(S)-Pramipexole and Its Enantiomer, Dexpramipexole: A New Chemoenzymatic Synthesis and Crystallographic Investigation of Key Enantiomeric Intermediates

A new chemoenzymatic method has been developed for the synthesis of (S)and (R)-N-(6-hydroxy-4,5,6,7-tetrahydrobenzo[d]thiazol-2-yl) acetamide, two key synthons for the preparation of (S)-pramipexole, an anti-Parkinson drug, and its enantiomer dexpramipexole, which is currently under investigation for the treatment of eosinophil-associated disorders. These two building blocks have been obtained in good yields and high enantiomeric excess (30% and >98% ee for the R-enantiomer, and 31% and >99% ee for the Sone) through a careful optimization of the reaction conditions, starting from the corresponding racemic mixture and using two consecutive irreversible transesterifications, catalyzed by Candida antarctica lipase type A. Single crystal X-ray analysis has been carried out to unambiguously define the stereochemistry of the two enantiomers, and to explore in depth their three-dimensional features.


Introduction
Pramipexole is a dopamine agonist approved for the treatment of Parkinson's disease and for restless legs syndrome; it is marketed as dihydrochloride salt under the brand name Mirapex [1][2][3].
Pramipexole or 2-amino-6-propylamino-4,5,6,7-tetrahydrobenzothiazole 1, has a stereogenic center on C6 (Figure 1), and only the isomer (S)-1 exhibits anti-Parkinson properties. The isomer (R)-1, also known as dexpramipexole, was studied some years ago for the treatment of ALS (amyotrophic lateral sclerosis) [4][5][6]; however, despite the promising results in early clinical studies, its development was discontinued in 2013, due to a lack of efficacy observed in a Phase 3 study [7,8]. More recently, a reduction in the absolute eosinophil count (AEC), coincidentally detected during the ALS-related clinical development, suggested a possible application of dexpramipexole for the treatment of eosinophil-associated disorders [9,10]. The optical purity of this compound is a crucial parameter in the synthetic process, due to the different activities of the two enantiomers.
Taking into account the promising results of our previous experiments [15], in which a Saccharomyces cerevisiae-catalyzed reduction of ketone 3 ( Figure 2) afforded the optically pure (R)-4 ( Figure 2), and considering the few chemoenzymatic approaches [16,17] available for the synthesis of pramipexole, we studied the resolution of the alcohol rac-4 by means of an enzymatic transformation. The enantiopure alcohol (R)-4 proved to be a useful synthon for the synthesis of (S)-1 dihydrochloride monohydrate, the commercial form of this active ingredient, in five steps and 21% overall yield [15].
The alcohols (R)-4 and (S)-4 ( Figure 2) were here studied by means of single-crystal X-ray diffraction, which represents the best approach to unambiguously characterize chiral compounds, as confirmed by our successful application of this technique to the structural analysis of several optically active molecules [18,19].

Lipase-Catalyzed Preparation of Alcohols (R)-4 and (S)-4
Two enzyme-catalyzed preparations of the suitable synthon of (S)-1 are reported in the literature. In a 2006 patent [17], the enzyme-catalyzed resolution of diamine rac-2 was described: among the screened enzymes, only Candida antarctica lipase A (CAL-A), under acylation conditions, and some lipases from Pseudomonas sp. afforded promising results (100% ee determined by HPLC analysis of the non-isolated diamine 2 or the corresponding amide 5) (Scheme 1A). The 2-acetylamino-4,5,6,7tetrahydrobenzothiazole-6-carboxylic acid 7 was resolved in 2007 by Riva et al. through a CAL-Bcatalyzed hydrolysis of the corresponding ester, affording the optically pure ester (S)-6 (by HPLC analysis) at 55% conversion [16] (Scheme 1, B). The (S)-acid can be converted into (S)-1 through a rearrangement, described in a 2005 patent [20]. The optical purity of this compound is a crucial parameter in the synthetic process, due to the different activities of the two enantiomers.
Usually, the rac-1 or its precursor, the amine rac-2 ( Figure 2), are resolved by fractional crystallization of a diastereomeric salt [11,12] or by preparative chiral high-performance liquid chromatography (HPLC) [13,14]. The optical purity of this compound is a crucial parameter in the synthetic process, due to the different activities of the two enantiomers.
