Enantioselective Synthesis of Vicinal (R,R)-Diols by Saccharomyces cerevisiae Butanediol Dehydrogenase

Abstract

Butanediol dehydrogenase (Bdh1p) from Saccharomyces cerevisiae belongs to the superfamily of the medium-chain dehydrogenases and reductases and converts reversibly R-acetoin and S-acetoin to (2R,3R)-2,3-butanediol and meso-2,3-butanediol, respectively. It is specific for NAD(H) as a coenzyme, and it is the main enzyme involved in the last metabolic step leading to (2R,3R)-2,3-butanediol in yeast. In this study, we have used the activity of Bdh1p in different forms—purified enzyme, yeast extracts, permeabilized yeast cells, and as a fusion protein (with yeast formate dehydrogenase, Fdh1p)—to transform several vicinal diketones to the corresponding diols. We have also developed a new variant of the delitto perfetto methodology to place BDH1 under the control of the GAL1 promoter, resulting in a yeast strain that overexpresses butanediol dehydrogenase and formate dehydrogenase activities in the presence of galactose and regenerates NADH in the presence of formate. While the use of purified Bdh1p allows the synthesis of enantiopure (2R,3R)-2,3-butanediol, (2R,3R)-2,3-pentanediol, (2R,3R)-2,3-hexanediol, and (3R,4R)-3,4-hexanediol, the use of the engineered strain (as an extract or as permeabilized cells) yields mixtures of the diols. The production of pure diol stereoisomers has also been achieved by means of a chimeric fusion protein combining Fdh1p and Bdh1p. Finally, we have determined the selectivity of Bdh1p toward the oxidation/reduction of the hydroxyl/ketone groups from (2R,3R)-2,3-pentanediol/2,3-pentanedione and (2R,3R)-2,3-hexanediol/2,3-hexanedione. In conclusion, Bdh1p is an enzyme with biotechnological interest that can be used to synthesize chiral building blocks. A scheme of the favored pathway with the corresponding intermediates is proposed for the Bdh1p reaction.

INTRODUCTION

Biocatalytic transformations using pure enzymes or whole-cell microorganisms offer mild and environmentally benign reaction conditions, as opposed to the chemical processes that use harsh conditions and produce residual metals. Several diols, such as 1,3-propanediol, 1,2-propanediol, 2,3-butanediol (2,3-BD), and 1,4-butanediol, are considered platform chemicals because of their many applications in the industry, including the synthesis of special chemicals and their use as precursors of polymers (1). Thus, 2,3-butanediol has been used as an antifreeze agent [the stereoisomeric form (2R,3R)-2,3-BD has a freezing point of −60°C] and as a precursor of several compounds, through reactions of dehydration (obtaining methyl ethyl ketone, which can be used as a fuel additive), dehydrogenation (giving flavoring compounds as acetoin and diacetyl), ketalization (yielding a potential gasoline blending agent), and esterification (giving precursors of drugs and plasticizers) (see references 2 and 3 for recent reviews). Several microbial systems have been optimized (by genetic engineering and by optimization of the fermentation conditions) for the production of 2,3-BD, such as Klebsiella pneumoniae (4), Klebsiella oxytoca or Paenibacillus polymyxa (2), and Escherichia coli (5). The manipulation of the carbon flux and cofactor regeneration in Bacillus amyloliquefaciens has resulted in a high titer of 2,3-butanediol from biodiesel-derived glycerol (6). Recent work has also established the potential of some acetogenic bacteria to produce 2,3-butanediol using CO and/or CO₂ plus H₂ (7). Saccharomyces cerevisiae has also been used recently as a producer of 2,3-BD through genetic engineering (8, 9). Thus, a pyruvate decarboxylase yeast mutant was used as a starting strain that was evolved for rapid glucose consumption, such that the resulting strain achieved 96.2 g/liter of 2,3-butanediol production in the fermentation medium (8).

The pure stereoisomers of 2,3-butanediol, (2S,3S)-2,3-BD, (2R,3R)-2,3-BD, and meso-2,3-BD, are useful as auxiliaries and can serve as building blocks in the asymmetric synthesis of chiral compounds with two stereogenic centers (10). The three stereoisomeric forms can be produced by microbial fermentation. Thus, (2R,3R)-2,3-BD can be obtained in high purity from sucrose fermentations by Paenibacillus polymyxa (11) and by metabolic engineering of E. coli and S. cerevisiae, through the introduction of biosynthetic pathways composed of endogenous and foreign genes (12, 13). (2S,3S)-2,3-BD has been obtained recently by fed-batch bioconversion from diacetyl using an E. coli strain expressing 2,3-BD dehydrogenase and glucose or formate dehydrogenases as cofactor-regenerating systems (14). meso-2,3-BD has been produced from a metabolically engineered E. coli from glucose fermentation under low oxygen (15). Several classes of oxidoreductases belonging to the superfamilies of medium-chain dehydrogenases and reductases (MDR), aldoketo reductases, and short-chain dehydrogenases and reductases have been used in those organisms to convert diacetyl and/or acetoin to 2,3-BD (16, 17). A recent biotechnological use for Bdh1p from Bacillus subtilis was the production of (3R)-acetoin and (3S)-acetoin with high enantiomeric excess from (2R,3R)-2,3-butanediol and meso-2,3-butanediol, respectively (18). They used Bdh1p together with an NAD⁺-regenerating system, formed by NADH oxidase from Lactobacillus brevis, to regenerate NAD⁺ from NADH by reducing O₂ to H₂O.

We have previously characterized Bdh1p as an MDR that can reversibly convert R-acetoin and S-acetoin to (2R,3R)-2,3-butanediol and meso-2,3-butanediol, respectively, using NAD(H) as a coenzyme (19). We have also shown that Bdh1p is the main enzyme involved in the last metabolic step leading to (2R,3R)-2,3-butanediol in yeast (20). In the present work, we have used the activity of Bdh1p in different conditions—(i) purified enzyme, (ii) yeast extracts, (iii) permeabilized yeast cells, and (iv) as a fusion protein with yeast formate dehydrogenase (Fdh1p)—to transform diacetyl, 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione to the corresponding diols. By using purified Bdh1p (together with exogenous Fdh to regenerate NADH), we have obtained enantiopure (2R,3R)-2,3-butanediol, (2R,3R)-2,3-pentanediol, (2R,3R)-2,3-hexanediol, and (3R,4R)-3,4-hexanediol. We have also engineered a yeast strain (by developing a new variant of the delitto perfetto methodology [21]) that overexpresses Bdh and Fdh activities in the presence of galactose. The use of this engineered strain allowed us to obtain several diols from the corresponding diketones in the presence of formate, without the need of a coenzyme-regenerating system. Although the use of yeast extracts or permeabilized cells from this strain did not allow the production of pure stereoisomers, we could separate and identify the hydroxyketones and diols arising from the reduction of the diketones. To allow the production of pure diol stereoisomers and to avoid the need of a coenzyme-regenerating system, we have constructed a chimeric fusion protein combining Fdh1p and Bdh1p. Finally, we propose a mechanism to explain the selectivity shown by Bdh1p toward the oxidation and reduction of the hydroxyl and ketone groups from (2R,3R)-2,3-pentanediol and 2,3-pentanedione and (2R,3R)-2,3-hexanediol and 2,3-hexanedione.

MATERIALS AND METHODS

Materials.

Restriction enzymes and T4 DNA ligase were from Roche (Basel, Switzerland). Vent and Thermococcus kodakarensis (KOD) polymerases were from New England BioLabs, Inc. (Beverly, MA) and Merck (Nottingham, United Kingdom), respectively. DNA oligomers were synthesized and purified by Sigma-Genosys (Haverhill, United Kingdom). Chemicals were purchased from Fluka or Sigma-Aldrich (St. Louis, MO). S. cerevisiae alcohol dehydrogenase and Candida boidinii formate dehydrogenase were from Sigma-Aldrich.

Synthesis and NMR analysis of 2,3-pentanediol stereoisomers.

To a stirred solution of 2,3-pentanedione (0.50 g, 5 mmol) in methanol (40 ml), a suspension of sodium borohydride (0.180 g, 4.8 mmol) in methanol (20 ml) was added over 60 min. After stirring for 3 h, the solvent was gently evaporated and the residue was treated with water (50 ml), neutralized, and extracted with chloroform (3 consecutive extractions with 30 ml). The solvent was removed obtaining 0.47 g (4.5 mmol; 90% yield) of 2,3-pentanediol. The residue was analyzed by ¹H and ¹³C nuclear magnetic resonance (NMR). To prepare (2R,3R)-2,3-pentanediol, we prepared a reaction mixture containing 48 mM 2,3-pentanedione, 1 mM NADH, 100 mM sodium formate, 25 U of Fdh, and 10 U of Bdh1p. The mixture was allowed to react for 23 h and then was extracted with chloroform. The chloroform was evaporated with a Rotavapor, and the oily extract was dissolved in deuterated chloroform before being analyzed by ¹H NMR.

