Chemoenzymatic Stereodivergent Synthesis of All the Possible Stereoisomers of the 2,3-Dimethylglyceric Acid Ethyl Ester

: 2,3-dihydroxy-2-methylbutyric acid, also known as 2,3-dimethylglyceric acid, consti-tutes the acyl and/or the alcoholic moiety of many bioactive natural esters. Herein, we describe a chemoenzymatic methodology which gives access to all the four possible stereoisomers of the 2,3-dimethylglyceric acid ethyl ester. The racemic ethyl α -acetolactate, produced by the N -heterocycle carbene ( N HC)-catalyzed coupling of ethyl pyruvate and methylacetoin was employed as the starting material. The racemic mixture was resolved through ( S )-selective reductions, promoted by the acetylacetoin reductase (AAR) affording the resulting ethyl (2 R ,3 S )-2,3-dimethylglycerate; the isolated remaining ( S )-ethyl α -acetolactate was successively treated with baker’s yeast to obtain the corresponding (2 S ,3 S ) stereoisomer. syn -2,3-Dimethylgliceric acid ethyl ester afforded by reduc-ing the rac - α -acetolactate with NaBH 4 in the presence of ZnCl 2 was kinetically resolved through selective acetylation with lipase B from Candida antarctica (CAL-B) and vinyl acetate to access to ( 2S ,3 R ) stereoisomer. Finally, the (2 R ,3 R ) stereoisomer, was prepared by C3 epimerization of the (2 R ,3 S ) stereoisomer recovered from the above kinetic resolution, achieved through the TEMPO-mediated oxidation, followed by the reduction of the produced ketone with NaBH 4 . The resulting 2,3-dimethylglycertate enriched in the (2 R ,3 R ) stereoisomer was submitted to stereospeciciﬁc acetylation with vinyl acetate and CAL-B in order to separate the major stereoisomer. The entire procedure enabled conversion of the racemic α -acetolactate into the four enantiopure stereoisomers of the ethyl 2,3-dihydroxy-2-methylbutyrate with the following overall yields: 42% for the (2 R ,3 S ), 40% for the (2 S ,3 S ), 42% for the (2 S ,3 R ) and 20% for the (2 R ,3 R ).