Taking into account the promising results of our previous experiments [15], in which a Saccharomyces cerevisiae-catalyzed reduction of ketone 3 (Figure 2) afforded the optically pure (R)-4 ( Figure 2), and considering the few chemoenzymatic approaches [16,17] available for the synthesis of pramipexole, we studied the resolution of the alcohol rac-4 by means of an enzymatic transformation. The enantiopure alcohol (R)-4 proved to be a useful synthon for the synthesis of (S)-1 dihydrochloride monohydrate, the commercial form of this active ingredient, in five steps and 21% overall yield [15].
The alcohols (R)-4 and (S)-4 ( Figure 2) were here studied by means of single-crystal X-ray diffraction, which represents the best approach to unambiguously characterize chiral compounds, as confirmed by our successful application of this technique to the structural analysis of several optically active molecules [18,19].

Lipase-Catalyzed Preparation of Alcohols (R)-4 and (S)-4
Two enzyme-catalyzed preparations of the suitable synthon of (S)-1 are reported in the literature. In a 2006 patent [17], the enzyme-catalyzed resolution of diamine rac-2 was described: among the screened enzymes, only Candida antarctica lipase A (CAL-A), under acylation conditions, and some lipases from Pseudomonas sp. afforded promising results (100% ee determined by HPLC analysis of the non-isolated diamine 2 or the corresponding amide 5) (Scheme 1A). The 2-acetylamino-4,5,6,7tetrahydrobenzothiazole-6-carboxylic acid 7 was resolved in 2007 by Riva et al. through a CAL-Bcatalyzed hydrolysis of the corresponding ester, affording the optically pure ester (S)-6 (by HPLC analysis) at 55% conversion [16] (Scheme 1, B). The (S)-acid can be converted into (S)-1 through a rearrangement, described in a 2005 patent [20]. Taking into account the promising results of our previous experiments [15], in which a Saccharomyces cerevisiae-catalyzed reduction of ketone 3 ( Figure 2) afforded the optically pure (R)-4 ( Figure 2), and considering the few chemoenzymatic approaches [16,17] available for the synthesis of pramipexole, we studied the resolution of the alcohol rac-4 by means of an enzymatic transformation. The enantiopure alcohol (R)-4 proved to be a useful synthon for the synthesis of (S)-1 dihydrochloride monohydrate, the commercial form of this active ingredient, in five steps and 21% overall yield [15].
The alcohols (R)-4 and (S)-4 ( Figure 2) were here studied by means of single-crystal X-ray diffraction, which represents the best approach to unambiguously characterize chiral compounds, as confirmed by our successful application of this technique to the structural analysis of several optically active molecules [18,19].
In 2014, we prepared the alcohol (R)-4 in 79% yield and >98% ee from the 6-ketone 3 by means of a bio-reduction catalyzed by Saccharomyces cerevisiae, common baker's yeast [15]. This synthon was used as a suitable precursor of (S)-1; the same (R)-4 was also used as starting material for the synthesis of (R)-1, after an inversion of configuration carried out through a Mitsunobu reaction, unfortunately characterized by a tedious purification process and low yield (27%) (Scheme 2).
Once the suitable enzyme has been identified, the well-known advantage [21][22][23] offered by an enzymatic kinetic resolution of a racemic alcohol is the obtainment of both enantiomers as optically pure substances, while the reduction of a ketone catalyzed by a microorganism, including Saccharomyces cerevisiae [15], affords only one enantiomer. The 6-hydroxy derivative of the 2aminotetrahydrobenzothiazole, compound 8 [15], could be a substrate of a hydrolytic enzyme, but preliminary attempts of resolution by means of common lipases resulted in a concurrent acylation of the 6-hydroxyl and 2-amino groups (Scheme 3). In 2014, we prepared the alcohol (R)-4 in 79% yield and >98% ee from the 6-ketone 3 by means of a bio-reduction catalyzed by Saccharomyces cerevisiae, common baker's yeast [15]. This synthon was used as a suitable precursor of (S)-1; the same (R)-4 was also used as starting material for the synthesis of (R)-1, after an inversion of configuration carried out through a Mitsunobu reaction, unfortunately characterized by a tedious purification process and low yield (27%) (Scheme 2).
In 2014, we prepared the alcohol (R)-4 in 79% yield and >98% ee from the 6-ketone 3 by means of a bio-reduction catalyzed by Saccharomyces cerevisiae, common baker's yeast [15]. This synthon was used as a suitable precursor of (S)-1; the same (R)-4 was also used as starting material for the synthesis of (R)-1, after an inversion of configuration carried out through a Mitsunobu reaction, unfortunately characterized by a tedious purification process and low yield (27%) (Scheme 2).
