from E. coli (35). FNR was used with NADP⁺ (100 μM) as the oxidant (kinetic constants for recombinant FNR are presented in Fig. S10 in the supplemental material). FNR accepts NAD⁺ as well as NADP⁺ as an oxidant. Although NAD⁺ is released by butyryl-CoA dehydrogenase, we found that only NADP⁺ provided in excess allows reliable measurements. Thus, it may be concluded that the NAD⁺ concentration provided by butyryl-CoA dehydrogenase action was too low for effective reoxidation of reduced ferredoxin, which, therefore, must be driven by an excess of NADP⁺ as an oxidant. When this system was utilized to provide oxidized ferredoxin, the specific activity of the butyryl-CoA dehydrogenase increased to 11.8 U/mg in the presence of 150 μM NADH, 50 μM crotonyl-CoA, and 5 μM ferredoxin. Thus, the kinetic parameters of the butyryl-CoA dehydrogenase for NADH, crotonyl-CoA, and ferredoxin were determined independently (Fig. 4).

Table 4.

Establishment of the optimum level for the butyryl-CoA dehydrogenase/Etf complex under anaerobic conditions^a

Expt no.	FAD (μM)	Fd (μM)	FNR (U)	NADP+ (μM)	Sp act (U/mg)	Fold increase in sp act
1	0				0.33	1
2	5				1.48	5
3	10				2.58	8
4	10	1			3.76	12
5	10	1	0.3		5.8	18
6	10	1	0.3	100	7.39	23
7	10	2	0.3	100	9.63	30
8	10	5	0.3	100	11.75	37

^aFAD, flavin adenine dinucleotide; Fd, ferredoxin from C. pasteurianum; FNR, ferredoxin:NADP reductase recombinantly produced in E. coli; NADP⁺, NADP.

Fig 4.

Michaelis-Menten kinetics of recombinant butyryl-CoA dehydrogenase. The parameters for the three substrates crotonyl-CoA (A), NADH (B), and ferredoxin (C) were determined independently. The insets show the Lineweaver-Burk plots of the primary data. Kinetics analyses were performed in 50 mM Tris-HCl (pH 7.5) supplemented with 10 mM MgCl₂, 50 mM NaCl, 10 μM FAD, 100 μM NADP⁺, FNR (0.3 U/ml), and 120 ng enzyme per ml. Note that only the level of the substrate in question was varied in the particular experiments, while the substrates not under consideration were kept constant at 50 μM crotonyl-CoA (B and C), 200 μM NADH (A and C), and 10 μM ferredoxin (A and B).

Butyrate production by recombinant strains.

The obtained data for the individual recombinant enzymes of the butyrate pathway encouraged us to create a single expression plasmid harboring all eight genes necessary for butyrate production (Fig. 1). As summarized in Table 5, the enzyme activities for the target enzymes were either absent or very low in the host strain and likewise in the recombinant strain prior to induction. The activities of the recombinant enzymes were readily detectable in the induced cells harboring pASG-ButD.wt. The enzyme activities in the cell extracts rapidly increased, reached maximum values 2 h postinduction, and slowly declined throughout the continuation of the experiment.

Table 5.

Specific activities for butyrate-forming enzymes in E. coli strains

Strain	Sp act (U/mg protein)a
	THIA1	HBD	CRT2	BCD	PBT	BUK
Rosetta	0.41	0.01	0.22	ND	ND	0.09
BUTD	13.80	0.38	64.3	0.043	1.95	13.22

^aTHIA1, thiolase; HBD, 3-hydroxybutyryl-CoA dehydrogenase; CRT2, crotonase; BCD, butyryl-CoA dehydrogenase complex; PBT, phosphate butyryl-transferase; BUK, butyrate kinase; ND, not detected.

A pronounced effect of the gene induction on the short-chain fatty acids accumulating in the medium was observed (Fig. 5B). After induction, the recombinant strain produced significant amounts of butyrate (up to 3.1 mM) during aerobic growth on a minimal medium, while the host strain exclusively produced acetate (up to 16 mM; Fig. 5A). Notably, the accumulation of acetate was strongly suppressed by butyrate formation (maximum < 1/3 of control). In the exponential phase, the recombinant strain produced biomass, CO₂, and 0.5 mol butyrate per mol of glucose.

Fig 5.

Induced butyrate formation by a recombinant E. coli strain. Escherichia coli Rosetta cells (Novagen) were transformed with the plasmid pASG-ButD.wt. The host cell (A) and the recombinant strain (B) were grown on M₉ medium supplemented with glucose. Glucose consumption (long-dash line), bacterial growth (solid line), acetate levels (dotted line), and butyrate levels (short-dash line) were monitored as indicated. The induction time is indicated by a gray arrow in panel B.

