Introduction

Dehydrogenases/reductases are found throughout across a wide range of organisms where they are involved in a broad spectrum of metabolic functions (Jörnvall 2008), and a system of short-, medium- and long-chain dehydrogenase/reductases has been recently identified based on molecular size, sequence motifs, mechanistic features and structural comparisons (Kavanagh et al. 2008; Persson et al. 2009). Many studies have been addressed to characterize alcohol dehydrogenases (ADHs) from thermophiles and hyperthermophiles, mainly to understand their evolution and structure/function/stability relationship (Radianingtyas and Wright 2003) and develop their biotechnological potential in the synthesis of chiral alcohols (Jones and Beck 1976; Keinan et al. 1986; Hummel 1999; Kroutil et al. 2004; Müller et al. 2005; Goldberg et al. 2007). Recently, ADHs displaying distinct substrate specificity, good efficiency and high enantioselectivity have been described, such as the NADP-dependent (R)-specific ADH from Lactobacillus brevis (LB-RADH) (Schlieben et al. 2005), the NAD-dependent ADH from Leifsonia sp. strain S749 (LSADH) (Inoue et al. 2006), and the NADP-dependent carbonyl reductase from Candida parapsilosis (Nie et al. 2007). These enzymes originate from mesophilic microorganisms and belong to the short-chain dehydrogenases/reductases (SDRs) superfamily (Kavanagh et al. 2008) which is characterized by ∼250 residue subunits, a Gly-motif in the coenzyme-binding regions, and a catalytic tetrad formed by the highly conserved residues Asn, Ser, Tyr and Lys (Filling et al. 2002; Schlieben et al. 2005; Persson et al. 2009). Representative examples of ADHs from thermophilic microorganisms are medium-chain enzymes, including Thermoanaerobacter brockii ADH (Korkhin et al. 1998), the ADH from Bacillus stearothermophilus strain LLD-R (BsADH) (Ceccarelli et al. 2004) and two archaeal enzymes, the ADH from Aeropyrum pernix (Guy et al. 2003) and Sulfolobus solfataricus (Giordano et al. 2005; Friest et al. 2010). However, two archaeal short-chain ADHs have been identified in Pyrococcus furiosus, an NADP(H)-dependent SDR (van der Oost et al. 2001; Machielsen et al. 2008), and an NAD(H)-preferring ADH, that belongs to the aldo-keto reductase superfamily (Machielsen et al. 2006; Zhu et al. 2006, 2009). Furthermore, two short-chain NAD(H)-dependent ADHs, TtADH and SaADH, identified in Thermus thermophilus HB27 and Sulfolobus acidocaldarius, respectively, have been recently purified and characterized in our laboratory (Pennacchio et al. 2008, 2010a, 2010b, 2011).

With the aim to find novel dehydrogenase/reductases that are both stable and NAD+ dependent, we focused our attention on the genomes of thermophilic organisms containing genes encoding putative ADHs belonging to the SDR superfamily and applied the criteria used for TtADH and SaADH to select enzymes with a high probability of being NAD+ dependent (Pennacchio et al. 2008). An open reading frame coding for a protein belonging to the SDR superfamily with relatively high sequence identity to that of most representative ADHs (LB-RADH, LSADH, TtADH and SaADH) was found in the genome of S. acidocaldarius, an aerobic thermoacidophilic crenarcheon which grows optimally at 80 °C and pH 2 (Chen et al. 2005). The amino acid sequence revealed the presence of a glutamic acid residue at position 39. Noteworthy, an aspartate residue at the homologous position plays a critical role in determining the preference of SDRs for NAD(H) as shown for the LB-RADH mutant G37D (Schlieben et al. 2005) and by the evidence that the SDRs LSADH (Inoue et al. 2006), TtADH and SaADH (Pennacchio et al. 2008, 2010a, b) are strictly NAD(H)-dependent and have an aspartate residue at the same position within the sequence (Fig. 1). Due to Glu and Asp chemical similarity, the putative dehydrogenase/reductase identified in S. acidocaldarius (SaADH2) was expected to display a preference for NAD(H) rather than NADP(H). This preference as well as an intrinsic thermostability are the features that make an oxidoreductase more attractive from an application perspective (Kroutil et al. 2004; Zhu et al. 2006; Huisman et al. 2010).

Fig. 1
figure 1

Multiple-sequence alignment of the S. acidocaldarius ADH (SaADH2) and ADHs belonging to the SDR family, including S. acidocaldarius ADH (SaADH; NCBI accession no. YP_256716.1), L. brevis ADH (LB-RADH; PDB code 1ZK4), Leifsonia sp. strain S749 ADH (LSADH; NCBI accession no. BAD99642) and T. thermophilus ADH (TtADH; PDB code 2D1Y). The sequences were aligned using the ClustalW2 program. Grey shading indicates residues highly conserved in the SDR family. The four members of the catalytic tetrad are indicated by a black background. The following positions are indicated by bold type: the glycine-rich consensus sequence and the sequence motif Dhx[cp] that (in all SDRs) have a structural role in coenzyme binding (Kallberg et al. 2002). The star indicates the major determinant of the coenzyme specificity. The LB-RADH G37D mutant shows preference for NAD+ over NADP+ (Schlieben et al. 2005)

Full size image

This report describes cloning, heterologous expression and structural characterization of the S. acidocaldarius saadh2 gene, which encodes the SDR SaADH2. The purified enzyme was characterized in terms of substrate specificity, kinetics and stability as well as enantioselectivity. SaADH2 was found to be highly efficient and enantioselective in reducing the diaryl diketone benzil to (R)-benzoin.

Materials and methods

Chemicals

NAD(P)+ and NAD(P)H were obtained from AppliChem (Darmstadt, Germany). Alcohols, aldehydes, ketones and keto esters were obtained from Sigma-Aldrich. 1-Butyl-3-methylimidazolium tetrafluoroborate (BMIMBF4) was a kind gift from Professor S. Cacchi. Recombinant BsADH was prepared as previously described (Fiorentino et al. 1998). Other chemicals were A grade substances from Applichem. Solutions of NADH and NAD+ were prepared as previously reported (Raia et al. 2001). All solutions were made up with MilliQ water.