Once the suitable enzyme has been identified, the well-known advantage [21][22][23] offered by an enzymatic kinetic resolution of a racemic alcohol is the obtainment of both enantiomers as optically pure substances, while the reduction of a ketone catalyzed by a microorganism, including Saccharomyces cerevisiae [15], affords only one enantiomer. The 6-hydroxy derivative of the 2aminotetrahydrobenzothiazole, compound 8 [15], could be a substrate of a hydrolytic enzyme, but preliminary attempts of resolution by means of common lipases resulted in a concurrent acylation of the 6-hydroxyl and 2-amino groups (Scheme 3).
Once the suitable enzyme has been identified, the well-known advantage [21][22][23] offered by an enzymatic kinetic resolution of a racemic alcohol is the obtainment of both enantiomers as optically pure substances, while the reduction of a ketone catalyzed by a microorganism, including Saccharomyces cerevisiae [15], affords only one enantiomer. The 6-hydroxy derivative of the 2-aminotetrahydrobenzothiazole, compound 8 [15], could be a substrate of a hydrolytic enzyme, but preliminary attempts of resolution by means of common lipases resulted in a concurrent acylation of the 6-hydroxyl and 2-amino groups (Scheme 3). Hence, a screening was carried out choosing as substrate the acetate rac-9, which is the acetyl ester of the alcohol rac-4, obtained by the sodium borohydride-mediated reduction of ketone 3. The most common commercially available lipases and a cross-linked preparation of the protease subtilisin (Alcalase ® CLEA) were used under hydrolytic conditions (phosphate buffer pH 7). Only the lipase from porcine pancreas (PPL) proved to be inactive, whereas the other tested enzymes were able to remove the acetyl group, albeit with an unsatisfactory ee (Table 1 and Scheme 4).
The alcohol (R)-4 for the synthesis of (S)-1 was obtained with all the enzymes, with the exception of the lipase from Candida cylindracea (CCL), which led to the opposite enantiomer (S)-4. The low ee, also observed under alcoholysis conditions, prompted us to study the enzyme-catalyzed transesterification of alcohol rac-4. The use of trifluoroethyl butyrate as acyl donor and the lipase from Pseudomonas fluorescens (PFL) or CAL-A as enzymes, which emulated the reaction conditions reported in the 2006 patent for the resolution of rac-2 [17], did not provide encouraging results: PFL did not work, and CAL-A afforded a low conversion (7%) to the corresponding butyrate, showing a 43% ee.
Furthermore, the irreversible transesterification [24,25], using vinyl acetate as acyl donor and tetrahydrofuran (THF) as solvent, afforded an unsatisfactory ee (Table 2 and Scheme 5), with the best Hence, a screening was carried out choosing as substrate the acetate rac-9, which is the acetyl ester of the alcohol rac-4, obtained by the sodium borohydride-mediated reduction of ketone 3. The most common commercially available lipases and a cross-linked preparation of the protease subtilisin (Alcalase ® CLEA) were used under hydrolytic conditions (phosphate buffer pH 7). Only the lipase from porcine pancreas (PPL) proved to be inactive, whereas the other tested enzymes were able to remove the acetyl group, albeit with an unsatisfactory ee (Table 1 and Scheme 4). Hence, a screening was carried out choosing as substrate the acetate rac-9, which is the acetyl ester of the alcohol rac-4, obtained by the sodium borohydride-mediated reduction of ketone 3. The most common commercially available lipases and a cross-linked preparation of the protease subtilisin (Alcalase ® CLEA) were used under hydrolytic conditions (phosphate buffer pH 7). Only the lipase from porcine pancreas (PPL) proved to be inactive, whereas the other tested enzymes were able to remove the acetyl group, albeit with an unsatisfactory ee (Table 1 and Scheme 4).

Scheme 4.
Enzyme-catalyzed hydrolysis of acetate rac-9. i) Ac2O, py; ii) hydrolytic enzyme, pH 7 buffer. The alcohol (R)-4 for the synthesis of (S)-1 was obtained with all the enzymes, with the exception of the lipase from Candida cylindracea (CCL), which led to the opposite enantiomer (S)-4. The low ee, The alcohol (R)-4 for the synthesis of (S)-1 was obtained with all the enzymes, with the exception of the lipase from Candida cylindracea (CCL), which led to the opposite enantiomer (S)-4. The low ee, also observed under alcoholysis conditions, prompted us to study the enzyme-catalyzed transesterification of alcohol rac-4. The use of trifluoroethyl butyrate as acyl donor and the lipase from Pseudomonas fluorescens (PFL) or CAL-A as enzymes, which emulated the reaction conditions reported in the 2006 patent for the resolution of rac-2 [17], did not provide encouraging results: PFL did not work, and CAL-A afforded a low conversion (7%) to the corresponding butyrate, showing a 43% ee.