Yeast and bacterial strains.

Escherichia coli XL1-Blue (Stratagene, La Jolla, CA) or DH5α was used for cloning experiments. The S. cerevisiae strains used in this study were derived from wild-type FY834α (MATα his3Δ200 ura3-52 leu2Δ1 lys2Δ202 trp1Δ63) (22) and from the Adh⁻ strain WV36-405 (MATa ura3-52 trp1 Δadh1 Δadh2 Δadh3 adh4::TRP1) constructed by Wolfgang Vogel (Neuherberg, Germany). Mutant strain EG2 (FY834α MATα his3Δ200 ura3-52 leu2Δ1 lys2Δ202 trp1Δ63 TRP1::bdh1) (19) was used to characterize additional NADH-dependent diacetyl reductases. Yeast strains with deletions of bdh1 (Δbdh1) and/or ara1 (Δara1) were constructed by the one-step gene replacement method (23) with a fragment containing the kanMX4 and/or natMX4 (24, 25) genes flanked by sequences identical to BDH1 and/or ARA1 (20).

Plasmids, DNA manipulations, cloning techniques, and transformation methods.

All DNA manipulations were performed under standard conditions as described previously (26). E. coli plasmid DNA was obtained by using a commercial kit provided by Sigma. The plasmids used to overexpress Bdh1p (pYES2-BDH1) and Bdh1p-(His)₆ (pYES2-BDH1-6His) have already been described (19, 20). The formate dehydrogenase gene (FDH1) from S. cerevisiae was amplified from yeast genomic DNA in a PCR with oligonucleotides o-FDH-fw and o-FDH-rv (Table 1). The band containing FDH1 was excised from an agarose gel, digested with BamHI and EcoRI, and cloned in the pYES2 vector digested with the same enzymes. The construct pYES2-FDH1 was sequenced to verify that no mutations had been introduced during the PCR. Plasmids pYES2-BDH1-6His and pYES2-FDH1 (see above) were used as starting points to construct the plasmid pYES2-FDH1-BDH1-6His, expressing the fusion protein containing Fdh1p in its N terminus and Bdh1p(His)₆ in its C terminus linked by the heptapeptide ENLYFQG. The gene coding for the fusion protein was constructed in two steps. First, we used oligonucleotides o-FDH-fw and FPint-Rv (Table 1) together with plasmid pYES2-FDH1 as a template, which in a PCR resulted in a fragment containing FDH1 fused to the sequence coding for the peptide linker (see above). Second, oligonucleotides FPint-Fw and bdh1-his together with plasmid pYES2-BDH1-6His as a template were used to construct by PCR a fragment that contained the sequence coding for Bdh1p(His)₆ fused to the aforementioned peptide linker. A mixture of these two fragments, containing a complementary region (underlined in the sequences in Table 1), was used as the template in a PCR with the external oligonucleotides o-FDH-fw and bdh1-his to obtain the complete gene coding for the fusion protein Fdh1p-Bdh1p(His)₆. The amplified fragment was excised from an agarose gel and cloned in pYES2 digested with BamHI and EcoRI. The correctness of the construct was verified by sequencing.

TABLE 1.

Oligonucleotide primers used in this study

Function	Primer name	Sequencea
Cloning of FDH1 and FDH1-BDH1-6His in vector pYES2	o-FDH-fw	5′ CGC GGA TCC AAT ATG TCG AAG GGA AAG G 3′
Cloning of FDH1 in vector pYES2	o-FDH-rv	5′ CCG GAA TTC TTA TTT CTT CTG TCC ATA AGC TCT GG 3′
Construction of the template for FDH1-BDH1-6His	FPint-Rv	5′ CAA AGC TCT CAT ACC TTG AAA ATA CAA ATT TTC TTT CTT CTG TCC ATA AGC TCT GG 3′
Construction of the template for FDH1-BDH1-6His	FPint-Fw	5′ CAG AAG AAA GAA AAT TTG TAT TTT CAA GGT ATG AGA GCT TTG GCA TAT TTC AAG AAG 3′
Cloning of FDH1-BDH1-6His in vector pYES2	bdh1-his	5′ GCG GAA TTC TTA ATG ATG ATG ATG ATG ATG CTT CAT TTC ACC GTG ATT GTT AGG 3′
Generation of strain WV36-405 URA3-NAT1::pbdh1-bdh1	Pbdh-ura-fw	5′ GAG CAG TCG GAA AGA TCA AGA AAG ACT ACG AGA ATC AAT AAA CGA GGC CAA TAC AAC AGA TCA CGT G 3′
Generation of strain WV36-405 URA3-NAT1::pbdh1-bdh1	Bdh-nat1-rv	5′ ATT AGT GAA GTG AAT ATC ACC CTT CTT GAA ATA TGC CAA AGC TCT CAT TCG ACA CTG GAT GGC GGC G 3′
Generation of strain WV36-405 PGAL1-BDH1	Pbdh-gal-Fw	5′GAG CAG TCG GAA AGA TCA AGA AAG ACT ACG AGA ATC AAT AAA CGA GGC ACG GAT TAG AAG CCG CCG AGC 3′
Generation of strain WV36-405 PGAL1-BDH1	Bdh-gal-bis-rv	5′ ATT AGT GAA GTG AAT ATC ACC CTT CTT GAA ATA TGC CAA AGC TCT CAT TAT CCG GGG TTT TTT CTC CTT GAC G 3′
Verification of the construct PGAL1-BDH1	o-gal1-fw	5′ CCG ACG GAA GAC TCT CCT CCG 3′
Verification of the construct PGAL1-BDH1	o-bdh1-rv	5′ GGT CCG CAC AAC TGC TGG CAG C 3′

^aThe restriction sites and self-complementary regions are underlined.

The plasmids were introduced into the yeast strains by the lithium acetate method (23), and the transformants were selected on synthetic complete (SC)-Ura medium supplemented with 2% glucose or galactose (in the case of the Adh⁻ strains).

Construction of a yeast strain overexpressing Bdh1p and Fdh1p by self-cloning.

A slight modification of the delitto perfetto technique (21) was performed to place BDH1 under the control of the GAL1 promoter in its chromosomal locus. The promoter and part of the BDH1 coding region from yeast strain WV36-405 were deleted, with a double marker (consisting of URA3 from Kluyveromyces lactis fused to natMX4 from Streptomyces noursei) obtained by PCR using plasmids pUG72 and pAG25 (obtained from Euroscarf) as templates. The oligonucleotides used to amplify the double marker by PCR were Pbdh-ura-fw and Bdh-nat1-rv (Table 1). The linear fragment obtained upon PCR (containing two flanking regions identical to the sequences from −520 to −473 and +1 to +48 from BDH1 and its promoter) was used to transform yeast strain WV36-405 by the lithium acetate method, and the transformants were selected on YPD agar medium (1% Bacto yeast extract, 2% Bacto peptone, 2% glucose-containing 2% agar) supplemented with 100 μg/ml of clonNAT (Hans-Knöll Institut, Leibniz-Institut für Naturstoff-Forschung und Infektionsbiologie, Jena, Germany). Several transformants that grew on clonNAT medium were checked by PCR to verify that the double marker was in the proper place; thus, yeast strain WV36-405 URA3-NAT1::P_bdh1-bdh1 was constructed. One of the transformants was grown in SC-Ura medium and transformed with a recombinogenic linear fragment containing the GAL1 promoter flanked by the sequence of the BDH1 promoter from −520 to −473 and by the sequence of BDH1 from +1 to +48. This fragment was obtained in a PCR with the oligonucleotides Pbdh-gal-Fw and Bdh-gal-bis-rv (Table 1) by using pYES2-BDH1 as a template. Upon transformation by the lithium acetate method, the mixture was plated on a YPD plate overnight. The plate was replica plated on an SC plate containing 100 μg/ml of uracil and supplemented with 1 mg/ml of 5-fluoroorotic acid (5-FOA) (27) the following day to allow for counterselection. The colonies that grew in this plate were replica plated on a YPD plate containing clonNAT, to discard spontaneous mutations of URA3 that allowed cells to grow on the 5-FOA plate. Several transformants growing on FOA, but not on clonNAT plates, were checked by PCR with oligonucleotides o-gal1-fw and o-bdh1-rv (Table 1), and one of the transformants was verified by sequencing. The new strain (WV36-405 P_bdh1::P_GAL1-BDH1) was transformed with plasmid pYES2-FDH1 and selected on SC-Ura plates, obtaining a yeast strain named DP-BDH1 overexpressing Bdh1p (because of the presence of its engineered chromosomal GAL1 promoter controlling BDH1) and Fdh1p (under the control of the GAL1 promoter contained in the pYES2 plasmid) on galactose. Crude extracts were prepared from this strain grown in SC-Ura medium plus 2% galactose until the beginning of the stationary phase. Permeabilized yeast DP-BDH1 cells were prepared by adding 0.1% (wt/vol) digitonin to the cell suspension, essentially as described previously (28).