Amplification and cloning of the saadh2 gene

Chromosomal DNA was extracted by caesium chloride purification as described by Sambrook et al. (1989). Ethidium bromide and caesium chloride were removed by repeated extraction with isoamyl alcohol and extensive dialysis against 10 mM Tris–HCl pH 8.0 and 1 mM EDTA, respectively. DNA concentration was determined spectrophotometrically at 260 nm, and the molecular mass checked by electrophoresis on 0.8 % agarose gel in 90 mM Tris–borate pH 8.0 and 20 mM EDTA, using DNA molecular size markers. The saadh2 gene was amplified by polymerase chain reaction (PCR) using oligonucleotide primers based on the saadh2 sequence of S. acidocaldarius strain DSM 639 (GenBank accession no. YP_255871.1). The following oligonucleotides were used: (5′-TGCATAGTAGCATATGTCATATCAGAGTTTG-3′) as the forward primer (the Nde I restriction site is underlined in the sequence) and the oligonucleotide (5′-AGGAATTCACATTACAGTACAGTTAAACCACC-3′) as the reverse primer. This latter oligonucleotide introduces a translational stop following the last codon of the ADH gene, followed by an EcoRI restriction site, which is underlined in the sequence shown. The PCR product was digested with the appropriate restriction enzymes and cloned into the expression vector pET29a (Novagen, Madison, Wisconsin, USA) to create the recombinant pET29a–saADH plasmid. The insert was sequenced in order to verify that mutations had not been introduced during PCR.

Expression and purification of recombinant SaADH2

Recombinant protein was expressed in Escherichia coli BL21(DE3) cells (Novagen) transformed with the corresponding expression vector. Cultures were grown at 37 °C in 2 L of LB medium containing 30 μg ml−1 kanamycin. When the A600 of the culture reached 1.4, protein expression was induced by addition of isopropyl β-d-1-thiogalactopyranoside to a concentration of 1.0 mM. The bacterial culture was incubated at 37 °C for a further 24 h. Cells were harvested by centrifugation, and the pellet stored at −20 °C until use. The cells obtained from 2 L of culture were suspended in 20 mM Tris–HCl buffer (pH 7.5) containing 0.1 mM phenylmethylsulfonyl fluoride (PMSF) and were lysed using a French pressure cell (Aminco Co., Silver Spring, MD) at 2,000 psi (1 psi = 6.9 kPa). The lysate was centrifuged, and the supernatant incubated in the presence of DNase I (50 μg/ml of solution) and 5 mM MgCl2 for 30 min at 37 °C, followed by protamine sulphate (1 mg/ml of solution) at 4 °C for 30 min. The nucleic acid fragments were removed by centrifugation, and the supernatant incubated at 70 °C for 15 min. The host protein precipitate was removed by centrifugation. The supernatant was dialysed overnight at 4 °C against 20 mM Tris–HCl, pH 8.4 (buffer A) containing 1 mM PMSF. The dialysed solution was applied to a DEAE-Sepharose Fast Flow (1.6 × 12 cm) column equilibrated in buffer A at 120 ml h−1. The active pool did not bind to the column matrix. The flowthrough fractions were combined and dialysed against buffer A, concentrated 5-fold with a 30,000 MWCO centrifugal filter device (Millipore), and applied to a Sephadex G-75 (1.6 × 30 cm) column equilibrated in buffer A containing 0.15 M NaCl. The active pool was dialysed against buffer A and concentrated to obtain 2.0 mg protein·ml−1 as described previously. SaADH2 was stored at −20 °C in buffer A containing 50 % glycerol, without loss of activity following several months of storage. SDS-PAGE was carried out according to the Laemmli method (1970). The subunit molecular mass was determined by electrospray ionization mass spectrometry (ESI-MS) with a QSTAR Elite instrument (Applied Biosystems, USA). The protein concentration was determined with a Bio-Rad protein assay kit using BSA as a standard.

Sucrose density gradient centrifugation

One hundred microlitres of a 1 mg/ml purified SaADH2 solution was layered on top of a 10-ml 5–20 % preformed sucrose gradient in 50 mM Tris–HCl, pH 8.4. The standards were also prepared by layering 100 μl of 1 mg/ml chymotripsinogen (25 kDa), BSA (67 kDa), T. thermophilus ADH (a 108-kDa tetramer) and S. solfataricus ADH (a 150-kDa tetramer) on top of another identical sucrose gradient. By using a Beckman SW 41 Ti rotor, the tubes containing the samples were spun at 37,000 rpm for 18 h at 4 °C in a Beckman LE-80 ultracentrifuge. Immediately after centrifugation, gradients were collected in 250-μl fractions and assayed for absorption at 280 nm to determine the position of the marker protein. The mass of SaADH2 was estimated from the position of the peak relative to those of the markers.

Enzyme assay

SaADH2 activity was assayed spectrophotometrically at 65 °C by measuring the change in absorbance of NADH at 340 nm using a Cary 1E spectrophotometer equipped with a Peltier effect-controlled temperature cuvette holder. The standard assay for the reduction reaction was performed by adding 5−25 μg of the enzyme to 1 ml of preheated assay mixture containing 20 mM ethyl 3-methyl-2-oxobutyrate (EMO) and 0.2 mM NADH in 37.5 mM sodium phosphate, pH 5.0. The standard assay for the oxidation reaction was performed using a mixture containing 18 mM cycloheptanol and 3 mM NAD+ in 50 mM glycine/NaOH, pH 10.5. Screening of the substrates was performed using 1 ml of assay mixture containing either 10 mM alcohol and 3 mM NAD+ in 50 mM glycine/NaOH, pH 10.5, or 10 mM carbonyl compound and 0.1 mM NADH in 37.5 mM sodium phosphate, pH 5.0. One unit of SaADH2 represented 1 μmol of NADH produced or utilized per minute at 65 °C, on the basis of an absorption coefficient of 6.22 mM−1 cm−1 for NADH at 340 nm.

Effect of pH on activity

The optimum pH value for the reduction and oxidation reactions was determined at 65 °C under the conditions used for EMO and cycloheptanol, respectively, except that different buffer were used. The concentration of each buffer solution was diluted to get the similar conductivity (∼1.1 mS). The pH was controlled in each assay mixture at 65 °C.

Kinetics

The SaADH2 kinetic parameters were calculated from measurements determined in duplicate or triplicate and by analysing the kinetic results using the program GraFit (Leatherbarrow 2004). The turnover value (k cat) for SaADH2 was calculated on the basis of a molecular mass of 27 kDa, assuming that the four subunits are catalytically active.

Thermophilicity and thermal stability

SaADH2 was assayed in a temperature range of 25–90 °C using standard assay conditions and 22 μg protein·ml–1 of assay mixture. The stability at various temperatures was studied by incubating 0.2 mg ml–1 protein samples in 50 mM Tris–HCl, pH 9.0, at temperatures between 25 and 95 °C for 30 min. Each sample was then centrifuged at 5 °C, and the residual activity assayed as described above.