Furthermore, the irreversible transesterification [24,25], using vinyl acetate as acyl donor and tetrahydrofuran (THF) as solvent, afforded an unsatisfactory ee (Table 2 and Scheme 5), with the best results being observed in the case of PFL-or CAL-A-catalyzed transformations (68% ee for (R)-9, 20-23% conversion).  These rather promising results encouraged us to perform a solvent screening: besides THF, rac-4 showed an appreciable solubility in acetone, methyl isobutyl ketone (MIBK), dimethyl sulfoxide, and dimethylformamide. Among them, the choice was addressed to the two less hydrophilic options, namely MIBK and acetone. Moreover, toluene was selected to study the outcome of the enzymecatalyzed transesterification when carried out in a hydrophobic solvent, characterized by a lesser ability to strip off the essential water from the surface of the enzymes [26], but unable to solubilize the substrate rac-4. In this case, lipases afforded good reaction rates but a low enantioselectivity. Conversely, when employing MIBK as solvent, moderate reaction yield and enantioselectivity were observed. In acetone, PFL and CCL were found to be inactive, most likely due to the higher hydrophilicity of acetone compared to MIBK; however, CAL-A showed a good enantioselectivity, albeit at the expense of the reaction rate. To establish if the results obtained with MIBK and acetone were satisfactory for practical purposes, the E (enantiomeric ratio) values were calculated according to Rakels et al. [27]. An enzymatic resolution displays a preparative significance only if its E is >10 [28][29][30][31]. While the E observed in the case of the CAL-A-catalyzed resolution in MIBK was 12, which is enough to obtain enantiomers with ee values >98% after a second enzymatic resolution [29], the E of the CAL-A-catalyzed transformation in acetone was 28, thus indicating the possibility to obtain more satisfactory results in terms of optical purity and yields (Table 3). These rather promising results encouraged us to perform a solvent screening: besides THF, rac-4 showed an appreciable solubility in acetone, methyl isobutyl ketone (MIBK), dimethyl sulfoxide, and dimethylformamide. Among them, the choice was addressed to the two less hydrophilic options, namely MIBK and acetone. Moreover, toluene was selected to study the outcome of the enzyme-catalyzed transesterification when carried out in a hydrophobic solvent, characterized by a lesser ability to strip off the essential water from the surface of the enzymes [26], but unable to solubilize the substrate rac-4. In this case, lipases afforded good reaction rates but a low enantioselectivity. Conversely, when employing MIBK as solvent, moderate reaction yield and enantioselectivity were observed. In acetone, PFL and CCL were found to be inactive, most likely due to the higher hydrophilicity of acetone compared to MIBK; however, CAL-A showed a good enantioselectivity, albeit at the expense of the reaction rate. To establish if the results obtained with MIBK and acetone were satisfactory for practical purposes, the E (enantiomeric ratio) values were calculated according to Rakels et al. [27]. An enzymatic resolution displays a preparative significance only if its E is >10 [28][29][30][31]. While the E observed in the case of the CAL-A-catalyzed resolution in MIBK was 12, which is enough to obtain enantiomers with ee values >98% after a second enzymatic resolution [29], the E of the CAL-A-catalyzed transformation in acetone was 28, thus indicating the possibility to obtain more satisfactory results in terms of optical purity and yields (Table 3). Following the findings by Herbst, Peper, and Niemeyer [26], we also studied the possibility to mitigate the tendency of acetone to remove the essential water molecules from the enzyme by adding water (1 µL of distilled water/mL of acetone) or mixing it with n-hexane, with the aim of reaching a good reaction rate together with a similar enantioselectivity to that obtained in pure acetone. The performed tests and the obtained results are summarized in Table 4. The results reported in Table 4 show that the addition of n-hexane had a great impact on both the enantioselectivity and the reaction rate. The higher the amount of n-hexane in the reaction medium, the higher the reaction rate was, but at the expense of the enantioselectivity. In particular, comparing pure acetone to the acetone/n-hexane 1:1 mixture, we observed an increase in the reaction rate (from 168 h for a 34% conversion to 32 h for a 69% conversion) and a decrease in the enantioselectivity (the E values changed from 28 to 14). These findings are in agreement with those reported by Herbst et al. [26]: the conversion rate of a kinetic resolution catalyzed by a lipase decreases with a stepwise increase of the solvent hydrophilicity (in our case with the increase of the acetone content); at the same time, the higher the hydrophilicity of the reaction medium, the higher the enantioselectivity. This behavior can be explained by the ability of the hydrophilic medium to extract the essential water of the enzymes, thus lowering their activity. At the same time, an enzyme with less essential water has a highly rigid structure, which allows it to better discriminate between the two enantiomers, leading to an increased enantioselectivity. In this context, it is interesting to note the linear dependence between the E-value and the amount of n-hexane in acetone, observed in the selected range ( Figure 3). Moreover, the drop in activity of CAL-A with the addition of 0.1% water to the solvent (336 h for a 32% conversion) was previously explained by Herbst et al. [26]: under certain conditions, the addition of water can lead to enzyme agglomeration up to denaturation.  According to the results summarized in Table 4 and Figure 3, the catalytic system employing the mixture acetone/n-hexane 4:1 as solvent can be considered a good compromise between enantioselectivity and reactivity, with the aim to synthesize enantiopure (R)-and (S)-4.