DISCUSSION

In this communication, we report the production of recombinant enzymes from C. difficile required for butyrate formation in E. coli in a systematic approach aiming to gain insights into the physiological context of butyrate formation in the donor organism. Kinetic parameters for the individual recombinant enzymes from C. difficile, which have not been studied so far, differ from the values available for C. difficile enzymes and orthologs from closely related species (compare data in Table 3 and Table 6). Individual differences in V_max values are likely related to the novel enzymatic test methods developed in the current work (e.g., for BCD, PBT, and BUK) or related to a better quality of the recombinant enzymes compared to that of the endogenous enzymes (e.g., THIA1 and CRT2) from C. difficile, which were purified in multistep protocols. The large differences in k_cat and K_m values of HBD compared to those of the reference enzyme from C. kluyveri are most like due to the fact that the latter enzyme is NADP dependent, while the enzyme from C. difficile is exclusively active with NAD.

Table 6.

Range of kinetic constants for butyrogenic enzymes from other clostridia

Enzymea	Substrate	Vmax (U/mg)	Km (μM)	kcat (s−1)	Reference(s)
THIA	Acetoacetyl-CoA	114–216	30–50	80–160	53, 54
	CoA		4.8–70
HBD	Acetoacetyl-CoA	290–450	50	195	13, 35
	NADPH		70
CRT	Crotonoyl-CoA	6.5 × 103	30	4.3 × 103	55
BCD	Butyryl-CoA	1.3–331	6–16	2–237	27, 42, 56
	Crotonyl-CoA	18	4	31	27
	Ferredoxin	5–14	∼10		25
	NADH	12	18	21	27
PBT	Butyryl-CoA	(1.3–1.6) × 103	40–110	700–860	57, 58
BUK	Butyrate	65–420	(14–20) × 103	110–270	10, 59, 60
	ATP		(1.4–5) × 103

^aTHIA1, thiolase; HBD, 3-hydroxybutyryl-CoA dehydrogenase; CRT2, crotonase; BCD, butyryl-CoA dehydrogenase complex; PBT, phosphate butyryl-transferase; BUK, butyrate kinase.

The major issue addressed in this work was whether or not the butyryl-CoA dehydrogenase from C. difficile belongs to the recently described novel subfamily of bifurcating enzymes (25, 26, 40) capable of coupling the exergonic reduction of crotonyl-CoA by NADH with the endergonic reduction of ferredoxin by NADH. In previous studies, ferredoxin produced as a product of crotonyl-CoA reduction was directly consumed by hydrogenase in order to release hydrogen, which was quantified for a description of enzyme activity (25). This assay is, however, restricted to anaerobic applications, and we were interested to investigate the enzyme activities in the presence of air also. Therefore, we used soluble ferredoxin:NADP⁺ oxidoreductase (FNR) from E. coli (41) as an auxiliary enzyme to reoxidize the reduced ferredoxin (see Eq. S2 in the supplemental material). Using this method to assay butyryl-CoA dehydrogenase under anoxic conditions, we found that oxidized ferredoxin is required for continuous crotonyl-CoA reduction in vitro. The dependence of NADH oxidation on both crotonyl-CoA and ferredoxin as oxidants strongly supports the notion that the butyryl-CoA dehydrogenase from C. difficile belongs to the recently discovered class of bifurcating enzymes (25, 26, 42). Since it is not yet possible to distinguish bifurcating acyl-CoA dehydrogenases from the nonbifurcating ones solely based on sequence analyses or other bioinformatical tools, a facile assay for bifurcating enzymes was desired. Since the FNR-mediated reoxidation of commercially available C. pasteurianum ferredoxin significantly stimulates bifurcation enzymes, its application is recommended whenever an electron bifurcation is in question.

Interestingly, the crotonyl-CoA reduction mediated by the C. difficile enzyme appeared to be decoupled from ferredoxin reduction in the presence of air. Here, molecular oxygen apparently serves as an electron acceptor and hydrogen peroxide is formed as a second product. In line with this observation, the recombinant enzyme from C. difficile was found to be highly effective in butyrate production in vivo under aerobic conditions.

The reduced ferredoxin formed in the bifurcating reaction under anaerobic conditions is reoxidized by H₂ generation or by reduction of CO₂ to CO in acetogenesis or to conserve additional energy in the form of a proton-motive force by the membrane-associated complex RNF and/or hydrogenase (27, 43). The capability to conserve additional energy in the reductive part of carbohydrate as well as amino acid fermentation (26, 30, 44, 45) might provide a significant growth advantage and readily explains the wide distribution of RNF-like FNR in the genomes of butyrogenic clostridia.

