Chemoenzymatic Synthesis of Tenofovir

We report on novel chemoenzymatic routes toward tenofovir using low-cost starting materials and commercial or homemade enzyme preparations as biocatalysts. The biocatalytic key step was accomplished either via stereoselective reduction using an alcohol dehydrogenase or via kinetic resolution using a lipase. By employing a suspension of immobilized lipase from Burkholderia cepacia (Amano PS-IM) in a mixture of vinyl acetate and toluene, the desired (R)-ester (99% ee) was obtained on a 500 mg scale (60 mM) in 47% yield. Alternatively, stereoselective reduction of 1-(6-chloro-9H-purin-9-yl) propan-2-one (84 mg, 100 mM) catalyzed by lyophilized E. coli cells harboring recombinant alcohol dehydrogenase (ADH) from Lactobacillus kefir (E. coli/Lk-ADH Prince) allowed one to reach quantitative conversion, 86% yield and excellent optical purity (>99% ee) of the corresponding (R)-alcohol. The key (R)-intermediate was transformed into tenofovir through “one-pot” aminolysis–hydrolysis of (R)-acetate in NH3-saturated methanol, alkylation of the resulting (R)-alcohol with tosylated diethyl(hydroxymethyl) phosphonate, and bromotrimethylsilane (TMSBr)-mediated cleavage of the formed phosphonate ester into the free phosphonic acid. The elaborated enzymatic strategy could be applicable in the asymmetric synthesis of tenofovir prodrug derivatives, including 5′-disoproxil fumarate (TDF, Viread) and 5′-alafenamide (TAF, Vemlidy). The molecular basis of the stereoselectivity of the employed ADHs was revealed by molecular docking studies.


INTRODUCTION
Nowadays, developing cost-efficient and sustainable methodologies for the synthesis of enantiomerically pure active pharmaceutical ingredients (APIs), which often determine production cost, accessibility, and drug quality (i.e., therapeutic efficacy and patient safety), is of prime concern for both academia and industry. 1 Moreover, chiral APIs are present in over 80% of drugs currently on the market, and therefore, their production in optically pure form is of paramount importance for target-oriented therapies. 2 One of the sustainable and transition-metal-free industrial methods, which increases cost efficiency and reduces waste in small-molecule API synthesis by shortening synthetic routes mainly via bypassing protecting group requirements, relies on biocatalysis. 3 Therefore, in a continuation of our interest in chemoenzymatic preparation of enantiomerically pure APIs, 4 we present novel biocatalytic routes toward (R)-9-(2-phosphonomethoxypropyl) adenine [(R)-PMPA], namely, tenofovir (TFV).
Tenofovir and its lipophilic prodrugs [i.e., tenofovir disoproxil fumarate (TDF) and tenofovir alafenamide (TAF)] (Figure 1) are powerful antiretroviral agents that already have gained the status of being "frontline drugs" for the treatment of human immunodeficiency virus (HIV) infection and chronic hepatitis B caused by HBV. 5 The anti-HIV activity of parent tenofovir, resulting from competitive inhibition of HIV reverse tran-scriptase with respect to dATP, is strictly related to the absolute configuration of its stereogenic center. In this context, the (R)configured title API is ca. 100-fold more active as a nucleoside reverse transcriptase inhibitor (NRTI) than its enantiomeric counterpart. 6 Therefore, the synthesis and administration of pharmacologically relevant TFV single enantiomer by HIVinfected patients are required for highly active antiretroviral therapy (HAART). 7 Most TFV manufacturing processes utilize adenine and (R)propylene carbonate as starting materials. 8 However, synthesis of (R)-propylene carbonate requires an asymmetric synthetic methodology, whereas N-alkylation of adenine with this chiral reagent generates ca. 10% of material loss due to undesired regioisomer formation. 9 Recently, to reduce the costs of the production of TFV, an alternative synthetic strategy utilizing 5amino-4-cyanoimidazole (HPI) as an intermediate was developed. 10 This method starts with hydrogen cyanide, which is highly toxic and requires a well-ventilated fume hood equipped with a HCN detector. In turn, despite significant efforts already being made toward developing chemoenzymatic synthesis of many antiviral agents and their synthetic intermediates, 11 tenofovir and its prodrugs lack a vast repertoire of such creative attempts. Until now, only two biocatalytic approaches have been reported and included a gram-scale Candida antarctica lipase B (CAL-B)-catalyzed kinetic resolution of racemic 1-(trityloxy)propan-2-ol 12 or enzymatic reduction of prochiral 1-(6-amino-9H-purin-9-yl) propan-2one by using the (R)-selective Codexis KRED P1B02. 13 Both methods suffer from severe limitations that render them not ideal for potential upscaling. Thus, the first protocol elongates a routine synthesis of tenofovir in about six synthetic steps. In contrast, the second approach requires commercial enzymes of unknown composition, which enforces dependency on a particular supplier.