The first CAL-A-catalyzed transesterification was carried out starting from the racemic alcohol rac-4. 91% ee (S)-4 and 75% ee (R)-9 were obtained (E-value of 22) at 58% conversion (about 300 h) after chromatographic purification. The 91% ee alcohol (S)-4 was used as substrate in a second CAL-A-catalyzed transesterification in order to obtain enantiopure (S)-4 (>99% ee at 12% conversion). The 75% ee acetate (R)-9 was hydrolyzed to yield enantioenriched (R)-4, used as substrate in a second CAL-A-catalyzed transesterification. This second enzymatic resolution was stopped at 78% conversion. (R)-9 was isolated by means of chromatographic purification, affording >98% ee (R)-4 after hydrolysis.  According to the results summarized in Table 4 and Figure 3, the catalytic system employing the mixture acetone/n-hexane 4:1 as solvent can be considered a good compromise between enantioselectivity and reactivity, with the aim to synthesize enantiopure (R)-and (S)-4.

X-ray Crystallography
The 3D structures of (R)-4 and (S)-4 were unambiguously determined by means of single-crystal X-ray diffraction and are reported in Figure 4 as ORTEP [32] representations. Both compounds crystallized in the monoclinic P2 1 space group: the asymmetric unit (ASU) of (R)-4 shows the (R)-alcohol interacting with an ordered water molecule, while in (S)-4 the ASU accommodates two independent molecules of the (S)-enantiomer.
Overall, the two compounds are characterized by a flat region corresponding to the thiazole ring of the tetrahydrobenzothiazole, with the amide moiety lying on the same plane; the C9 methyl group deviates slightly, forming an angle of 6. (the values in the square brackets refer to the A-labeled molecule). The different θ values indicate that the cyclohexene ring assumes opposite half-chair conformations in the two enantiomers, while the two hydroxyl groups remain equatorially oriented in both structures. Moreover, the puckering data evidence small differences in the shape of the cyclohexenes in (S)-4, which justifies the presence of two distinct molecules in the crystal ASU. The superposition of the two forms of (S)-4 is shown in Figure 5, which reveals that their most significant structural difference is represented by the torsion angles ε (ε = H1-O1-C5-C6) involving the hydroxyl group, which are 66.7(1) • and 27.1(1) • .

X-ray Crystallography
The 3D structures of (R)-4 and (S)-4 were unambiguously determined by means of single-crystal X-ray diffraction and are reported in Figure 4 as ORTEP [32] representations. Both compounds crystallized in the monoclinic P21 space group: the asymmetric unit (ASU) of (R)-4 shows the (R)alcohol interacting with an ordered water molecule, while in (S)-4 the ASU accommodates two independent molecules of the (S)-enantiomer.  Overall, the two compounds are characterized by a flat region corresponding to the thiazole ring of the tetrahydrobenzothiazole, with the amide moiety lying on the same plane; the C9 methyl group deviates slightly, forming an angle of 6.1(1)° in (R)-4, and 8.1(1)° and 8.8(1)° in (S)-4. The cyclohexene moiety is in half-chair conformation, with the shared C2-C3 double bond belonging to the plane formed by the heterocycle. The Cremer-Pople [33] ring puckering parameters for the six-membered while the two hydroxyl groups remain equatorially oriented in both structures. Moreover, the puckering data evidence small differences in the shape of the cyclohexenes in (S)-4, which justifies the presence of two distinct molecules in the crystal ASU. The superposition of the two forms of (S)-4 is shown in Figure 5, which reveals that their most significant structural difference is represented by the torsion angles ε (ε = H1-O1-C5-C6) involving the hydroxyl group, which are 66.7(1)° and 27.1(1)°. Figure 5. Superposition of the two independent molecules of the asymmetric unit of (S)-4.