The effect of chelating agents on enzyme stability was studied by measuring the activities before and after exhaustive dialysis of the enzyme against buffer A, containing 1 mM EDTA, and then against buffer A alone. An aliquot of the dialysed enzyme was then incubated at 70 °C in the absence and presence of 1 mM EDTA, and the activity assayed at different times.

Effects of compounds on enzyme activity

The effects of salts, metal ions and chelating agents on SaADH2 activity were investigated by assaying the enzyme in the presence of an appropriate amount of each compound in the standard assay mixture used for the oxidation reaction.

The effects of organic solvents were investigated by measuring the activity in enzyme samples (0.2 mg ml–1 in 100 mM sodium phosphate, pH 7.0) immediately after the addition of organic solvents at different concentrations and after incubation for 6 and 24 h at 50 °C. The percentage activity for each sample was calculated by comparison with the value measured prior to incubation. The volume of the solution in a tightly capped test tube did not change during incubation.

Enantioselectivity

The enantioselectivity of SaADH2 was determined by examining the reduction of acetophenone, the bicyclic ketones benzyl and 2,2′-dichlorobenzil using an NADH regeneration system consisting of BsADH and a substrate alcohol (see below). The reaction mixture contained 2 mM NAD+, 5 mM carbonyl compound, 4 to 18 % v/v alcohol, 250 μg SaADH2 and 50 μg of BsADH in 1 ml of 100 mM sodium phosphate, pH 7.0. The reactions were carried out at 50 °C for 24 h in a temperature-controlled water bath. Upon termination of the reaction, the reaction mixture was extracted twice with ethyl acetate. The percentage of conversion and enantiomeric purity of the product were determined on the basis of the peak areas of ketone substrates and alcohol products by HPLC, on a Chiralcel OD-H column (Daicel Chemical Industries, Ltd., Osaka, Japan). The absolute configuration of product alcohols was identified by comparing the chiral HPLC data with the standard samples. Products were analyzed with isocratic elution, under the following conditions: hexane/2-propanol (9:1) (mobile phase), flow rate of 1 ml min−1, detection at 210 nm. Retention times were the following: 5.6, 11.9 and 16.7 for benzil, (S)- and (R)-benzoin, respectively; 6.98, 12.73 and 15.13, for 2,2′-dichlorobenzil, (S)- and (R)-1,2-bis(2-chlorophenyl)-2-hydroxyethanone, respectively and 5.27, 6.2 and 7.5 for acetophenone, (R)- and (S)-1-phenylethanol, respectively. The absolute stereochemistry of 1,2-bis(2-chlorophenyl)-2-hydroxyethanone enantiomers was assigned by analogy to the values of (S)- and (R)-benzoin.

Structural characterization

The purified protein was concentrated to 8–10 mg ml−1 in 20 mM Tris/HCl buffer at pH 8.4. The protein solution was added with NADH to obtain a 8.5-mg ml−1 protein solution containing 2.2 mM NADH in the above buffer. Crystallization was performed by hanging-drop vapour diffusion method at 296 K. The best crystals were grown by mixing 1 μl of protein solution with the same volume of a solution containing 2.0 M ammonium sulphate in 0.1 M HEPES buffer (pH 7.5). The crystals belong to P21 space group with unit cell parameters a = 83.80 Å, b = 60.48 Å, c = 107.1 Å and β = 99.8°. Crystals were transferred to a stabilizing solution containing 20 % glycerol as cryoprotectant and X-ray data were then collected in-house at 100 K using a Rigaku Micromax 007 HF generator producing Cu Kα radiation and equipped with a Saturn944 CCD detector. Intensity data up to 1.75 Å resolution were processed and scaled using program HKL2000 (HKL Research). Data collection statistics are shown in Table 1.

Table 1 Data collection and refinement statistics
Full size table

The crystal structure was solved by MR methods by using as a search model the coordinates of a putative glucose/ribitol dehydrogenase from Clostridium thermocellum (PDB code 2HQ1), which showed the best sequence alignment with SaADH2 (39 % sequence identity). Model building was first performed with ARP/WARP package using the warpNtrace automated procedure. Later model building and refinement was performed by the program CNS version 1.1. The final structure shows R and R-free factors of 16.8 and 19.1 %, respectively. Refinement statistics are summarized in Table 1. Fold similarities searches were performed with the DALI server (http://ekhidna.biocenter.helsinki.fi/dali_server).

Results

Expression and protein purification

Analysis of the S. acidocaldarius genome (Chen et al. 2005) for genes encoding short-chain ADHs resulted in identification of a putative oxidoreductase gene. The sequence of the 27,024-Da protein, named SaADH2, showed the highest level of identity to four typical SDRs, LB-RADH (29 % identity), LSADH (34 %), TtADH (35 %) and SaADH (30 %; Fig. 1). The saadh2 gene was successfully expressed in E. coli cells, yielding an active enzyme accounting for about 9 % of the total protein content of the cell extract (Supplementary material Table S1). Host protein precipitation at 70 °C was found to be the most effective purification step. An overall purification of 5.4-fold was achieved from crude-cell-free extracts with an overall yield of 48 %. SDS-PAGE of the purified protein showed a single band corresponding to a molecular mass of ∼29 kDa (data not shown).

The quaternary structure of the enzyme was investigated by sucrose density gradient centrifugation. SaADH2 sedimented as one peak nearly overlapping that of TtADH (Supplementary Material Fig. S1). The plot of molecular mass of the markers vs fraction number (inset Fig. S1) allowed to determine an apparent molecular mass of 109 ± 10 kDa for SaADH2. The molecular mass of the subunit determined by ESI-MS analysis proved to be 27,024.0 Da (average mass), in agreement with the theoretical value of the sequence.

Optimal pH

The pH dependence of SaADH2 in the reduction and oxidation reaction was analysed (Fig. S2). The SaADH2 activity was found to be closely dependent on pH in the reduction reaction, displaying a peak of maximum activity at around pH 5.0. The oxidation reaction showed a less marked dependence on pH, displaying a peak with a maximum at around pH 10.0.