Introduction
The curative effects of traditional pharmaceutics are frequently related to the activity of secondary metabolites produced by plants or microorganisms. Because of the low concentration of such substances within natural sources, many efforts are devoted to the identification of their chemical structure in order to develop synthetic strategies which could allow their therapeutic exploitation. Knowing the precise structure is also of pivotal importance to understand the mechanisms of action and to design derivatives with best pharmacological performance [1,2]. Many bioactive natural products are chiral compounds produced by living organisms as single stereoisomers whose artificial enantiomers often result less active, if not noxious [3,4]. From an economic and environmental point of view, a sustainable industrial synthesis of these metabolites should require highly efficient and selective reactions so as to reduce the number of steps, simplify the purification procedures and consequently reduce waste formation and energy costs [5]. From this perspective, biocatalysis nowadays offers a broad range of easily accessible enzymes to perform challenging reactions with excellent results in terms of yield and selectivity [6][7][8]. Moreover, thanks to the recent advances in bioinformatic and protein engineering which have expanded the biocatalytic toolbox [9], some exquisite examples of total enzymatic syntheses have been recently reported [10][11][12]. Moving in this field, we recently highlighted that the combined use of a thiamine diphosphate (ThDP)-dependent lyase and a NADH-dependent dehydrogenase enables the preparation of enantiopure 1-substituted-1,2-propanediols [13]. Some of the compounds obtained in this work are secondary metabolites or metabolite moieties, produced by living organisms [14][15][16]. One representative example is 2,3-dihydroxy-2-methylbutanoic acid, also known as 2,3-dimethylglyceric acid, whose different stereoisomeric forms are contained in a number of bioactive natural esters ( Figure 1).
Catalysts 2021, 11, x FOR PEER REVIEW 2 of 10 simplify the purification procedures and consequently reduce waste formation and energy costs [5]. From this perspective, biocatalysis nowadays offers a broad range of easily accessible enzymes to perform challenging reactions with excellent results in terms of yield and selectivity [6][7][8]. Moreover, thanks to the recent advances in bioinformatic and protein engineering which have expanded the biocatalytic toolbox [9], some exquisite examples of total enzymatic syntheses have been recently reported [10][11][12]. Moving in this field, we recently highlighted that the combined use of a thiamine diphosphate (ThDP)dependent lyase and a NADH-dependent dehydrogenase enables the preparation of enantiopure 1-substituted-1,2-propanediols [13]. Some of the compounds obtained in this work are secondary metabolites or metabolite moieties, produced by living organisms [14][15][16]. One representative example is 2,3-dihydroxy-2-methylbutanoic acid, also known as 2,3-dimethylglyceric acid, whose different stereoisomeric forms are contained in a number of bioactive natural esters ( Figure 1). For instance, the phytotoxin phomozin, responsible for the stem cankering of sunflower during infection by Phomopsis helianthi is an ester of orsellinic acid and (2S,3S)-2,3-dimethylglyceric acid [17,18]. In goncarins A and B, two secoiridoids from Gonocaryum calleryanum, the 3-hydroxyl group and the carboxylic group of 2,3-dimethylglyceric acid, are involved in ester linkages with complementary functional groups of the secoiridoid part, resulting in macrocyclic lactones with anti-inflammatory activity [19,20]. In some clerodane diterpenoids [21][22][23] and furoeudesmane sesquiterpenes [24] with feeding stimulating and antifeedant activity, respectively, the alcoholic terpenoids part is esterified with 2,3-dimethylglyceric acid. Likewise, pyrrolizidine [25] and dehydropyrrolizidine alkaloids pointed out as potential hepatotoxic metabolites [26] or antitumor prodrugs [27] show ester linkages with 2,3-dimethylglyceric acid. Furthermore, the steroidal alkaloids protoveratrines B and C [28] and neogermbudine [29] are worth mentioning. As for the above clerodanes and furoeudesmanes, these plant metabolites also show a 2,3-dimethylglycerate ester in position 3, which seems to be responsible for their documented neurotoxic activity [29,30]. It is worth noting that most the above studies report only the relative stereochemistry of the dimethylglycerate fragment (except for phomozin, where the absolute stereochemistry was ascertained) [17]. This means that the stereoselective access to all the four stereoisomers of the 2,3-dimethylglyceric acid, For instance, the phytotoxin phomozin, responsible for the stem cankering of sunflower during infection by Phomopsis helianthi is an ester of orsellinic acid and (2S,3S)-2,3dimethylglyceric acid [17,18]. In goncarins A and B, two secoiridoids from Gonocaryum calleryanum, the 3-hydroxyl group and the carboxylic group of 2,3-dimethylglyceric acid, are involved in ester linkages with complementary functional groups of the secoiridoid part, resulting in macrocyclic lactones with anti-inflammatory activity [19,20]. In some clerodane diterpenoids [21][22][23] and furoeudesmane sesquiterpenes [24] with feeding stimulating and antifeedant activity, respectively, the alcoholic terpenoids part is esterified with 2,3-dimethylglyceric acid. Likewise, pyrrolizidine [25] and dehydropyrrolizidine alkaloids pointed out as potential hepatotoxic metabolites [26] or antitumor prodrugs [27] show ester linkages with 2,3-dimethylglyceric acid. Furthermore, the steroidal alkaloids protoveratrines B and C [28] and neogermbudine [29] are worth mentioning. As for the above clerodanes and furoeudesmanes, these plant metabolites also show a 2,3-dimethylglycerate ester in position 3, which seems to be responsible for their documented neurotoxic activity [29,30]. It is worth noting that most the above studies report only the relative stereochemistry of the dimethylglycerate fragment (except for phomozin, where the absolute stereochemistry was ascertained) [17]. This means that the stereoselective access to all the four stereoisomers of the 2,3-dimethylglyceric acid, would allow the determination of the absolute configuration of the above bioactive natural products, making it possible to design asymmetric total synthetic pathways. The literature reports only a few examples of stereoselective preparation of 2,3-dimethylglyceric acid esters. The racemic syn and anti methyl esters were prepared by OsO 4 oxidation of the corresponding transand cis-2methyl-2-butenoates, respectively [20]. Although it uses inexpensive substrates, this route relies on the use of a toxic oxidant and does not show enantioselectivity. On the other hand, the enantioselective preparation of the (2R,3R)-and (2S,3S)-2,3-dimethylglycerate ethyl esters via addition of the sterically hindered 2-t-butyl-5-methyl-2-phenyl-1,3-dioxolan-4one litium enolate to acetaldehyde reported by Greiner et al. [18] has a low atom economy because of the large amount of unrecoverable waste produced by employing the chiral auxiliary [17]. More recently, an iron(II) complex was exploited as catalyst along with aqueous H 2 O 2 oxidant as a green alternative to the OsO 4 for the enantioselective cis-hydroxylation of the phenyl trans-2-methyl-2-butenoate yielding the corresponding (2S,3R) diol with 87% yield and >99% ee [31]. Finally, as above mentioned, we recently reported the enzymatic synthesis of the (2R,3S)-2,3-dimethylglygeric acid ethyl ester through the enzymatic reduction of the ethyl (R)-α-acetolactate previously prepared by benzoin-type condensation of methylacetoin and ethyl pyruvate catalyzed by a thiamine diphosphate-dependent lyase [13]. Inspired by this last work, we herein report a stereodivergent chemoenzymatic strategy for the preparation of all the four stereoisomers of the 2,3-dimethylglyceric acid ethyl ester starting from cheap and safe reagents, using easily available biological and chemical catalysts.