It might be justified to put the finding of a bifurcating butyryl-CoA dehydrogenase in C. difficile into a more general metabolic context. The metabolic capabilities of C. difficile have been studied intensively in the past decades, and the knowledge of these physiological data may allow the postulation of a central function of the electron-transport system provided by bifurcating butyryl-CoA dehydrogenases and membrane-bound RNF-like FNR in anaerobes.

C. difficile is capable of growing in a defined medium with glucose as the principal source of energy and carbon (46). Notably, the organism produces little gas. Rather, acetate formation at a much higher level (∼70 mM final) than butyrate production (∼ 25 mM) was observed (T. Selmer, unpublished data). This suggests that a significant de novo synthesis of acetate from CO₂ and/or butyrate oxidation takes place. In line with this proposal, two of the three C. difficile strains available in the DSMZ collection were isolated as acetogens from the rumen of new-born lambs (47). Combined with the genome-derived data, a new hypothesis regarding the energy conservation in acetogenesis by C. difficile can be formulated. Under heterotrophic growth conditions with glucose, de novo synthesis of a third acetate from CO₂ essentially requires 8 electrons arising from glucose oxidation (2 NADHs and 2 reduced ferredoxins) and is, therefore, sufficient for the redox balance. One of the two reduced ferredoxins is utilized for the reduction of CO₂ to CO, while the second reduced ferredoxin is either reoxidized by RNF, which forms NAD(P)H with the concomitant generation of a proton-motive force to conserve energy, or used in the reduction of CO₂ to formate. Another two NADHs are used for the reduction of the formyl group, yielding the methyl group of acetyl-CoA. Butyrogenesis, in turn, can accept only 4 electrons, and the redox balance is disturbed when no hydrogen is released. Nevertheless, butyrogenesis may offer an effective alternative route in energy conservation for growing cultures, where excesses of redox equivalents are needed for biomass production, or in complex media, where alternative electron acceptors (e.g., amino acids) are available.

Under autotrophic growth conditions with H₂ and CO₂, reduced ferredoxin is provided by hydrogenases (HydA; EC 1.12.7.2[CD630_33140]) and can drive the generation of a proton-motive force by RNF as well as acetogenesis directly (26, 48, 49). In line with this proposal, it is worth noting that RNF is also present in acetogenic clostridia lacking butyrate synthetic genes, e.g., C. ljungdahlii (3). Accepting the central role of the RNF in clostridia, an interesting functional proposal of the butyrate pathway may be given on the basis of its central role in the anaerobic oxidation of volatile fatty acids upon growth in syntrophic communities. To our knowledge, C. difficile is the only organism whose capability of anaerobic butyrate oxidation in complex media has been demonstrated in pure cultures (50). The oxidation of butyryl-CoA to crotonyl-CoA with NAD⁺ is endergonic and must be driven by the concomitant oxidation of ferredoxin, while all other steps yielding two molecules of acetyl-CoA might proceed smoothly and allow a net substrate-level phosphorylation of ADP. Acetogenesis serves as the electron sink necessary to balance redox potentials, and the strict metabolic coupling of butyrate oxidation and acetogenesis was convincingly shown in thermophilic mixed cultures when methanogenesis was suppressed (51).

The metabolic pathways converting acetyl-CoA to butyryl-CoA are identical in solventogenic and nonsolventogenic clostridia. Remarkably, a bifurcating butyryl-CoA dehydrogenase is probably lacking in C. acetobutylicum (4), which possesses the capability to reduce crotonyl-CoA but does not harbor the rnf genes. The butyryl-CoA dehydrogenase complex of C. acetobutylicum has been previously shown to be rate limiting in the solventogenesis of genetically engineered E. coli strains harboring its solventogenic genes. Its oxygen sensitivity has been suggested to cause the failure of butan-1-ol production in growing cultures of recombinant E. coli strains (16). Furthermore, replacement of this enzyme with trans-enoyl-CoA reductase (Ter) increased butanol production (23). The data presented here show clearly that recombinant expression of C. difficile butyrate genes was effective in creation of an E. coli strain which produced significant quantities of butyric acid under aerobic growth conditions. Thus, the implementation of bifurcating butyryl-CoA dehydrogenase in an artificial metabolic pathway efficiently provides butyryl-CoA as an intermediate for solvent production. Furthermore, the reduced ferredoxin thus formed might provide additional advantages. Given that functional RNF complexes are introduced in the host, additional ATP is generated without impairment of product yield (see Fig. S11 in the supplemental material). The latter could be used in order to provide metabolic energy for improved solvent resistance in genetically engineered strains (for a review, see reference 52).