Thermophilicity and thermal stability

The effect of temperature on SaADH2 activity is shown in Fig. 2. The reaction rate increases up to 78 °C and then decreases rapidly due to thermal inactivation. This optimal temperature value is similar to that of TtADH (73 °C) and SaADH (75 °C) (Pennacchio et al. 2008, 2010a, b), and lower than that of P. furiosus aldo-keto reductase (100 °C) (Machielsen et al. 2006). The thermal stability of SaADH2 was determined by measuring the residual enzymatic activity after 30 min of incubation over a temperature range from 25 to 95 °C (Fig. 2). SaADH2 was shown to be quite stable up to a temperature of ∼75 °C, above which its activity decreased abruptly, resulting in a T1/2 value (the temperature of 50 % inactivation) of ∼88 °C.

Fig. 2
figure 2

Effect of temperature on activity and stability of SaADH2 monitored by dehydrogenase activity. The assays at the increasing temperature values (triangles) were carried out as described in Materials and methods, using cycloheptanol as the substrate. The thermal stability (circles) was studied by incubating 0.2 mg ml−1 protein samples in 50 mM Tris–HCl, pH 9.0 for 30 min at the indicated temperatures. Activity measurements were carried out under the conditions of the standard assay using cycloheptanol as the substrate. The assay temperature was 65 °C. The percentage of residual activity was obtained by the ratio to the activity without heating

Full size image

Coenzyme and substrate specificity

The enzyme showed no activity with NADP(H) and full activity with NAD(H).

The specificity of SaADH2 for various alcohols, aldehydes and ketones was examined (Table 2). The enzyme showed a poor activity on a discrete number of aliphatic linear and branched alcohols, such as 2-propyn-1-ol, 4-methyl-1-pentanol, the S enantiomers of 2-butanol and 2-pentanol as well as on aliphatic cyclic and bicyclic alcohols, except for isoborneol and cycloheptanol which rank first and second, respectively, among the tested alcohols. Benzyl alcohol and substituted benzyl alcohols were found to be poor substrates. Among the aromatic secondary alcohols tested SaADH2 showed a low activity on (S)-1-phenylethanol and no activity with the R enantiomer, whereas displayed similar poor activities towards the (S)- and (R)- forms of α-(trifluoromethyl)benzyl alcohol, 1-(2-naphthyl)ethanol and methyl and ethyl mandelates. Moreover, the enzyme showed poor activity on (±)-1-phenyl-1-propanol and its ortho-chloro derivative, and para-halogenated 1-phenylethanols. However, SaADH2 showed a relatively high activity with 1-indanol and α-tetralol, but a poor activity on the β-hydroxy ester ethyl (R)-4-chloro-3-hydroxybutyrate.

Table 2 Substrate specificity of SaADH2 in the oxidation and reduction reactionsa
Full size table

The enzyme was not active on aliphatic and aromatic aldehydes, and on aliphatic linear, and branched ketones (data not shown). However, it was active on aliphatic cyclic and bicyclic ketones such as cyclohexanone, methyl-substituted cyclohexanones and decalone (Table 2). Two aryl diketones, 1-phenyl-1,2-propanedione and benzil were good substrates of SaADH2 which also showed a relatively high reduction rate with 2,2-dichloro- and 2,2,2-trifluoroacetophenone and penta-substituted fluoro-acetophenone (Table 2). The electronic factor accounts for the relatively high activity measured with the halogenated acetophenones, as compared to the apparent zero activity observed with acetophenone. The electron withdrawing character of fluorine (or chlorine) favours hydride transfer, inductively decreasing electron density at the acceptor carbon C1. On the other hand, the corresponding ketones of cycloheptanol and isoborneol, cycloheptanone and camphor did not show any apparent activity due to the deactivating effect exerted by the electron donating alkyl groups on the acceptor carbon C1. SaADH2 was also active on aliphatic and aryl α-keto esters but not on β-ketoesters (Table 2).

Kinetic studies

The kinetic parameters of SaADH2 determined for the most active substrates are shown in Table 3. Based on the specificity constant (k cat/K m), isoborneol is the best substrate in the oxidation reaction; compared to the alicyclic cycloheptanol and aromatic bicyclic alcohols, it displays a higher affinity to the active site probably due to its alicyclic bridged structure. However, the enzyme shows a 6-fold greater preference for (S)- than (R)-1-indanol and a 23-fold greater preference for (S)- than (R)-α-tetralol. In the reduction reaction, 2,2-dichloroacetophenone was preferred 22-fold more than 2,2,2-trifluoroacetophenone and 2′,3′,4′,5′,6′-pentafluoroacetophenone, and 1.6- and 6-fold more than the two diketones, benzil and 1-phenyl-1,2-propanedione, respectively, due to its higher affinity. Ethyl 3-methyl-2-oxobutyrate shows the highest turnover among the carbonyl compound tested, although it binds to the catalytic site with relatively low affinity. Moreover, the specificity constant value is 6-fold higher for NADH than NAD+.

Table 3 Steady-state kinetic constants of SaADH2
Full size table

Effects of various compounds

The effects of salts, ions and reagents on SaADH2 activity were studied by adding each compound to the standard assay mixture. The enzyme activity in the presence of 1 mM of the Li+, Na+, K+, Ca++, Mg++ and Mn++ chlorides was 115, 113, 102, 101, 101 and 110 %, respectively, and in the presence of 1 mM of the sulphate of heavy metal ions such as Fe++, Co++, Ni++, Cu++ and Zn++ was 93, 108, 104, 102 and 107 %, correspondingly, when compared to the enzyme activity measured in the absence of additional metal ions. The presence of 5 % ionic liquid (BMIMBF4) inactivated by 50 % the enzyme presumably due to a competition of the BF −4 ion with the coenzyme phosphate moiety for the anion-binding site of the enzyme.

The addition of 4 mM iodoacetate and 1 mM Hg++ had no significant influence on the enzyme activity, which resulted in 101 and 113 % of the control activity, respectively, suggesting that the only Cys residue per monomer, C90, has no functional role. Even the metal-chelating agents did not affect enzyme activity. The enzyme activity in the presence of either 1 mM o-phenanthroline, or 10 mM EDTA was 98 and 87 %, respectively, suggesting that either the protein does not require metals for its activity or the chelating molecule was not able to remove the metal under the assay conditions. Furthermore, the enzyme showed no loss in activity following exhaustive dialysis against EDTA. The EDTA-dialysed enzyme turned out to be quite stable at 70 °C for 5 h, both in the absence and the presence of EDTA (data not shown).