Results and Discussion
Within a previous study we reported the enzymatic synthesis of optically pure ethyl ester of the (2R,3S)-2,3-dimethylglyceric acid [13]. Searching in the literature for characterization data, we realized the biological relevance of the different stereoisomers of this acid as well as the limited number of stereoselective synthetic routes for this compound. Moved from these observations, we envisaged the racemic ethyl α-acetolactate (Scheme 1, compound 3) as the potential starting point for a stereodivergent synthesis leading to all the four possible stereoforms of ethyl 2,3-dimethylglycerate (Scheme 1, product 4). The starting compound 3, can be easily produced from the cross-benzoin type coupling of 2,3butanedione and ethyl pyruvate (Scheme 1, compounds 1 and 2, respectively) promoted by the N-heterocycle carbene (NHC) catalysts generated in situ by treating the thiamine hydrochloride with trimethylamine [32] (Scheme 1, reaction a).

Enzymatic Kinetic Resolution of the Racemic α-Acetolactate: Synthesis of Ethyl
Taking into account the recently highlighted preference of the NADH-dependent acetylacetoin reductase (AAR) for the (R) enantiomer of ethyl α-acetolactate [13], we engaged the kinetic resolution of 3 through the enantioselective reduction of the carbonyl group. The optimized reaction performed in 50 mM phosphate buffer at pH 6.5 in the presence of sodium formate (5 equivalents) and formate dehydrogenase (FDH) for the NADH recycle, afforded the expected (2R,3S)-4 in 42% isolated yield (>95%, ee; d.r. 92%) (Scheme 1, reaction b). The enantiopure (2S)-3 recovered from the above reaction mixture (45% yield, > 95% ee), could have been enantioselectively reduced in order to obtain the (2S,3S) or to the (2S,3R) stereoisomers of 4. To the best of our knowledge, the (R)-selective enzymatic reduction of 3 was never reported in the literature, while it is known that whole cells of baker's yeast (BY) are able to reduce racemic 3, giving a 50/50 diasteromeric mixture of (2S,3S)-and (2R,3S)-4 [33]. Following this example, we treated an aqueous solution of (S)-3 with BY in the presence of glucose. After 6 h at 30 • C, the expected (2S,3S)-4 (ee > 95%) was obtained in 90% isolated yield (Scheme 1, reaction c).  Once we paved the way for the synthesis of two stereoisomers of 4, we moved to investigate a route to obtain the other two stereoisomers, namely the syn-(2S,3R)and the anti-(2R,3R)-4. We started exploring the effect of coordinating metals on the reduction of racemic 3 with NaBH 4 . While the reaction conducted in diethyl-ether/ethanol in the absence of strong coordinating metals afforded a syn/anti mixture (d.r. 30/70), the addition of ZnCl 2 (1 equiv.) to the reaction mixture, led to the diastereoselective formation of the only syn-4 (Scheme 1, reaction d) due to a chelation control [34]. The so-obtained racemic syn-4 contains one of the two desired stereoisomers, namely (2S,3R)-4. In order to avoid troublesome and inconvenient chromatographic separation, we studied the enzymatic acylation as an approach for the kinetic resolution of this racemate. We did not find in literature, examples of the enzymatic acylations conducted on the compound 4. On the contrary, the kinetic resolution of the less hindered ethyl 3-hydroxybutyrate was successfully performed by using Candida antarctica lipase B (CAL-B) as the catalyst, and vinyl acetate as the acylating agent [35]. Therefore, we extended this approach to the racemic syn-4. Taking into account that CAL-B is notoriously inactive toward tertiary alcohols [36], it remained to be verified how the hindrance of the quaternary center should have affected the rate and the stereoselectivity of the acetylation of the hydroxyl group on position 3. Fortunately, the reaction performed with CAL-B (20% w/w) in vinyl acetate without additional solvents led, after 4 h, to the complete conversion of the (2S,3R)-4 to the corresponding 3-O-acetyl derivative (2S,3R)-5 (Scheme 1, reaction e). We also verified that prolonged reaction time (10 h) did not reduce the enantiomeric excess. The product was isolated in 42% yield by column chromatography, and at the same time, the unreacted (2R,3S)-4 was recovered (41% yield, >95% ee). The acetyl derivative 5 was then dissolved in cyclohexane and reacted with ethanol (3 equivalent) in the presence of CAL-B (20% w/w). Thus, the desired (2S,3R)-4 was isolated in pure form (95% yield, >95% ee) simply by filtering out the enzyme and evaporating the solvents (Scheme 1, reaction f).