The analysis of the intermolecular contacts shows considerable differences between the two crystals, even though hydrogen bonds remain the principal driving force in the solid-state assembly ( Figure 6). In (R)-4, four H-bonds consolidate the crystal packing, with the water playing a central role in bridging the alcohol molecules: its oxygen interacts with the amidic hydrogen, while the two hydrogens form contacts involving the oxygens of the hydroxyl and amide functions. Overall, the water molecules create channels, crossing the crystal along the c axis ( Figure 7). Another H-bond is established between the hydrogen of the hydroxyl group and the nitrogen of the thiazole ring. The crystal structure of (S)-4 is stabilized by just two hydrogen bonds, between the amidic hydrogen and the hydroxyl oxygen, and between the alcoholic hydrogen and the C=O of the amide group. No intermolecular interactions are observed between the two independent molecules of the ASU. A summary of the H-bonds of (R)-4 and (S)-4 is reported in Table 5, along with a complete account of their geometry. The analysis of the intermolecular contacts shows considerable differences between the two crystals, even though hydrogen bonds remain the principal driving force in the solid-state assembly ( Figure 6). In (R)-4, four H-bonds consolidate the crystal packing, with the water playing a central role in bridging the alcohol molecules: its oxygen interacts with the amidic hydrogen, while the two hydrogens form contacts involving the oxygens of the hydroxyl and amide functions. Overall, the water molecules create channels, crossing the crystal along the c axis ( Figure 7). Another H-bond is established between the hydrogen of the hydroxyl group and the nitrogen of the thiazole ring. The crystal structure of (S)-4 is stabilized by just two hydrogen bonds, between the amidic hydrogen and the hydroxyl oxygen, and between the alcoholic hydrogen and the C=O of the amide group. Table 5, along with a complete account of their geometry.      Table 5. Hydrogen-bond geometry of (R)-4 and (S)-4 (arbitrary atom-numbering scheme used in Figure 4). Values in square brackets refer to the A-labeled molecule in the ASU of (S)-4.   Table 5. Hydrogen-bond geometry of (R)-4 and (S)-4 (arbitrary atom-numbering scheme used in Figure 4). Values in square brackets refer to the A-labeled molecule in the ASU of (S)-4.
The employed HPLC unit was a Merck-Hitachi (Hitachi Ltd., Tokyo, Japan), equipped with a UV detector model L-4250, pump system model L-6200 and a chromato-integrator model D-2500. The column employed in the analyses was a Daicel Chiralpak IA (Daicel Group, Chiral Technologies Europe, Illkirch Cedex, France). The dimension of the column is 250 mm × 4.6 mm, 5 µm. The elution was in isocratic mode with n-hexane/2-propanol 8:2. The flow rate was 0.7 mL/min. All the samples were measured at λ = 254 nm and 25 • C. Representative retention times (R t ): (R)-4 10.7 min; (S)-4 15.3 min; (R)-and (S)-9 were not resolved and eluted at about 13.6 min. In order to establish the enantiomeric excess of acetate 9, the compound was hydrolyzed with sodium hydroxide in methanol (1% w/v) to the corresponding alcohol 4.
High-resolution mass spectra (HRMS) were recorded on Impact HDTM UHR-QqToF (Bruker Daltonics, Billerica, MA, USA); acquisition parameters were set as follows: Funnel1 RF 400 Vpp, Funnel2 RF 400 Vpp, Hexapole RF 600 Vpp, transfer time 95 µs, pre-pulse storage 12 µs, 40-650 m/z window. Calibration was performed online using ESI-L Low Concentration Tuning Mix (Agilent Technologies); the samples were dissolved in methanol at a final concentration of 0.1 µg/mL and injected at an infusion rate of 3 µL/min. Data acquisition and analysis were accomplished with Bruker Compass Data Analysis 4.1 software (Bruker Compass, Billerica, MA, USA).