Stability in organic solvents

The effects of common organic solvents, such as acetonitrile, DMSO, 1,4-dioxane and ethyl acetate on SaADH2 were investigated at 50 °C, at two different time points and 18 % concentration (Fig.  3 ). SaADH2 activated after 6 and 24 h incubation in aqueous buffer (120 and 105 % the initial values, respectively), and inactivated by 15 and 30 % in the presence of 18 % acetonitrile following incubation for 6 and 24 h, respectively. A slightly reduced activity (∼95 %) remained after 6 and 24 h incubation in aqueous solution containing 18 % ethyl acetate. Interestingly, significant increases in enzyme activity occurred after 6 h incubation in the presence of 18 % DMSO (>140%) and 1,4-dioxane (135 %). After 24 h incubation in the presence of these two solvents, the enzyme activity remained nearly unchanged with respect to the initial value.

Fig. 3
figure 3

Effects of various solvents on SaADH2. Samples of enzyme (0.20 mg/ml) were incubated at 50 °C in the absence and presence of the organic solvents at 18 % concentrations, and the assays were performed after 6 h (white bars) and 24 h (grey bars). The activity assays were performed at 50 °C as described in Materials and methods using cycloheptanol as the substrate. The data obtained in the absence and presence of organic solvents are expressed as percentage of activity relative to the value determined prior to incubation

Full size image

Enantioselectivity

The enantioselectivity of SaADH2 was tested on benzil using an NADH recycling system consisting of thermophilic NAD(H)-dependent BsADH (Fig. 4). The latter enzyme is mainly active on aliphatic and aromatic primary and secondary alcohols and aldehydes (Guagliardi et al. 1996), but not on aliphatic and aromatic ketones, nor on the carbonyl substrates of SaADH2 and corresponding alcohols (data not shown). Since 2-propanol is not a substrate of SaADH2, it may be a suitable substrate for BsADH in NADH recycling, as well as being used as a co-solvent. Bioconversions were carried out for 24 h using 5 mM benzil as substrate and 2-propanol at three different concentrations. Sodium phosphate buffer, pH 7.0 and 50 °C was chosen as a compromise between cofactor stability and catalytic activity of the two ADHs at suboptimal pH (Pennacchio et al. 2008, 2011). Chiral HPLC analysis of the extracts obtained from the bioconversions showed that benzil was reduced by the archaeal enzyme to the (R)-enantiomer of benzoin with a level of conversion of 35, 100 and 99 % and an enantiomeric excess (ee) of 85, 82 and 80 % using 4, 14 and 18 % (v/v) 2-propanol as ancillary substrate, respectively.

Fig. 4
figure 4

Coenzyme recycling in the production of chiral diaryl alcohol with SaADH2 utilizing B. stearothermophilus ADH (BsADH) and 1-butanol

Full size image

In addition to 2-propanol (k cat/K m = 9 s−1 mM−1), BsADH oxidizes other alcohols with even greater efficiency such as ethanol (k cat/K m = 64 s−1 mM−1), 1-propanol (286 s−1 mM−1), 1-butanol (437 s−1 mM−1), 1-pentanol (64 s−1 mM−1) and 1-hexanol (64 s−1 mM−1; Raia, unpublished data). We therefore tested these alcohols as alternative hydride source for NADH recycling and also to improve the solubility of the substrate in the aqueous phase. Moreover, these alcohols and the respective aldehydes are not substrates of SaADH2. Thus, only the cofactor is the co-substrate of SaADH2 and BsADH.

Figure 5 summarizes the results of the bioconversions carried out using different alcohol substrates at 18 % concentrations. As for the case of 2-propanol, benzil was reduced to the (R)-alcohol with excellent conversion (>99 %) but modest enantioselectivity using ethanol and 1-propanol. However, in the presence of 1-butanol, 1-pentanol and 1-hexanol excellent optical purity (98–99 % ee) and levels of conversion decreasing from 98 to 93 % and 66 %, respectively, were obtained. This suggests that ethanol and linear/branched propanol acted as good substrates and solvents for the bioconversion system, but negatively affected the prochiral selectivity of SaADH2 that was instead enhanced by longer chain-linear alcohols (four to six carbons). The decrease in conversion rate with 1-pentanol and 1-hexanol could be due to different effects: (1) substrate excess inhibition, (2) BsADH deactivation and (3) reduced substrate solubility.

Fig. 5
figure 5

Reduction of benzil catalyzed by SaADH2. Biotransformations were carried out at 50 °C with different alcohols at 18 % (v/v) concentrations. The reactions were stopped after 24 h by addition of ethyl acetate. The dried extracts were analysed by chiral HPLC to determine the relative conversion (RC, grey bars) and the enantiomeric excess (ee, dark grey bars). Error limit, 2 % of the stated values. The alcohols are plotted according to increasing log P (P = partition coefficient) values; from left to right: log P = −0.24, 0.07, 0.28, 0.8, 1.3 and 1.8. Log P values were from Laane et al. (1987)

Full size image

The SaADH2 enantioselectivity was further examined with other aromatic ketones using the system BsADH/1-butanol. Table 4 shows that SaADH2 preferably reduced benzil to (R)-benzoin with an ee of 98 % and 98 % conversion after 24 h of reaction, and that the selectivity was reduced when 2-propanol instead of 1-butanol was used. However, 2,2′-dichlorobenzil was reduced to (R)-1,2-bis(2-chlorophenyl)-2-hydroxyethanone with lower conversion (33 %) and ee of 91 %, despite the apparent inactivity under different assay conditions (Table 2). Presumably, the presence of chlorine, a bulky atom with electron-withdrawing properties, disturbs the fitting of the diketone molecule in the substrate-binding pocket as well as the interactions that stabilize the transition state. Acetophenone was reduced to (S)-1-phenylethanol with a 4 % conversion and an ee of 92 %, while no conversion was observed for camphor, the ketone corresponding to the highly preferred alcohol isoborneol (Table 3). Kinetic tests showed that neither racemic benzoin nor (R)-benzoin was oxidized in the presence of NAD+ as well as both were not reduced in the presence of NADH. Accordingly, the HPLC analysis indicated that no further reduction of benzoin occurred since no trace of hydrobenzoin could be found within the detection limits (data not shown).

Table 4 Asymmetric reduction of carbonyl compounds by SaADH2
Full size table

Structural characterization

The determined crystal structure reveals a tetrameric organization similar to other members of the short-chain dehydrogenase/reductase (SDR) family. Although SaADH2 has low sequence similarities with other SDR proteins, it shares common structural features with them (Kallberg et al. 2002; Kavanagh et al 2008). The overall structure is typical of SDRs that is an α/β fold with characteristic Rossmann-fold motifs for the binding of the dinucleotide cofactor at the N-terminal region of the chain. Indeed, the structure confirmed that SaADH2 is a NAD-dependent enzyme with a cofactor-binding domain composed of a seven-stranded parallel sheet flanked by helices on each side.