Media and growth conditions.

E. coli was grown at 37°C in Luria-Bertani medium supplemented with 50 μg/ml of ampicillin to select for the desired plasmid constructs. The yeast strains derived from WV36-405 containing pYES2-derived plasmids were grown at 28°C on a rotatory shaker (250 rpm) in SC-Ura supplemented with 2% galactose.

Bdh1p, Bdh1p(His)₆, and Fdh1p-Bdh1(His)₆ purification and molecular weight determination.

Bdh1p, Bdh1p(His)₆, and Fdh1p-Bdh1(His)₆ were purified from yeast strains WV36-405 Δara1(pYES2-BDH1), WV36-405Δara1 Δbdh1(pYES2-BDH1-6His), and WV36-405Δara1 Δbdh1(pYES2-FDH1-BDH1-6His), respectively, as described previously (19, 20). The last step of the purification protocol served to determine the M_r of the fusion protein Fdh1p-Bdh1p(His)₆ by gel filtration chromatography. A Hi-Load (26/60) SuperDex 200 prep-grade column from GE Healthcare equilibrated with 50 mM HEPES (pH 7) and 150 mM NaCl was used. The column was eluted at a flow rate of 1.75 ml/min with the equilibration buffer.

Enzyme activities, chemical transformations, and coenzyme-regenerating systems.

Enzyme activities were determined spectrophotometrically by measuring the change of absorbance at 340 nm and 25°C, corresponding to the oxidation of NADH (ε₃₄₀ = 6,220 M⁻¹ · cm⁻¹). One unit of activity corresponds to 1 μmol of NAD⁺ formed per minute. To determine the steady-state parameters, the initial velocities were measured in duplicate at eight different substrate concentrations, and the k_cat and K_m for the different substrates (Table 2) were determined by using the nonlinear regression program Grafit 5.0 (Erithacus Software Ltd., Horley, United Kingdom). All reported values are the means ± standard errors (SEs) from at least three separate experiments. The specific activity of Bdh1p was measured in 33 mM sodium phosphate buffer at pH 7 in the presence of 50 mM (R,S)-acetoin and 0.2 mM NADH. To ascertain the stereoselectivities of the enzymes, reaction mixtures were prepared in 5 ml tubes with O-ring from Nirco (Barberà del Vallés, Spain) with continuous agitation at room temperature. The initial composition of the mixtures when using pure Bdh1p was 50 mM diketone, 200 U of Bdh1p, 1 mM NADH in 33 mM sodium phosphate (pH 7) buffer, and an NADH-regenerating system containing 100 mM sodium formate and 4 U of formate dehydrogenase from Candida boidinii (29). When using crude extracts, yeast strain DP-BDH1 [WV36-405 P_GAL1-BDH1(pYES2-FDH1)] was grown in SC-Ura plus galactose until the beginning of the stationary phase and the pelleted cells were disrupted with glass beads. The extracts, containing 150 U of Bdh1 and 3 U of Fdh activities, were incubated with 50 mM diketone together with 1 mM NADH and 100 mM formate. Yeast cells, permeabilized with 0.1% digitonin (28), were used to drive the reduction of 50 mM diketone in the presence of 2% galactose, 1 mM NADH, 100 mM formate, and 33 mM sodium phosphate (pH 7). When using the fusion protein Fdh1p-Bdh1p(His)₆, the reaction solution contained 39 U of Bdh and 0.3 U of Fdh, which were incubated with 50 mM diketone together with 5 mM NAD⁺ and 100 mM formate as an NADH-regenerating system. For the oxidation of the diols with pure Bdh1p, we used an NAD⁺-generating system containing α-ketoglutarate, ammonium chloride, and glutamate dehydrogenase in 33 mM sodium phosphate buffer at pH 7 (29).

TABLE 2.

Steady-state kinetic constants of yeast Bdh1p toward acetoin and several vicinal diketones^a

Compound(s)	kcat (s−1)	Km (mM)	kcat/Km (s−1 · M−1)
R- and S-acetoin	1,625 ± 83	3 ± 0.6	5.4 × 105 ± 1.3 × 105
Diacetyl	1,735 ± 165	60 ± 2	2.9 × 104 ± 0.4 × 104
2,3-Pentanedione	826 ± 83	33 ± 7	2.5 × 104 ± 0.8 × 104
2,3-Hexanedione	55 ± 8	6 ± 3	9 × 103 ± 6 × 103
3,4-Hexanedione	22 ± 1	0.7 ± 0.2	3 × 104 ± 1 × 104

^aAcetoin and diketone reduction activities were measured at 25°C in 33 mM sodium phosphate (pH 7.0) with 0.2 mM NADH.

Western blotting and in-gel assays of Bdh activity of yeast extracts.

Bdh1p(His)₆ and Fdh1p-Bdh1(His)₆ expressed from yeast strains WV36-405Δara1 Δbdh1(pYES2-BDH1-6His) and WV36-405Δara1BDH1(pYES2-FDH1-BDH1-6His) were detected by Western blotting using a mouse monoclonal anti-His₆ antibody according to a previously described method (20).

Yeast protein extracts from strains expressing Bdh1p(His)₆ and Fdh1p-Bdh1(His)₆ in different genetic backgrounds were loaded on precast (pH 3 to 9) isoelectric focusing (IEF) gels from Bio-Rad (Criterion). Their Bdh activities were visualized on the gels by incubating them with 100 mM (2R,3R)-2,3-butanediol and 2 mM NAD⁺, together with 0.08 mg/ml of phenazine methosulfate and 0.8 mg/ml of nitroblue tetrazolium, essentially as described previously (20).

Analytical methods.

The reaction mixtures were extracted with chloroform to recover the hydroxyketones and diols, together with 1-hexanol as an internal standard, essentially as described previously (30). A slight modification was introduced to quantify R- and S-acetoins and the different 2,3-butanediol stereoisomers obtained from the overnight reaction mixtures. To a 1-ml reaction mixture 1-hexanol was added up to 8 mM, together with 2.5 g of K₂CO₃, followed by two successive extractions with 2 and 4 ml of chloroform. Both extractions were necessary to attain linearity in the recovered acetoins and 2,3-butanediols from standard additions to the reaction mixtures. Alternatively, the analytes were extracted from the reaction mixtures with two successive ethyl acetate extractions (19). The different stereoisomers of the hydroxyketones and diols were resolved on a chiral column (Supelco β-DEX 120; 30-m length and 0.25-mm inner diameter) coupled to a Hewlett-Packard gas chromatograph (GC) equipped with a mass spectrophotometer as a detector, under conditions previously described (19). The identities of the products were verified by known standards (when available) and by mass spectrometry. To confirm the resolution between (3R)-hydroxy-2-pentanone and (2R)-hydroxy-3-pentanone, we also used an HP-5MS Agilent column (30-m length and 0.25-mm inner diameter) coupled to a Hewlett-Packard gas chromatograph equipped with a mass spectrophotometer as a detector. The following temperature program was used: isotherm at 45°C for 2 min and three ramps of 5°C/min up to 100°C, 15°C/min up to 150°C, and 30°C/min up to 275°C. The identities of the products were verified by mass spectrometry (MS).

RESULTS

Kinetic parameters and stereoselectivity of Bdh1p.

Table 2 gives the kinetic parameters determined for Bdh1p toward acetoin, diacetyl, 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione. The specificity constant k_cat/K_m decreased from acetoin to 2,3-hexanedione and then increased for 3,4-hexanedione. The reaction of purified Bdh1p with diacetyl specifically produced (2R,3R)-2,3-butanediol, with a 95% yield, and a minor amount of R-acetoin (Fig. 1A). To determine the stereoisomeric purity of the (2R,3R)-2,3-butanediol obtained, we added a known concentration of meso-2,3-butanediol and (2S,3S)-2,3-butanediol to half of the reaction mixture obtained upon the reaction of Bdh1p with diacetyl (and the NADH-regenerating system). An equal volume of buffer was added to the other half of the reaction. By comparing the areas of the known added concentrations of meso-2,3-butanediol and (2S,3S)-2,3-butanediol (Fig. 2B) and the corresponding ones in Fig. 2A, we estimated a stereoisomeric purity of (2R,3R)-2,3-butanediol greater than 99.6% (Fig. 2).