General Information
All commercially available reagents were used as received without further purification, unless otherwise stated. Formate dehydrogenase form Candida boidinii (0.45 U/mg) was purchased from Fluka. The Candida antarctica lipase B Lipozyme 435 ® (CAL-B) was obtained from Novozymes. The recombinant acetylacetoin reductase (AAR) was obtained as described [13]. The baker's yeast was purchased from Lesaffre Italia. Reactions were monitored by TLC on silica gel 60 F254 with detection by charring with phosphomolybdic acid. Flash column chromatography was performed on silica gel 60 (230-400 mesh). 1 H and 13 C NMR spectra were acquired at room temperature on spectrometers operating at 300 and 400 MHz; CDCl 3 was employed as a solvent. The chemical shifts (δ) are given in ppm by taking as reference the solvent signal. High-resolution mass spectrometry (HRMS) analyses were performed in positive ion mode on an Agilent 6520 HPLC-Chip Q/TOF-MS nanospray system equipped with a time-of-flight, quadrupole or hexapole unit as analyzer.
Optical rotation values were acquired at 20 ± 2 • C in CHCl 3 as solvent; [α] 20 D values are given in 10 -1 deg cm 2 g -1 . GC analyses were performed using a flame ionization detector and a Megadex 5 column (25 m × 0.25 mm). The samples, free from solvents (about 1 mg), were dissolved in trifluoroacetic anhydride (0.1 mL) and the solution was kept at room temperature for 20 min. After dilution with dichloromethane (1.0 mL), 1.0 µL of the resulting solution was injected. The products were detected using the following temperature program: from 80 • C, 10 • C min −1 up to 200 • C. For retention times, see Supplementary Materials (S10-S16).

AAR Activity Assay
The enzyme activity was measured by following the disappearance of NADH (decrease of absorbance at 340 nm) during the reduction of racemic 3 as follows. To a solution of NADH (0.2 mM) and racemic 3 (5 mM) in 50 mM phosphate buffer at pH 6.5 (1 mL) the AAR was added and the change in absorbance at 340 nm was monitored for 3 min. One activity unit (U) is defined as the enzyme amount needed to reduce 1 µmol of (S)-3 in one minute, under the above reaction conditions.

Conclusions
The herein reported chemoenzymatic methodology allows accessing to the ethyl esters of the four possible stereoisomers of the biologically relevant 2,3-dimethylglyceric acid. The products were obtained as pure enantiomers (ee > 95%) with good overall yields (from 20 to 42%). All the coproducts of the kinetic resolution steps are employed as intermediates for the preparation of one of the other enantiomers, minimizing waste production. All the reagents and the catalysts employed are commercially available, other than AAR, whose gene sequence and cloning procedure are, however, known. In conclusion, this study contributes to demonstrate how a synergistic integration of chemical and biocatalytic approaches could be a winning strategy for the asymmetric synthesis of stereochemically dense products.
Author Contributions: Conceptualization, F.P. and P.P.G.; methodology, G.D.C. and A.F.; investigation, F.P. and V.C.; writing-original draft preparation, P.P.G.; writing-review and editing, C.T. and L.A.L.; funding acquisition, C.T., L.A.L. and O.B. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by University of Ferrara, 2020 call for funds 5 × 1000 year 2018.