Overall tertiary and quaternary structure

The structural comparison of SaADH2 with other SDRs demonstrated that SaADH2 exhibits a high degree of overall structural homology. Indeed, when aligning the monomeric structures by the DALI server the first 100 best superimpositions showed RMS deviations on Cα atoms ranging from 1.5 to 2.5 Å (with a minimum of 222/255 aligned residues). Some structural neighbours are superimposed onto a single chain of SaADH2 in Fig. S3. Major conformational differences are found on the molecular surface and at the terminal region of the polypeptide chain.

Analysis of the crystal packing (using PISA server:http://www.ebi.ac.uk/pdbe/prot_int/pistart.html), indicates that the tetramer is the biologically active form. The tetramer shows 222 point group symmetry and approximate dimensions of 80 × 70× 55 Å (Fig. S4). The largest dimeric interface is the one involving AB or CD assemblies (∼1,630 Å2). The central elements in this interface are the long α6 helices (residues 156–179, Fig. S5) facing each other from adjacent subunits. Other segments involved in this interface are α4 and α5 helices (Fig. S5). The second large interface involving AC or BD assemblies (∼1,250 Å2) mainly comprises the C-terminal regions (residues 216–255) of the two subunits. The very terminal segment (residue 253–255) also contributes to the smallest interface involving AD or BC assemblies (∼445 Å2); indeed, this segment forms inter-subunit interactions with residues belonging to 150–153 segment. It is worth noting that residues from the 209–218 region of the structure also mediate inter-subunit interactions in this interface.

Cofactor-binding site

A highly defined electron density was observed for the whole NAD(H) molecule thus allowing unambiguous identification of the positions of the cofactor in all the four molecules of the asymmetric unit (Fig. S6). The cofactor-binding site is located at the C-terminal end of the central parallel β-sheet.

The NAD(H) cofactor specificity is explained by the presence of Glu39 whose side chain oxygen atoms form hydrogen bonds with two oxygen atoms of the adenosine ribose of NAD(H).

The adenine ring of NAD binds in a hydrophobic pocket on the enzyme surface formed by the side chains of Val116, Leu40, Val66, Ala93 and Gly15. Hydrogen bonds are formed by N6 of adenine and the Asp65 side chain and by N1 and the peptide nitrogen of Val66. The segment 14-TGAGSGIG-21 corresponds to the glycine-rich sequence (TGxxx[AG]xG) present in NAD-binding site of “classical” SDR (Kavanagh et al 2008). This region interacts with the adenine ribose and the diphosphate moiety of NAD(H). The phosphate groups also interact with Thr193-Ala194. The 2′- and 3′-hydroxyls of the second ribose, attached to the nicotinamide ring, form a bifurcated hydrogen bond with Lys162. The 3′-hydroxyl is also hydrogen bonded to the active site Tyr158. The syn face of the nicotinamide ring displays vdW contacts with the hydrophobic side-chains of Ile20, Pro188 and Val191. The amide part of the nicotinamide portion is anchored to the protein by hydrogen bonds to Thr193 and Val191.

Substrate-binding site

The SaADH2 contains the conserved catalytic residues, Ser145, Tyr158 and Lys162, located in the active-site cleft with a spatial arrangement that is common to that found in other SDRs. SaADH2 also shows the conserved Asn residue (Asn117) which has been suggested as a key residue in a catalytic mechanism based on a tetrad instead of a triad (Filling et al. 2002). Its side chain atoms form hydrogen bonds with main chain atoms of Ile95.

In many SDRs the substrate-binding site is a deep cleft with a floor created by the NADH molecule and the right-side and left-side walls formed by the substrate-binding region (residues 193-221 in SaADH2 numbering) and the loop between β5 and β6 strands (Fig. 6).

Fig. 6
figure 6

Overall tertiary structure of one of the subunits of SaADH2. Bound NADH is shown in stick representation (magenta). The substrate-binding loop, the stretches 152–155 and 96–101 are highlighted in blue, yellow and red, respectively. The arrows point toward the substrate-binding cleft; the directions of approach are shown in magenta and black (see the text for details)

Full size image

In three out of four subunits of the asymmetric unit there is a glycerol molecule occupying the position that is approximately suited for the binding of substrate. Indeed, the glycerol O2 and O3 oxygen atoms are hydrogen bonded to the oxygen of the NAD carboxamide and to the oxygen of Tyr158 side chain, respectively. Glycerol was used as cryoprotectant agent and therefore diffused through the water channels of the crystals up to the binding site. The lacking of glycerol binding in one of the subunits (chain B) of the tetramer is due to crystal packing which restricts the access to NADH in subunit B. Indeed, a nearby symmetric molecule (#B) inserts its 50–57 region into the substrate-binding region (residues 193–221) that covers the access to the NADH site. On the other hand, the other subunits of the tetramer do not show direct interactions with adjacent symmetric molecules through the substrate-binding loop.

It is worth noting that, despite the different crystal environments experienced by the four molecules of the asymmetric unit, the substrate-binding loops adopt similar conformations in the four subunits of the tetramer with pairwise RMSDs calculated on Cα atoms lying in the range 0.36–0.68 Å.

Furthermore, the structure of the loop (in blue in Fig. 6) represents the most significant difference between SaADH2 and the other SDR structures aligned with it. Indeed, as can be seen in Fig. S3, this loop (in cyan) is remarkably different from the other structures. In most of SDR structures the substrate-binding loop contains two segments of α-helix on opposite sides of the loop. In addition, part of the loop is often disordered thus revealing a considerable mobility. In SaADH2 the loop shows the following features: (1) it is well defined in the electron density; (2) it contains only one stretch of α-helix lacking the one usually present at the C-terminal end and (3) it partially obstructs the access to the substrate-binding cleft indicated with a magenta arrow in Fig. 6. Indeed, a detailed comparison with other SDR aligned structures reveals that three stretches (193–221, 96–101 and 152–155) of SaADH2 structure adopt a different conformation by getting closer to the NADH nicotinamide ring which becomes inaccessible to solvent/substrate from the top of the molecule (magenta arrow in Fig. 6). The cofactor ring is, however, still accessible to ligands through a side cavity (black arrow in Fig. 6) lined by residues Met96, Ser145, Ile146, Ala147, Phe153, Ala154, Tyr158, Gly189, Thr190, Ile195, Gly196, Leu197, Leu210 and Met214 (Fig. S7). The mouth of the cavity is composed of atoms from residues 190, 195–197, 210 and 214, as confirmed by CastP server (http://sts.bioengr.uic.edu/castp/calculation.php).