FIG 1.

Production of diols by purified Bdh1p. Gas chromatograms were obtained after extraction with chloroform of the products obtained from a reaction mixture containing 50 mM diketone, 1 mM NADH, 100 mM sodium formate, and 4 U of FDH in 33 mM sodium phosphate, pH 7, after the addition of 200 U of pure Bdh1p, together with 1-hexanol as an internal standard. Total reaction time was 20 h.

FIG 2.

Stereoisomer purity of the (2R,3R)-2,3-butanediol obtained with pure Bdh1p and diacetyl. Panels A and B correspond to the enlarged region of 2,3-butanediols stereoisomers before (A) and after (B) adding external (2S,3S)-2,3-butanediol and meso-2,3-butanediol. (A) Enlarged chromatogram corresponding to a reaction with diacetyl, Bdh1p, and an NADH-regenerating system where the concentration of the reaction product, (2R,3R)-2,3-butanediol, was 13.7 mM. (B) Enlarged chromatogram containing the reaction mixture from panel A, to which (2S,3S)-2,3-butanediol and meso-2,3-butanediol were added as controls. The final concentration of (2R,3R)-2,3-butanediol was 12.9 mM, while the concentrations of (2S,3S)-2,3-butanediol and meso-2,3-butanediol were 0.1 mM each.

When using Bdh1p with 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione, only one diol each was obtained (Fig. 1B to D), identified as 2,3-pentanediol, 2,3-hexanediol, and 3,4-hexanediol, respectively, by their mass spectra. Since there are not commercial sources of pure stereoisomers of these compounds that could be used as standards, we assigned the configuration of the stereoisomer of 2,3-pentanediol produced by Bdh1p by comparing the ¹H NMR spectrum obtained from 2,3-pentanediol obtained chemically with the one from the diol obtained with pure Bdh1p (see Fig. S1 in the supplemental material).

The ¹H and ¹³C NMR spectra of chemically prepared 2,3-pentanediol showed the presence of two diastereoisomers, (2RS,3RS) and (2RS,3SR) in their racemic forms, showing a good agreement with NMR data described before (31). Thus, the pair (2R,3S) plus (2S,3R) [abbreviated (2RS,3SR)] showed the chemical shifts at δ 3.81 ppm (for H-2), δ 3.55 (for H-3), δ 1.15 (for H-1), and δ 1.00 (for H-5) (see Fig. S1B). In its turn, the pair (2R,3R) plus (2S,3S) [abbreviated (2RS,3RS)] showed the chemical shifts at δ 3.61 ppm (for H-2), δ 3.27 (for H-3), δ 1.20 (for H-1), and δ 1.00 (for H-5) (see Fig. S1B). From the areas of the corresponding signals, the two isomers were obtained in a ratio of 2.14, with (2RS,3SR) being the major compounds (see Fig. S1B). When the same mixture of 2,3-pentanediols was loaded in a GC equipped with a chiral column, the four stereoisomers were resolved, with those that were eluted at retention times (t_R) of 14.86 and 15.51 min less abundant being than the other two (see Fig. S2 in the supplemental material). Consequently, these minority stereoisomers should correspond to the pair (2RS,3RS).

Moreover, when we prepared 2,3-pentanediol by the action of Bdh1p and the NADH-regenerating system, only one 2,3-pentanediol stereoisomer was obtained, which was eluted at 15.56 min (see Fig. S3 in the supplemental material) and with a NMR spectrum coincident with the one from the pair (2RS,3RS). The chemical shifts for this compound were at δ 3.58 ppm (for H-2), δ 3.24 (for H-3), δ 1.17 (for H-1), and δ 0.98 (for H-5) (see Fig. S1A). Since the NMR signals obtained for this compound are coincident with the ones from the pair (2RS,3RS), this stereoisomer should be (2R,3R)-2,3-pentanediol or (2S,3S)-2,3-pentanediol. Given the stereoselectivity displayed by Bdh1p toward diacetyl, we assign the peak eluting at 15.51 min to (2R,3R)-2,3-pentanediol (also see Discussion) and automatically the stereoisomer that was eluted at a t_R of 14.86 min to (2S,3S)-2,3-pentanediol (see Fig. S2).

Search for other yeast NADH-dependent diacetyl reductases.

A protein extract of a yeast strain with a deletion of bdh1, namely, EG2 (FY834α TRP1::bdh1) (19), grown in YPD displayed a diacetyl reductase activity of 0.73 U/mg (measured in 50 mM diacetyl and 0.2 mM NADH). The extract was loaded in an IEF gel, and the diacetyl reductase activities were visualized by activity staining with diacetyl and NADH (results not shown). A clear band with a pI lower than the pI of Bdh1p (Bdh1p was not visible in this gel since BDH1 was disrupted in the strain EG2) indicated that another enzyme (or enzymes) was able to reduce diacetyl in an NADH-dependent reaction. Therefore, we developed a protocol to purify this diacetyl reductase activity from the yeast strain EG2. The protocol consisted of a DEAE-Sepharose column followed by a hydroxyapatite column and by a Cibachron Blue 3GA chromatography. We obtained a homogeneous protein with an estimated M_r of approximately 42,000 (by SDS-PAGE), which upon digestion with trypsin was analyzed by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) (under conditions described by González et al. [20]). The protein was identified as Adh2p, and thus, we analyzed by GC-MS the products obtained in the reduction of diacetyl in the presence of Adh2p, which resulted in S-acetoin. Given the similarity between Adh2p, Adh1p, and Adh3p, we cloned and overexpressed Adh1p and Adh3p (by means of the galactose-inducible pYES2 vector) in the WV36-405 strain, finding that both displayed diacetyl reductase activity, also yielding S-acetoin. Figure S4 in the supplemental material shows that diacetyl treated with yeast Adh yielded S-acetoin. When the filtrate obtained upon this reduction was treated with Bdh1p, meso-2,3-butanediol was obtained (see Fig. S5). Moreover, 2,3-pentanedione treated with yeast Adh yielded (2S)-hydroxy-3-pentanone (see Fig. S6). When the filtrate obtained upon this reduction was treated with Bdh1p, (2S,3R)-2,3-pentanediol was obtained (see Fig. S7). The residual amount of 2,3-pentanedione from the filtrate yielded (3R)-hydroxy-2-pentanone and (2R,3R)-2,3-pentanediol (see Fig. S7). Thus, since Bdh1p is enantioselective, yielding R-acetoin from diacetyl and (2R,3R)-2,3-butanediol from R-acetoin, we decided to use the Adh-deficient strain WV36-405 when working with extracts and permeabilized cells to minimize the production of S-acetoin, which would yield meso-2,3-butanediol upon the action of Bdh1p.

Diols and hydroxyketones obtained from vicinal diketones by yeast extracts and permeabilized cells.

Although purified Bdh1p yields enantiopure vicinal diols upon the reduction of the corresponding diketones, an NADH-regenerating system was necessary to displace the reaction toward the diols. To avoid the need for an external source of formate dehydrogenase activity to regenerate NADH, we constructed a yeast strain that overexpressed Bdh1p and yeast formate dehydrogenase (Fdh1p), such that it would not be necessary to purify Bdh1p or to add exogenous formate dehydrogenase. As a first step to construct such a strain, we used a modification of the delitto perfetto technique (21) developed in our laboratory to change the chromosomal promoter of BDH1 by the GAL1-10 promoter. Then, a multicopy inducible plasmid with cloned FDH1 under the control of the GAL1 promoter was used to transform the previous yeast strain, such that the presence of galactose in the growth medium induced the Bdh and Fdh activities. The resulting yeast strain was named DP-BDH1 [WV36-405 P_GAL1-BDH1(pYES2-FDH1)]. To ascertain the overexpression of Bdh and Fdh activities in strain DP-BDH1, we grew it and its parental control strain [WV36-405 P_BDH1-BDH1(pYES2)] in galactose-containing medium. At the end of their logarithmic phases, we obtained the yeast extracts by disrupting the cells with glass beads and determined their Bdh and Fdh activities. While the Bdh and Fdh specific activities from the control strain, WV36-405 P_BDH1-BDH1(pYES2), were 0.26 and less than 0.001 U/mg, respectively [measured at 5 mM NAD⁺ and 100 mM (2R,3R)-butanediol or 100 mM formate], the corresponding ones from strain DP-BDH1 were 1.22 and 0.13 U/mg, respectively. Yeast strain DP-BDH1 was permeabilized with 0.1% digitonin by following the technique of Cordeiro and Freire (28) and incubated with the different diketones and galactose-containing medium. When using the protein extracts or the digitonin-permeabilized yeast cells of DP-BDH1, grown in galactose, on the different diketones, the products obtained were a mixture of all possible stereoisomers (Fig. 3 and 4). Thus, (2R,3R)-2,3-butanediol, meso-2,3-butanediol, and traces of (2S,3S)-2,3-butanediol were the products obtained upon the reduction of diacetyl with permeabilized cells, with a yield of 65% (adding the 3 butanediol stereoisomers) (Fig. 4). Furthermore, the additional enzymatic activities present in the yeast extracts and permeabilized cells yielded several diol stereoisomers from 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione (Fig. 3 and 4). For instance, the four stereoisomers of 2,3-hexanediol could be detected from their mass spectra upon the reaction of 2,3-hexanedione with DP-BDH1 extracts or permeabilized cells (Fig. 3C and 4C).