Nuclear magnetic resonance (NMR) spectra were recorded on an AVANCE 500 spectrometer (Bruker, Billerica, MA, USA) equipped with a 5 mm broadband inverse probe (BBI) with field z-gradient operating at 500.13 and 125.76 MHz for 1 H and 13 C, respectively. NMR spectra were recorded at 300 K in CDCl 3 (isotopic enrichment 99.95%) solution. The data were collected and processed by XWIN-NMR software (3.5, Bruker) running on a PC with Microsoft Windows 7 (Redmond, WA, USA). The samples (10 mg), were dissolved in CDCl 3 (0.75 mL) in a 5 mm NMR tube. Acquisition parameters for 1D were as follows: 1 H spectral width of 5000 Hz and 32 K data points providing a digital resolution of ca. 0.305 Hz per point, relaxation delay 10 s; 13 C spectral width of 29,412 Hz and 64 K data points providing a digital resolution of ca. 0.898 Hz per point, relaxation delay 2 s. The experimental error in the measured 1 H-1 H coupling constants was ± 0.5 Hz. Chemical shifts (δ) of the 1 H NMR and 13 C NMR spectra are reported in ppm using the signal for residual solvent resonance as internal standard ( 1 H NMR: CDCl 3 7.26 ppm; 13 C NMR: CDCl 3 77.0 ppm (central line)). The splitting pattern abbreviations are as follows: s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet, and bs, broad signal. For two-dimensional experiments, Bruker microprograms using gradient selection (gs) were applied. All two-dimensional spectra (COSY and HSQC) were acquired with 2048 data points for t 2 and 256 for t 1 increments. The confirmation of the structure of compounds 9 was obtained by NMR analysis, in particular through proton and carbon 1D-NMR spectra, as well as by 2D-NMR homo-correlation (COSY) and hetero-correlation (HSQC) experiments.
The NMR and HRMS spectra, and the HPLC chromatograms are available in the Supplementary Materials.

Chemistry
Compounds 4 and 8 were synthesized according to the methods previously described in Ref. [15].

General Procedure for Enzyme-Catalyzed Hydrolysis
rac-2-Acetamido-4,5,6,7-tetrahydrobenzo[d]thiazol-6-yl acetate (rac-9) (13 mg; 0.05 mmol) was suspended in phosphate buffer pH 7 (5 mL) in a 10 mL glass vial followed by addition of the enzyme (an amount corresponding to about 100 U). The mixture was shaken at 300 rpm and room temperature. A sample of about 50 µL was collected regularly, diluted with water, and extracted with ethyl acetate. The organic phase was concentrated under reduced pressure. The residue was dissolved in 2-propanol (about 50 µL) and a portion (10 µL) was injected in a HPLC system equipped with a Chiralpak IA column. To a solution of rac-4 (500 mg, 2.4 mmol) in anhydrous acetone (200 mL) and n-hexane (50 mL), vinyl acetate (0.7 mL, 7.6 mmol) and CAL-A (3 g, 4860 U) were added sequentially. The reaction mixture was shaken at room temperature and 300 rpm in a screw-cap flask. The reaction progress was monitored by HPLC, equipped with a Chiralpak IA column. At 58% conversion (reaction time: about 300 h), the enzyme was removed by filtration and the solvent was evaporated at reduced pressure. The residue was purified by silica gel column chromatography. By elution with n-hexane/ethyl acetate 6:4, acetate (R)-9 was recovered (328 mg, 54%). Elution with dichloromethane/methanol 95:5 afforded enantioenriched (S)-4 (190 mg, 38%, 91% ee). A solution of 1% w/v sodium hydroxide in methanol (15 mL, 3.8 mmol) and enantioenriched (R)-9 (300 mg, 1.18 mmol), obtained from the enzyme-catalyzed resolution described in Section 3.3.4, was stirred at room temperature. The reaction progress was monitored by TLC analysis (dichloromethane/methanol 95:5) until starting material disappearance (2 h). The pH of the reaction mixture was adjusted to 7 by addition of 1 M HCl and methanol was removed under reduced pressure. Water was added (5 mL) and the obtained aqueous phase was extracted with ethyl acetate (5 × 5 mL). The collected organic phases were dried over sodium sulfate; after filtration, the solvent was removed at reduced pressure, affording enantioenriched (R)-4 (212 mg, 1.0 mmol, 85%) which was used directly in the next step without any further purification. The 75% ee of enantioenriched (R)-4 was determined by HPLC analysis on a Chiralpak IA column. To a solution of enantioenriched (S)-4 (150 mg, 0.71 mmol, 91% ee), obtained as described in Section 3.3.4, in anhydrous acetone (60 mL) and n-hexane (15 mL), vinyl acetate (210 µL, 2.3 mmol) and CAL-A (0.9 g, 1458 U) were sequentially added. The reaction mixture was shaken at room temperature and 300 rpm in a screw-cap flask. The reaction progress was monitored by HPLC equipped with a Chiralpak IA column. The reaction was stopped at 12% conversion (reaction time: about 48 h), the enzyme was removed by filtration and the solvent was evaporated at reduced pressure. The residue was purified