The benzil molecule is an interesting substrate of the enzyme that reduces it to R-benzoin in the presence of NADH with excellent conversion (>99 %). Several attempts were performed to crystallize a ternary complex of the enzyme bound to a benzil molecule, but they were unsuccessful. To envision how the enzyme binds the substrate and to probe the stereoselectivity of the hydride transfer we have manually modelled the benzil molecule in the active site.

The starting structure for the alpha-diketone modelling was the crystal structure present in the small molecule structure database (Allen 2002). This structure shows a conformation of the molecule with the two aromatic rings forming a dihedral angle of 114° around the diketo bond. A single benzil carbonyl group is bonded to two substituents: a smaller phenyl group and a larger carbophenyl group. In order to obtain a (R)-alcohol, the hydride has to be added to the Si face of the carbonyl ketone and this means that the smaller group is located opposite to the bulky carboxamide group of NADH.

A rigid docking of the benzil compound performed by the patchdock server identifies an orientation in the substrate site which is consistent with the stereoselectivity of the reduction reaction leading to (R)-benzoin. However, it has to be noted that in the best docked conformational ensemble of benzil, the carbonyl C atom, to be attacked by hydride, is still rather far from the hydrogen at the NADH C4 position (>5.5 Å). In addition, the oxygen atom of the carbonyl is not hydrogen bonded to the catalytic Tyr158 side chain (4.7 Å apart) as expected in case of a fully productive state of the enzyme–substrate complex.

Due to the steric hindrance of the benzil molecule an efficient accommodation in the substrate site would require a structural rearrangement of residues from the three segments (193–221, 96–101, and 152–155) which line the site and/or a change in the conformation of the diketone. A better packing of benzil in the site can be indeed obtained when the torsional angle around the carbon–carbon bond linking the two carbonyl moieties is reduced to values around 80–90°. Although approximate, this manually docked substrate position identifies a small pocket deep into the active site (Fig. 7) which is lined by Ser145, Leu146, Ala147, Tyr158, Phe153, Ala154 and Gly189. This pocket can accommodate the small phenyl group of benzil. Indeed, this phenyl ring would display mostly favourable contacts with Phe153, Ile146 and Ala147; the phenyl ring lies with its ring between two main-chain oxygen atoms from residues Phe153 and Glu189. On the other hand, the large carbophenyl group of benzil would point toward the channel opened to the solvent and lined by Met96, Thr190, Ile195, Gly196, Leu197, Leu210 and Met214. However, this bulky carbophenyl group would display steric clashes with the NADH and Thr190 when the intercarbonyl dihedral angle is about 80°. Some bad contacts can be relieved by a deviation of the phenyl group from the plane defined by the carbonyl moiety as well as by small rearrangements of the carboxyamide group of the nicotinamide ring (Fig. 7).

Fig. 7
figure 7

Surface representation of a single chain of SaADH2. The NADH molecule is highlighted in magenta, as both stick and surface representations. The black arrow indicates the channel by which the nicotinamide ring is accessible. The green box is a close-up view of the binding site with the benzil molecule (in orange) manually docked into it. The figure was prepared using the PyMOL software (www.pymol.org)

Full size image

Conformational rearrangements with respect to the crystal structure of benzil can be accessible at a reasonable energetic cost since both experimental and theoretical studies have indicated a significant flexibility of the molecular structure even depending on the environmental conditions (Pawelka et al. 2001; Lopes et al. 2004).

The presence of the deep pocket in the active site, well suited for an aromatic ring, also explains why, in the case of acetophenone and its derivatives, the alcohol obtained by the reduction reaction is the (S)-enantiomer.

Discussion

In search of novel dehydrogenase/reductases that are both stable and NAD+ dependent, a short-chain dehydrogenase (SDR) from S. acidocaldarius was identified and overexpressed. The enzyme was successfully purified and functionally/structurally characterized.

The enzyme is endowed with a high thermal stability showing a T1/2 value of ∼88 °C. It is rather stable in common organic solvents also exhibiting an increased activity in presence of DMSO and 1,4-dioxane. The activation of thermophilic enzymes by loosening of their rigid structure in the presence of protein perturbants is a well-documented phenomenon (Fontana et al. 1998; Liang et al. 2004). The two organic solvents discussed above (Fig. 3) may induce a conformational change in the protein molecule to a more relaxed and flexible state that is optimal for activity.

The analysis of the effects of ions reveals that SaADH2 does not require metals for its activity or structural stabilization; this is typical of non-metal SDR enzymes, although the well-known LB-RADH shows strong Mg++ dependency (Niefind et al. 2003).

Kinetic data show that SaADH2 is a strictly NAD(H)-dependent oxidoreductase with discrete substrate specificity. SaADH2 is more similar to the archaeal thermophilic TtADH and SaADH (Pennacchio et al. 2008; 2010a, b), than to the two representative SDR mesophilic ADHs, i.e. LB-RADH (Schlieben et al. 2005) and LSADH (Inoue et al. 2006), which are active on a variety of aliphatic as well as aromatic alcohols, ketones, diketones and keto esters. Indeed, the SaADH2 catalytic activity against various putative substrates was checked. Poor activity was observed on aliphatic linear and branched alcohols as well as on most of cyclic and bicyclic alcohols with the exception of cycloheptanol and isoborneol (Tables 2 and 3). Significant activity was found on alicyclic alcohols fused to an aromatic ring such as tetralol and indanol substrates. The kinetic data show that for most of chiral substrates SaADH2 is (S)-stereospecific and that the physiological direction of the catalytic reaction is reduction rather than oxidation (Table 3). The enzyme was active on aliphatic cyclic and bicyclic ketones such as cyclohexanone, methyl-substituted cyclohexanones and decalone (Table 2). It also showed a good reduction rate on halogenated acetophenone derivatives (Table 2) probably due to the electron withdrawing character of halogens which favours hydride transfer by decreasing electron density at the carbonyl carbon.