FIG 3.

Production of diols by yeast extracts overexpressing Bdh1p and Fdh1p. Gas chromatograms were obtained after extraction with chloroform of the products obtained from a reaction mixture containing 50 mM diketone, 1 mM NADH, and 100 mM sodium formate in 33 mM sodium phosphate, pH 7, 20 h after the addition of an extract of yeast strain DP-BDH1 containing 3 U of Fdh and 150 U of Bdh together with 1-hexanol as an internal standard.

FIG 4.

Production of diols by permeabilized yeast cells overexpressing Bdh1p and Fdh1p. Gas chromatograms were obtained after extraction with chloroform of the products obtained from a reaction mixture containing 2% galactose, 50 mM diketone, 5 mM NAD⁺, and 100 mM sodium formate in 33 mM sodium phosphate, pH 7, after the addition of permeabilized DP-BDH1 yeast cells together with 1-hexanol as an internal standard. Total reaction times were 1 h for diacetyl and 2,3-pentanedione and 20 h for 2,3-hexanedione and 3,4-hexanedione.

Identification of the stereoisomeric forms of the hydroxyketones and vicinal diols.

In the case of the acetoin and 2,3-butanediol stereoisomers obtained upon reduction of diacetyl (Fig. 3A and 4A), the assignment of the stereoisomeric forms was made by comparison with the retention times of the pure (2R,3R)-2,3-butanediol, meso-2,3-butanediol, and (2S,3S)-2,3-butanediol and also after the oxidation of pure (2R,3R)-2,3-butanediol to R-acetoin (19) and of meso-2,3-butanediol to S-acetoin. In the case of 2,3-pentanedione (Fig. 4B), the MS of the peaks that were eluted at 7.06 and 7.77 min (m/z 59, 43, and 31, in decreasing abundance) identified them as 3-hydroxy-2-pentanones (32), while the peak that was eluted at 7.64 min (m/z 45, 57, and 43, in decreasing abundance) was identified as 2-hydroxy-3-pentanone (32). The stereoisomeric form that was eluted at a t_R of 7.06 min was assigned to (3R)-hydroxy-2-pentanone since its retention time was the same as the product obtained with pure Bdh1p and 2,3-pentanedione (after 1 h of reaction only). Thus, the 3-hydroxy-2-pentanone that was eluted at 7.77 min must be (3S)-hydroxy-2-pentanone. The 2-hydroxy-3-pentanone that was eluted at 7.64 min was assigned to (2S)-hydroxy-3-pentanone, since this retention time was the same as that for the product obtained upon reduction of 2,3-pentanedione with pure Ara1p (33). Further evidence for the assignments of the hydroxyketones was obtained from the oxidations of (2R,3R)-2,3-pentanediol and (2R,3R)-2,3-hexanediol with pure Bdh1p (see below).

The 2,3-pentanediol peaks observed upon reduction of 2,3-pentanedione with yeast extracts or permeabilized cells overexpressing Bdh1p and Fdh1p (Fig. 3B and 4B) were assigned according to the signal obtained with pure Bdh1p on 2,3-pentanedione (see above) for (2R,3R)-2,3-pentanediol (eluted at t_R = 16.87 min) (Fig. 4B). The identities of the stereoisomers (2R,3S)-2,3-pentanediol and (2S,3R)-2,3-pentanediol were established by treating 2,3-pentanedione with Ara1p followed by Bdh1p: by our previous work (33), we knew that (2S)-hydroxy-3-pentanone should be obtained from Ara1p and (2S,3R)-2,3-pentanediol after treating this hydroxyketone with Bdh1p. This compound corresponds to the peak which was eluted at a t_R of 18.19 min (Fig. 4B), so the peak which was eluted at a t_R of 18.07 min should be (2R,3S)-2,3-pentanediol (Fig. 4B). No (2S,3S)-2,3-pentanediol was observed in these experiments.

With regard to the 2,3-pentanediol stereoisomers obtained chemically, the identities of the different stereoisomers are marked in Fig. S2 in the supplemental material. The elution times are different from the ones obtained in the reduction of 2,3-pentanedione by yeast extracts (and permeabilized cells) because the reaction mixtures were extracted with ethyl acetate and were run at different days.

The same experimental strategy and reasoning were followed to assign the hydroxyketones and 2,3-hexanediols derived from 2,3-hexanedione (Fig. 4C). First, the four peaks which were eluted from a t_R of 20 to 21 min were assigned to 2,3-hexanediol by their MS (with m/z 55, 73, 45, and 43, in decreasing abundance), characteristic of 2,3-hexanediols (34). Second, the treatment of 2,3-hexanedione with pure Bdh1p allowed us to assign (2R,3R)-2,3-hexanediol to the peak at a t_R of 20.25 min, and the treatment of 2,3-hexanedione with Ara1p followed by Bdh1p allowed us to assign (2S,3R)-2,3-hexanediol to the 20.74-min peak (Fig. 4C). The remaining stereoisomers were assigned as in the case of the reduction of 2,3-pentanedione. In fact, the same order of elution of the 2,3-hexanediol stereoisomers was also obtained by Schröder et al., working with a β-cyclodextrin chiral column (34).

In the case of the diols produced upon the reduction of 3,4-hexanedione (Fig. 4D), the assignation of the stereoisomeric forms was made as in the case of diacetyl reduction.

Properties of the fusion protein Fdh1p-Bdh1p(His)₆.

The fusion protein Fdh1p-Bdh1p(His)₆ was purified from the yeast strain WV36-405Δara1 Δbdh1(pYES2-FDH1-BDH1-6His) by means of an Ni-Sepharose column, followed by gel filtration on a Hi-Load SuperDex 200 that also served to determine its M_r. The estimated M_rs for Fdh1p-Bdh1p(His)₆ and Bdh1p(His)₆ were 133,000 and 70,500, respectively (Fig. 5C). Given the M_rs of Fdh1p (M_r = 41,700) and Bdh1p (M_r = 41,500) monomers, the active form of the fusion protein is a dimer of two Fdh1-Bdh1p(His)₆ polypeptides. The fusion protein is rather unstable, as can be deduced from the Western blot analysis (Fig. 5A, lane 2). A main band of an apparent M_r close to 95,000 [consistent with the M_r of an Fdh1p-Bdh1p(His)₆ polypeptide] and several bands (presumably degradation products) of lower M_rs are visible in the blot. Under these conditions, Bdh1p(His)₆ is stable, as only one band of the correct M_r is visible in the blot (Fig. 5A, lanes 3 and 4). Furthermore, the presence of Fdh1p in the fusion protein decreases the pI of Bdh1p(His)₆p (compare lanes 3 and 4 with lane 2 in Fig. 5B). Thus, the activity bands seen in the gel (Fig. 5B, lane 2) confirms that Fdh1p-Bdh1p(His)₆ had a lower pI than Bdh1p(His)₆ in two genetic backgrounds, with endogenous Bdh1p (Fig. 5B, lane 3) and without endogenous Bdh1p (Fig. 5B, lane 4). Moreover, we determined the kinetic constants of Fdh1p-Bdh1p(His)₆ toward R- and S-acetoins as a K_m of 2.2 mM (determined at 0.2 mM NADH) and k_cat of 31 s⁻¹. From the specific activity of Fdh1p-Bdh1p(His)₆ toward formate (at 5 mM NAD⁺), an apparent k_cat of 0.25 s⁻¹ was estimated at 25°C. Thus, the k_cat of the fusion protein for acetoin was approximately 50 times lower than the k_cat of Bdh1p (Table 2), while the K_m values were similar. The k_cat for formate was 24 times lower than the published value obtained at 30°C (35).