by silica gel column chromatography. By elution with n-hexane/ethyl acetate 6:4, acetate (R)-9 was recovered (18 mg, 10%). Elution with dichloromethane/methanol 95:5 afforded enantiopure (S)-4 (123 mg, 82%, >99% ee). The chemical-physical properties of (S)-4 were in agreement with the reported ones [15] To a solution of enantioenriched (R)-4 (200 mg, 0.94 mmol, 75% ee), obtained as described in Section 3.3.5, in anhydrous acetone (80 mL) and n-hexane (20 mL), vinyl acetate (280 µL, 3.02 mmol) and CAL-A (1.2 g, 1944 U) were added sequentially. The reaction mixture was shaken at room temperature and 300 rpm in a screw-cap flask. The reaction progress was monitored by HPLC equipped with a Chiralpak IA column. The reaction was stopped at 78% conversion (reaction time: about 144 h), the enzyme was removed by filtration and the solvent was evaporated at reduced pressure. The residue was purified by silica gel column chromatography. By elution with n-hexane/ethyl acetate 6:4, acetate (R)-9 was recovered (174 mg, 72%,). Elution with dichloromethane/methanol 95:5 afforded (S)-4 (30 mg, 15%). The hydrolysis of enantiopure (R)-9 (100 mg, 0.39 mmol), obtained from the enzyme-catalyzed resolution described in Section 3.3.7, was carried out under the same conditions described in Section 3.3.5 affording enantiopure (R)-4 (74 mg, 0.35 mmol, 90%). The >98% ee of enantiopure (R)-4 was determined by HPLC analysis on a Chiralpak IA column. The chemical-physical properties of (R)-4 were in agreement with the reported ones [15].

X-ray Crystallography
Diffraction-quality crystals were obtained at room temperature by slow evaporation of an ethanol/water 1:1 solution for (R)-4, and of a methanol/water 1:1 solution for (S)-4. After a week, transparent crystals precipitated as prisms.
X-ray diffraction data for (R)-4 were collected at room temperature on a Bruker Apex II CCD diffractometer, using graphite-monochromatized Mo-Ka X-radiation (λ = 0.71073 Å); for (S)-4, a Bruker-Axs CCD-based three-circle diffractometer was used, working with the same X-ray source (λ = 0.7107 Å) and environmental conditions. The collected information was processed with the SAINT [34] software for data reduction (including intensity integration, background, Lorentz and polarization corrections) and for the determination of accurate unit-cell dimensions. Absorption effects were empirically evaluated by the SADABS [35] software, and absorption correction was applied to the data. The structures were solved by direct methods (SIR-14) [36] and completed by iterative cycles of full-matrix least squares refinement on Fo 2 and DF synthesis using SHELXL-17 [37] on the WingX.v2014.1 suite [38]. The hydrogen atoms bonded to carbons were included at geometrically calculated positions and refined using a riding model.

Conclusions
The bicyclic compound pramipexole is a pharmacologically active substance, showing anti-Parkinson effects only when its stereocenter has the (S)-configuration. Conversely, its (R)enantiomer is still under investigation to identify possible therapeutic applications. For these reasons, we planned to prepare the optically pure synthons of (R)-and (S)-pramipexole, by means of the enzymatic resolution of the suitable secondary alcohol. This result was achieved through a double kinetic resolution, catalyzed by CAL-A; the reaction conditions were studied and optimized, also considering the choice of the solvent. Starting from the racemic alcohol 4, a 4:1 v/v mixture of acetone/n-hexane allowed us to obtain the (R)-alcohol 4 and (S)-alcohol 4 in >98% ee and >99% ee and 30% and 31% overall yields, respectively. The single-crystal X-ray analysis of these compounds unambiguously defined the 3D structural features of the two enantiomers and revealed a considerably different crystal packing.
These two advanced chiral intermediates can be converted to (S)-and (R)-pramipexole following the synthetic route previously optimized in our laboratory. A comparison between the two biocatalytic approaches put in evidence that the previous method, consisting in the enantioselective reduction of the suitable ketone 3 catalyzed by Saccharomyces cerevisiae, provided only the (R)-alcohol 4 in >98% ee. The (S)-alcohol 4 was obtained, starting from the (R)-alcohol 4, through an inversion of configuration operated by means of the Mitsunobu reaction that furnished a complex crude reaction mixture. After a careful purification process the (S)-alcohol 4 was recovered in poor yield (27%). By contrast, the new chemoenzymatic approach afforded both the (R)-and (S)-alcohol 4 in good yields (30% and 31% overall yields, respectively) starting from rac-4; this is a remarkable result considering that, with an enzymatic kinetic resolution of a racemate, the maximum theoretical yield for each enantiomer is 50%.