Interestingly, two aryl diketones, 1-phenyl-1,2-propanedione and benzil are good substrates of SaADH2. Although the natural substrates and function of SaADH2 remain unknown, the catalyzed reduction of the benzil α-diketone to (R)-benzoin with high yield and optical purity makes SaADH2 an interesting and biotechnologically relevant enzyme. Bioconversion studies have shown that SaADH2 can be conveniently coupled with a thermophilic bacillar ADH (BsADH) resulting in an efficient regeneration of NADH. Moreover, it is noteworthy that the SaADH2 enantioselectivity changes from moderate to excellent on going from ethanol to 1-hexanol (Fig. 5), i.e. that enantioselectivity increases as the hydrophobicity of the medium increases. The finely tuned solvent dependence of the prochiral selectivity shown by SaADH2 is remarkable and agrees with the general observation that enzyme selectivity (enantiomeric, prochiral and regiomeric) can change, or even reverse, from one solvent to another (Carrea and Riva 2000; Klibanov 2001). Overall, the solvent enantioselectivity study emphasizes the versatility and high efficiency of the BsADH/alcohol substrate system in recycling the reduced cofactor.

α-Hydroxy ketones are highly valuable building blocks for many applications for the fine chemistry sector as well as pharmaceuticals (Hoyos et al. 2010). It is noteworthy that optically pure (R)-benzoin is useful as urease inhibitor (Tanaka et al. 2004), and to prepare amino alcohols used as important chiral synthons (Aoyagi et al. 2000), or to synthesize different heterocycles (Wildemann et al. 2003). Many examples of stereoselective reduction of several benzils to the corresponding benzoins using whole microbial cells have been recently reviewed (Hoyos et al. 2010). The microorganisms Bacillus cereus, Pichia glucozyma, Aspergillus oryzae, Fusarium roseum, Rhizopus oryzae (at pH 4.5–5.0) and B. cereus Tm-r01 were reported to give (S)-benzoin while the microorganisms R. oryzae (at pH 6.8–8.5) and Xanthomonas oryzae were reported to transform benzil to (R)-benzoin with different yields and optical purity (Hoyos et al. 2010 and references therein). However, recombinant B. cereus NADPH-dependent benzil reductase belonging to SDR superfamily which reduces benzil to 97 % optically pure (S)-benzoin in vitro was characterized (Maruyama et al. 2002). To our knowledge, this study represents the first biochemical characterization of an NAD(H)-dependent archaeal short-chain ADH reducing the diketone compound benzil to (R)-benzoin with high yield and optical purity.

The functional characterization of SaADH2 was complemented by the determination of the crystal structure of SaADH2–NADH complex at 1.75 Å resolution.

The determined crystal structure reveals a tetrameric organization that corroborates the results obtained by sucrose density gradient centrifugation. The structural comparison of SaADH2 with other SDRs demonstrated that SaADH2 exhibits a high degree of overall structural homology despite the low sequence similarities. The largest differences with other SDR structures are on the surface and at the C-terminal region of the polypeptide chain which contributes to the shaping of the active site.

The reductase activity measurements showed that the protein is not active with NADP(H), having a definite preference for NAD(H) cofactor. This cofactor specificity is consistent with structural results. Indeed, the side chain oxygen atoms of Glu39 form hydrogen bonds with two oxygen atoms of the adenosine ribose of NAD(H), thus hampering the binding of NADP(H). Therefore, the Glu residue plays the same role of the corresponding Asp residue found in many SDR structures.

The active site of SaADH2 structure contains the conserved catalytic residues, Ser145, Tyr158 and Lys162, which adopt position and orientation similar to those found in other SDRs (Schlieben et al. 2005). Besides the catalytic triad, also Asn117 is structurally conserved; this residue has been suggested to play a role in maintaining the active-site architecture and the building up of a proton-relay system in SDR structures (Filling et al. 2002; Schlieben et al. 2005).

A comparison with other structures of the SDR superfamily reveals that the substrate-binding loop (residues 193–221) is the most different region of the subunit structure (Fig. S3). This is not unexpected since the conformation, the length and the composition of the substrate-binding loop are rather variable in SDR proteins thus conferring different substrate specificity. Besides this loop, also the segment 152–155 shows a different conformation by getting closer to the NADH nicotinamide ring (Fig. 6). Interestingly, this region contains a Phe residue (Phe153) which points deep into the active site and can play a relevant role for substrate specificity. The region 96–101 is also closer to the nicotinamide ring thus contributing to the reduced accessibility of the site where hydride transfer has to occur (Fig. 6). As a consequence, the manual modelling of the bulky benzil molecule in the active site, guided by the position of glycerol molecule bound and by preliminary results of rigid docking, cannot avoid the presence of unfavourable contacts. However, it can be suggested that minor changes in the three segments of the protein structure discussed above as well as in the conformation of the benzil α-diketone may well accommodate this substrate in the site (Fig. 7).

The elucidation of the ternary complex structure will ultimately provide details of the conformational changes in both the enzyme and the substrate necessary to obtain a fully productive state which guarantees high enantioselectivity. Nonetheless, the modelling explains the enantioselectivity of the benzil reduction. The (R)-benzoin is obtained since the hydride transfer occurs on the Si-face of the carbonyl ketone attacked. There is a deep pocket in the active site which can accommodate the small phenyl group of benzil (Fig. 7). Indeed, this phenyl ring would display mostly favourable contacts with Phe153, Ile146 and Ala147. On the other hand, the large carbophenyl group of benzil would point toward the channel opened to the solvent (Fig. 7) and lined with hydrophobic residues (Fig. S7).

Furthermore, the modelling of benzil provides clues on the different stereoselective course of the reduction reaction for other substrates such as acetophenone and its derivatives. Indeed, the deep pocket in the active site, well suited for an aromatic ring, can efficiently accommodate the phenyl group of the acetophenone whereas the methyl group can point toward the channel; this orientation of the substrate in the site leads to the (S)-1-phenylethanol product (Table 4).

In conclusion, a new short-chain dehydrogenase from S. acidocaldarius was identified and overexpressed. The purified enzyme was shown to possess remarkable thermal resistance as well as high enantioselectivity on the aryl diketone benzil molecule. It shows many advantages with regard to its preparative application: ease of purification, preservability of the biocatalyst and absolute preference for NAD(H). It is also amenable for coupling with a thermophilic bacillar ADH (BsADH) in an efficient and sustainable bioconversion process based on coenzyme recycling, where both enzymes have only the coenzyme as co-substrate. The remarkable resistance of SaADH2 to organic solvents proves the sturdiness of this biocatalyst and suggests exploratory investigations on conversions of poorly water-soluble prochiral substrates. In addition, the availability of the crystal structure of the enzyme provides the opportunity to rationalize the stereospecificity and stereoselectivity of the enzyme and to design proper mutations to broaden the scope of possible substrates.