FIG 5.

Determination of the molecular properties of the fusion protein Fdh1p-Bdh1p(His)₆. (A) Western blot analysis of yeast protein extracts developed with an anti-His antibody. Lane 1, M_r standards; lane 2, 15 μg from strain WV36-405Δara1 Δbdh1(pYES2-FDH1-BDH1-6His) extract; lane 3, 17 μg from strain WV36-405 Δara1BDH1(pYES2-BDH1-6His) extract; lane 4, 21 μg from strain WV36-405 Δara1 Δbdh1(pYES2-BDH1-6His) extract. (B) IEF gel (pH 3 to 9) stained with 100 mM (2R,3R)-2,3-butanediol. Lane 1, pI standards; lane 2, 30 mU of butanediol dehydrogenase activity from strain WV36-405 Δara1 Δbdh1(pYES2-FDH1-BDH1-6His) extract; lane 3, 7 mU of Bdh from yeast strain WV36-405 Δara1 BDH1(pYES2-BDH1-6His) extract; lane 4, 6 mU of Bdh from strain WV36-405 Δara1 Δbdh1(pYES2-BDH1-6His) extract. (C) Size exclusion chromatography of purified Bdh1p(His)₆ and fusion protein Fdh1p-Bdh1p(His)₆ on a SuperDex 200 prep-grade column. a, β-amylase (200 kDa); b, yeast alcohol dehydrogenase (150 kDa); c, bovine serum albumin (66 kDa); d, carbonic anhydrase (29 kDa); e, cytochrome c (12.4 kDa).

Hydroxyketones and diols produced by the fusion protein Fdh1p-Bdh1p(His)₆.

The purified fusion protein Fdh1p-Bdh1p(His)₆ yielded a single form of diol, namely, (2R,3R)-2,3-butanediol, upon diacetyl reduction (Fig. 6A), with a conversion of 31%. Moreover, when starting with either 2,3-pentanedione, 2,3-hexanedione, or 3,4-hexanedione, a single stereoisomeric form of diol was obtained also (Fig. 6B to D). Since these diols have the same retention times as the diols obtained with pure Bdh1p, they have been tentatively assigned as the corresponding (R,R) stereoisomers. The reduction of the diketones by Fdh1p-Bdh1p(His)₆ also yielded hydroxyketones. In the case of diacetyl, the hydroxyketone was identified as R-acetoin, while in the case of 2,3-pentanedione and 2,3-hexanedione, they were identified by their mass spectra as 3-hydroxy-2-pentanone and 3-hydroxy-2-hexanone, respectively. We assigned them the R configuration (Fig. 6) because we assume that for these compounds, the fusion protein displays the same enantioselectivity as shown toward diacetyl.

FIG 6.

Stereoisomeric composition of the products obtained from reaction mixtures containing vicinal diketones and purified fusion protein Fdh1p-Bdh1p(His)₆. Shown are stereoisomer compositions of the products obtained with a reaction mixture containing purified fusion protein Fhd1p-Bdh1p(His)₆, 37 U of Bdh activity, 0.3 U of Fdh activity, 5 mM NAD⁺, and100 mM sodium formate in 33 mM sodium phosphate buffer at pH 7, together with 50 mM concentrations of the compounds diacetyl (A), 2,3-pentanedione (B), 2,3-hexanedione (C), and 3,4-hexanedione (D). Total reaction time was 20 h.

Oxidation of the vicinal diols with Bdh1p.

We first studied the products obtained in the oxidation reaction of (2R,3R)-2,3-butanediol by Bdh1p in the presence of NAD⁺ and an NAD⁺-regenerating system consisting of α-ketoglutarate, ammonium chloride, and glutamate dehydrogenase. We obtained R-acetoin as expected from the oxidation of (2R,3R)-2,3-butanediol (results not shown). Since (2R,3R)-2,3-pentanediol, (2R,3R)-2,3-hexanediol, and (3R,4R)-3,4-hexanediol are not commercially available, they were obtained in situ from the reaction products upon reduction of 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione by Bdh1p [presumably the (R,R)-stereoisomers; see above]. When using (2R,3R)-2,3-pentanediol and NAD⁺ with Bdh1p, together with the NAD⁺-regenerating system, one peak was obtained, with an MS pattern compatible with 2-hydroxy-3-pentanone (Fig. 7C). We also analyzed the reaction products obtained upon the reduction of 2,3-pentanedione with Bdh1p (after a 1-h reaction) and observed a peak, identified as 3-hydroxy-2-pentanone (Fig. 7A). The insets in Fig. 7A and C show only the most abundant and selective ions for (3R)-hydroxy-2-pentanone (m/z 59 [black]) and (2R)-hydroxy-3-pentanone (m/z 45 [red]) for clarity. Both compounds should have the R-configuration and show close retention times on the chiral column. We obtained an additional proof of their identities by loading them (as a chloroform extract) on an HP-5MS column, which resolved two peaks corresponding to (3R)-hydroxy-2-pentanone and (2R)-hydroxy-3-pentanone (results not shown). Furthermore, a mixture of the four stereoisomeric forms of 2,3-pentanediol was oxidized with Bdh1p, and the most reactive species were (2R,3R)-2,3-pentanediol, yielding essentially (2R)-hydroxy-3-pentanone (t_R = 6.36 min) but also a minor amount of (3R)-hydroxy-2-pentanone (t_R = 6.31 min) and (2S,3R)-2,3-pentanediol, which yielded (2S)-hydroxy-3-pentanone (t_R = 6.76 min). A minor amount of (3S)-hydroxy-2-pentanone (t_R = 6.91 min) was obtained from the oxidation of (2R,3S)-2,3-pentanediol by Bdh1p (see Fig. S8 in the supplemental material). We did a parallel study with (2R,3R)-2,3-hexanediol, with similar results. Thus, the reduction of 2,3-hexanedione yielded (3R)-hydroxy-2-hexanone (Fig. 8A and C), while the oxidation of (2R,3R)-2,3-hexanediol yielded (2R)-hydroxy-3-hexanone (Fig. 8B and D). Figure 8C and D show only the most abundant and selective ions for 3-hydroxy-2-hexanone (m/z 55 [black]) and 2-hydroxy-3-hexanone (m/z 45 [red]). However, since the reduction of 2,3-hexanedione was not complete (Fig. 8A), some (3R)-hydroxy-2-hexanone was carried over together with (2R,3R)-2,3-hexanediol, which is visible in the chromatogram (Fig. 8B). With all these data on the oxidation and reduction of vicinal diols and diketones with Bdh1p, we propose as a mechanism that the more reactive carbon upon oxidation (of the diol) and reduction (of the diketone) would be that corresponding to the most internal hydroxyl (or keto) group (Fig. 9). Furthermore, in order to resolve all the stereoisomeric forms of 2-hydroxypentanone and 3-hydroxypentanone (as well as the corresponding hydroxyhexanones), we induced the racemization and isomerization through enolate species of the hydroxyketones by adding NaOH to the products obtained upon oxidation of (2R,3R)-2,3-pentanediol and (2R,3R)-2,3-hexanediol (or reduction of the diketones). Figure 10 shows that the chiral column successfully resolved all the hydroxyketone stereoisomers.

FIG 7.

Stereoisomeric composition of the products from the reduction andoxidation of 2,3-pentanedione/(2R,3R)-2,3-pentanediol with Bdh1p. (A) Stereoisomer composition of the products obtained in a reaction mixture containing Bdh1p, 50 mM 2,3-pentanedione, 1 mM NADH, and an NADH-regenerating system (100 mM sodium formate and 1 U of formate dehydrogenase) in 33 mM sodium phosphate, pH 7, after the addition of 40 U of pure Bdh1p together with 1-hexanol as an internal standard, with a total reaction time of 1 h. The inset shows the relative abundances of the ions with m/z 59 (in black) and 45 (in red) for the peak that was eluted at 6.1 min from the chiral column. (B) Stereoisomer composition of the products obtained in a reaction mixture containing Bdh1p, 50 mM 2,3-pentanedione, and the same components of reaction solution as for panel A, after the addition of 40 U of pure Bdh1p together with 1-hexanol as an internal standard, with a total reaction time of 20 h. (C) Stereoisomer composition of the products obtained in a reaction mixture containing Bdh1p, (2R,3R)-2,3-pentanediol, 5 mM NAD⁺, and an NAD⁺-regenerating system (50 mM α-ketoglutarate, 100 mM ammonium chloride, and 6 U of glutamate dehydrogenase) in 33 mM sodium phosphate, pH 7, after the addition of 6 U of Bdh1p together with 1-hexanol as an internal standard, with a total reaction time of 20 h. The inset shows the relative abundances of the ions with m/z 59 (in black) and 45 (in red) for the peak that was eluted at 7.1 min from the chiral column. The differences in retention times were due to running some samples on different days.

FIG 8.

Stereoisomeric composition of the products from the reduction and oxidation of 2,3-hexanedione/(2R,3R)-2,3-hexanediol with Bdh1p. (A) Stereoisomer composition of the products obtained in a reaction mixture containing Bdh1p, 50 mM 2,3-hexanedione, 1 mM NADH, and an NADH-regenerating system (100 mM sodium formate and 3 U of formate dehydrogenase) in 33 mM sodium phosphate, pH 7, after the addition of 40 U of Bdh1p together with 1-hexanol as an internal standard, with a total reaction time of 20 h. (B) Stereoisomer composition of the products obtained in a reaction mixture containing Bdh1p and (2R,3R)-2,3-hexanediol, 5 mM NAD⁺, and an NAD⁺-regenerating system (50 mM α-ketoglutarate, 100 mM ammonium chloride, and 6 U of glutamate dehydrogenase) in 33 mM sodium phosphate, pH 7, after the addition of 6 U of Bdh1p with 1-hexanol as an internal standard, with a total reaction time of 20 h. (C) Relative abundances of the ions with m/z 55 (in black) and 45 (in red), for the peak that was eluted at 11 min from the chiral column (from panel A). (D) Relative abundancies of the ions with m/z 45 (in red) and 55 (in black) for the peak that was eluted at 10.7 min from the chiral column (from panel B).

FIG 9.

Reaction schemes for the oxidation and reduction of (2R,3R)-2,3-pentanediol and 2,3-pentanedione by Bdh1p. The scheme accounts for the relative abundances of (R)-3-hydroxy-2-pentanone and (R)-2-hydroxy-3-pentanone in the reduction of 2,3-pentanedione and in the oxidation of (2R,3R)-2,3-pentanediol, with the C-3 position being the most reactive center in both directions. The dashed arrows correspond to putative pathways that although not observed would be difficult to rule out completely.

FIG 10.

Separation of the four hydroxy-pentanones and four hydroxy-hexanones obtained upon racemization and isomerization of the products obtained from the oxidations of (2R,3R)-2,3-pentanediol and (2R,3R)-2,3-hexanediol by Bdh1p. (A) A 40 mM concentration of 2,3-pentanedione and a 40 mM concentration of 2,3-hexanedione were separately reduced by 6 U of Bdh1p in the presence of 1 mM NADH, 100 mM sodium formate, and 4 U of formate dehydrogenase from Candida boidinii in 33 mM sodium phosphate buffer, pH 7. After 20 h of reaction, the products were recovered by filtration on an Amicon column, and the filtrates [containing the (2R,3R)-2,3-diols] were subjected to an overnight oxidation by 6 U of Bdh1p in the presence of 5 mM NAD⁺, 50 mM α-ketoglutarate, 100 mM ammonium chloride, and 6 U of glutamate dehydrogenase. The final reaction products were made at 1 N in NaOH to induce racemization and isomerization and were loaded (after chloroform extraction) on a β-cyclodextrin gas chromatography column coupled to a mass spectrophotometer. (B) Relative abundances of the ions with m/z 59 (black) and 45 (red) for the peaks that were eluted at 5.67, 5.72, 6.16, and 6.27 min from the chiral column, labeled 1, 2, 3, and 4. (C) Relative abundances of the ions with m/z 55 (blue) and 45 (red) for the peaks that were eluted at 9.03, 9.21, 9.59, and 10.31 min from the chiral column, labeled 5, 6, 7, and 8.

DISCUSSION

The main aims of the present work were to develop a system able to produce (R,R)-diols from vicinal diketones, to determine the stereoselectivity of Bdh1p toward the reduction of the two carbonyl groups, to determine the kinetic constants of the reactions, and to elucidate its preference toward the keto groups of the vicinal diketones. In the first part of the study, we showed that pure Bdh1p reduced both carbonyl groups from diacetyl to stereoisomerically pure (2R,3R)-2,3-butanediol (Fig. 2A and B). Furthermore, when using 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione with pure Bdh1p, only one stereoisomeric form of the corresponding diols was identified in the reaction mixtures. Given the stereoselectivity displayed by Bdh1p toward diacetyl and since there are not commercial sources of those pure stereoisomers, we tentatively assigned them to the corresponding (R,R) stereoisomers. We performed, however, NMR and GC experiments to assign the configuration of the 2,3-pentanediol stereoisomer produced by Bdh1p from 2,3-pentanedione. There are several reasons to conclude that the stereoisomer is (2R,3R)-2,3-pentanediol. Thus, the NMR experiments on the four 2,3-pentanediol stereoisomers obtained chemically show the signals from the pair (2R,3S) plus (2S,3R) and the pair (2R,3R) plus (2S,3S). The NMR signals show that the pair (2R,3R) plus (2S,3S) is in a ratio twice as low as that of the pair (2R,3S) plus (2S,3R). Since the same mixture loaded in a GC showed two peaks in a ratio twice as low as that of the two other peaks, those peaks correspond to the pair (2R,3R) plus (2S,3S). Thus, the results from the NMR data alone do not discard a mixture of (2R,3R)-2,3-pentanediol and its enantiomer (2S,3S)-2,3-pentanediol on the reduction of 2,3-pentanedione by Bdh1p (see Fig. S1 in the supplemental material). However, only one stereoisomer was obtained when the reaction mixture was loaded in the chiral column (see Fig. S2). Consequently, the product can be (2R,3R)-2,3-pentanediol or (2S,3S)-2,3-pentanediol. Given the stereospecificity shown by Bdh1p toward the production of (2R,3R)-2,3-butanediol from diacetyl, the production of pure (2S,3S)-2,3-pentanediol from 2,3-pentanedione, without any production of the (2R,3S)-2,3-pentanediol or (2S,3R)-2,3-pentanediol, seems very unlikely.

The steady-state kinetic studies performed with pure Bdh1p and several vicinal diketones show that the specificity constant k_cat/K_m decreased from diacetyl to 2,3-hexanedione and then increased for 3,4-hexanedione (Table 2). Bdh1p displayed the highest specificity constants toward the substrates known to be transformed by Bdh1p in the Saccharomyces cerevisiae metabolism, i.e., acetoin, diacetyl, and 2,3-pentanedione.

Although the use of pure Bdh1p yielded only one stereoisomeric form of the diols, it was necessary to use formate and exogenous formate dehydrogenase to regenerate NADH. To avoid this drawback, we explored the possibility of using a yeast strain derived from FY834 (22) which could overexpress their endogenous BDH1 and FDH1. To choose a convenient genetic background, we tested first whether there were any diacetyl reductases other than Bdh1p in the yeast extracts derived from the wild-type strain FY834. We detected Adh2p with an NADH-dependent reductase activity that yielded S-acetoin from diacetyl. We also showed that Adh1p and Adh3p could also reduce diacetyl to S-acetoin in NADH-dependent reactions. Furthermore, with a commercial preparation of yeast Adh, we also obtained S-acetoin from diacetyl (see Fig. S4 in the supplemental material). Consequently, we decided to work with a yeast strain with the genetic background Δadh1 Δadh2 Δadh3 (WV36-405) to avoid the production of meso-2,3-butanediol, which would be obtained from S-acetoin by Bdh1p (19) (see Fig. S5). We developed a modified delitto perfetto technique (21) to overexpress Bdh1p and Fdh1p in strain WV36-405. The counterselection marker used, URA3 from Kluyveromyces lactis, which is different in sequence from URA3 from S. cerevisiae, minimized the homologous recombination events to the endogenous URA3 locus, while the selection marker NAT1 yields fewer spontaneous mutants than the kanMX4 marker (25). Thus, by placing the GAL1-10 promoter controlling BDH1 and transforming the resultant strain with a multicopy galactose-inducible plasmid carrying FDH1, we obtained a modified yeast strain (named DP-BDH1) that overexpressed Bdh1p and Fdh1p in the presence of galactose. We measured a 4.5-fold increase in Bdh specific activity toward (2R,3R)-2,3-BD and a >100-fold increase in Fdh specific activity toward formate in the corresponding protein extracts in comparison to the values of the parental strain transformed with an empty plasmid. We used this engineered strain in the form of protein extracts and permeabilized cells to drive the reductions of diketones without exogenously added Fdh. Although the diketones were reduced, the obtained diols were a mixture of different stereoisomers. Thus, with diacetyl, meso-2,3-butanediol and (2R,3R)-2,3-butanediol were identified as products by using protein extracts or permeabilized cells (Fig. 3 and 4). A mixture of different stereoisomeric forms of 2,3-pentanediol, 2,3-hexanediol, and 3,4-hexanediol were obtained in the reductions of 2,3-pentanedione, 2,3-hexanedione, and 3,4-hexanedione, respectively (Fig. 3 and 4). Although it was advantageous to use DP-BDH1 with no need for coenzyme-regenerating processes, the diastereoisomeric composition of the products precludes its use to obtain pure (R,R)-diols. Mixtures of stereoisomeric alcohols are formed because of the presence of other enzymes that could convert the same substrates but with different stereoselectivities. Thus, the activities of other diketone reductases (NADH and also NADPH dependent in the permeabilized cells) in yeast would be responsible for the production of S-hydroxyketones and the different stereoisomeric forms of the diols (20, 32). In fact, the reduction of diacetyl and 2,3-pentanedione for only 1 h allowed the observation of S-acetoin, (2S)-hydroxy-3-pentanone, and (3S)-hydroxy-2-pentanone (Fig. 4A and B). Then, Bdh1p and/or other reductases would reduce these mixtures of hydroxyketones to the corresponding mixture of diols. A mixture of stereoisomeric alcohols derived from the reduction of α- and β-ketoesters has previously been observed in work with yeast cells (17).

To avoid these interfering activities and the use of exogenous Fdh, we constructed a bifunctional fusion protein containing Fdh and Bdh activities with a His tag at its C terminus, Fdh1p-Bdh1p(His)₆. By using this protein, pure (2R,3R)-2,3-butanediol, (2R,3R)-2,3-pentanediol, (2R,3R)-2,3-hexanediol, and (3R,4R)-3,4-hexanediol (Fig. 6) were obtained, but with lower yields [31% with (2R,3R)-2,3-butanediol] than when using pure Bdh1p and exogenous formate dehydrogenase [95% conversion of diacetyl into (2R,3R)-2,3-butanediol]. The fusion Fdh1p-Bdh1p(His)₆ behaved as a dimer in a gel filtration column (Fig. 5C), consistent with the fact that Bdh1p is a dimer (19) and the orthologous formate dehydrogenase from Candida boidinii is also a dimer (36). The stability of the fusion protein was checked by Western blotting and zymogram analyses (Fig. 5A and B), showing that the protein suffered proteolysis but maintained Bdh activity on the zymogram. The use of this fusion protein allowed the production of pure (R,R)-diols from vicinal diketones, without the need for adding an external cofactor-regenerating system, but with a low yield. In fact, the k_cat values of the Bdh and Fdh activities were 50 and 24 times lower, respectively, for the fusion protein than for the individual enzymes. However, this strategy allowed the rapid purification of the fusion protein and the production of enantiopure (R,R)-diols. However, further work needs to be done to increase the stability of the fusion protein and the expression level and/or the turnover rate of Fdh1p. A similar approach for cofactor regeneration was used recently by Hölsch and Weuster-Botz (37) to convert a prochiral ketone to (S)-1-(pentafluorophenyl)-ethanol. They used a bifunctional protein composed of an NADP⁺-dependent mutant of formate dehydrogenase linked to an NADPH reductase in the whole-cell reduction of pentafluoroacetophenone. In their case, however, the activity of the oxidoreductase was almost not affected by the fusion.

A different genetic engineering approach was taken by Kim et al. (8) and by Lian and coworkers (13) that constructed pyruvate decarboxylase mutant yeast strains as a starting point to develop high-titer producers of 2,3-butanediol. Thus, more than 96.2 g/liter of 2,3-butanediol by fed-batch fermentation with glucose (8) and more than 100 g/liter of (2R,3R)-2,3-BD from glucose and galactose were obtained (13). However, those engineered strains did not produce (2R,3R)-2,3-pentanediol, (2R,3R)-2,3-hexanediol, or (3R,4R)-3,4-hexanediol.

We also studied the specificity of Bdh1p toward the two hydroxy groups of (2R,3R)-2,3-pentanediol and (2R,3R)-2,3-hexanediol upon oxidation with NAD⁺, as well as for the reverse reaction, the reduction of 2,3-pentanedione and 2,3-hexanedione with NADH. The reduction of 2,3-pentanedione after 1 h of reaction yielded (3R)-hydroxy-2-pentanone as an intermediate (Fig. 7A). When the reduction reaction was complete (Fig. 7B), the selectivity was studied in the oxidation direction, yielding (2R)-hydroxy-3-pentanone. Furthermore, we prepared the four 2,3-pentanediol stereoisomers by borohydride reduction of 2,3-pentanedione, finding that the most abundant species (from their NMR signals) were the enantiomeric pair (2R,3S)-2,3-pentanediol and (2S,3R)-2,3-pentanediol (see Fig. S1 in the supplemental material), which, when loaded into the chiral column, yielded the two most abundant peaks with longer retention times (see Fig. S2). To characterize the configurations of these two stereoisomers, we reduced 2,3-pentanedione with Ara1p (the aldo-keto reductase AKR31C), which produced (2S)-hydroxy-3-pentanone (our work [33]). The reduction of this hydroxyketone with Bdh1p gave a single peak that given the stereoselectivty of Bdh1p should be (2S,3R)-2,3-pentanediol, with a retention time of 16.87 min (results not shown). Consequently, we could assign the remaining peak at a t_R of 16.63 min to (2R,3S)-2,3-pentanediol. The identities of the peaks which were eluted at t_Rs of 14.86 and 15.51 min were assigned from the reduction of 2,3-pentanedione with Bdh1p, obtaining only one stereoisomer that corresponds to the one that was eluted at a t_R of 15.51 min (see Fig. S3). After our previous discussion (see above), we can safely assume that this stereoisomer should be (2R,3R)-2,3-pentanediol and the one that was eluted at a t_R of 14.86 min should then be (2S,3S)-2,3-pentanediol. Upon oxidation of the mixture of the four 2,3-pentanediol stereoisomers, it was observed that the most reactive species were (2R,3R)-2,3-pentanediol and (2S,3R)-2,3-pentanediol (see Fig. S8), yielding (2R)-hydroxy-3-pentanone and (2S)-hydroxy-3-pentanone and indicating the preference toward the oxidation of the most internal carbon with R configuration.

A result consistent with these observations was obtained when studying the oxidation and reduction of (2R,3R)-2,3-hexanediol and 2,3-hexanedione by Bdh1p. The enzyme showed a preference for the functional group bound to the most internal carbon, namely, C-3, in the oxidation and reduction reactions (Fig. 8). Thus, 3-hydroxy-2-hexanone was identified in the reduction direction and 2-hydroxy-3-hexanone in the oxidation direction by its mass spectra. With the characterization of all these intermediates and products, we propose the mechanism shown in Fig. 9, which indicates that the most internal carbon (C-3 for 2,3-pentanedione and hexanedione) is the most reactive, since it is the one that predominantly reacts in both directions. Although we did not detect (2R)-hydroxy-3-pentanone in the reduction direction of 2,3-pentanedione by Bdh1p (Fig. 7A), it would be difficult to discard it completely (Fig. 9, dashed arrows). For the same reason, we drew a discontinuous line in the oxidation of (2R,3R)-2,3-pentanediol to (3R)-hydroxy-2-pentanone, which is much less important than the oxidation of (2R,3R)-2,3-pentanediol to (2R)-hydroxy-3-pentanone (see Fig. S8 in the supplemental material). Since no three-dimensional (3D) structure of any (2R,3R)-2,3-butanediol dehydrogenase has been published yet, it would be speculative to give a structural reason for this preference. A recent article (38) reports the crystallization of the orthologous (2R,3R)-2,3-butanediol dehydrogenase from Bacillus coagulans; therefore, we will be able to discuss more deeply the differential reactivity of both stereogenic centers once the 3D structure of the enzyme is solved.

Finally, we showed the resolution power of the β-cyclodextrin column that was able to resolve the four hydroxyketones derived from 2,3-pentanedione and the four derived from the 2,3-hexanedione (Fig. 10). Although R-3-hydroxy-2-pentanone and R-2-hydroxy-3-pentanone were not completely resolved with the chiral column under the conditions used, we obtained a better resolution and confirmed their identities by the use of an HP-5MS column (results not shown).