2.1. Lipase-Catalyzed Kinetic Resolution of Racemic Alcohols rac-4a and -4b. First, the syntheses of both ketones 3a and 3b and the corresponding racemic alcohols rac-4a and -4b were performed in analogy to the methods already reported in the literature 15 (for details, see Supporting Information). In this regard, to obtain 1-(6-chloro-9H-purin-9-yl) propan-2-one (3a), alkylation of commercially available 6-chloropurine (1) with 1.1 equiv of chloroacetone in the presence of 1 equiv of anhydrous potassium carbonate (K 2 CO 3 ) suspended in dry dimethylformamide (DMF) was performed, furnishing the carbonyl product 3a in 58% yield. In turn, the synthesis of 1-(6iodo-9H-purin-9-yl) propan-2-one (3b) was accomplished in a 2-step reaction sequence by the treatment of 1 with an aqueous solution of hydrogen iodide, followed by K 2 CO 3 -mediated alkylation of the isolated 6-iodopurine (2) with chloroacetone in the same manner as for 1. In this case, the desired iodo-ketone 3b was synthesized in 32% total yield after two steps. The resulting ketones 3a and 3b were then chemically reduced using a suspension of 1.2 equiv of sodium borohydride (NaBH 4 ) in a mixture of methanol and acetonitrile (2:1 v/v), thus furnishing racemic alcohols rac-4a in 75% yield and rac-4b in 91% yield. Prior to developing the enzymatic kinetic resolution step, the syntheses of both racemic acetates rac-5a and -5b as analytical standards were performed using a conventional esterification protocol employing 1.5 equiv of acetic anhydride (Ac 2 O, in the case of rac-5a) or acetyl chloride (AcCl, in the case of rac-5b), 1.5 equiv of triethylamine (Et 3 N) as a base, and a catalytic amount of 4-(dimethylamino)pyridine (DMAP) in dry dichloromethane (CH 2 Cl 2 ). In this case, the desired rac-5a was isolated in 89% yield, whereas rac-5a was isolated in 79% yield.
In the next step, the enzymatic KR of rac-4a and -4b was investigated by using a set of commonly used commercially Conditions: rac-4a and -4b 0.12 mmol, lipase 5 mg (42 mg/mmol), MTBE 2 mL, vinyl acetate 1 mL (92 equiv), 40°C, 800 rpm (magnetic stirrer). b Conversion values (%) (i.e., consumption of substrate rac-4a and -4b) were determined by GC analyses after derivatization of crude mixture with BSA as a silylating reagent; for confirmation, the percent conversion was calculated from the enantiomeric excess of the unreacted alcohol (ee s ) and the formed acetate (ee p ) according to the formula conv. = ee s /(ee s + ee p ). c Determined by HPLC analyses using columns packed with chiral stationary phases. d Calculated according to Chen et al. 16 (Table 1). Lipasecatalyzed KRs of rac-4a and -4b were carried out for either 16 or 24 h.
The preliminary enzyme screening revealed that the best results in terms of the enantiomeric excess (ee) of the desired (R)-1-(6-halo-9H-purin-9-yl) propan-2-ols [(R)-(−)-4a and -4b] as well as the enantioselectivity (E) of the acetylation of rac-4a and -4b were obtained when the lipase from Burkholderia cepacia (BCL) immobilized on diatomaceous earth (Amano PS-IM) was employed as biocatalyst. This lipase preparation displayed E values of 348 in the case of 6-chloro derivative rac-4a and 214 in the case of 6-iodo derivative rac-4b, furnishing enantiomerically enriched acetates (R)-(−)-5a (97% ee) and (R)-(+)-5b (98% ee) with 51% and 42% conversion after 24 h (Table 1, entry 8 vs entry 17), respectively. Comparable optical purities were obtained with the native lipase from Pseudomonas fluorescens (Amano AK) (Table 1, entries 9 and 18), however with significantly lower conversion after 24 h. Hence, our attention was paid to Amano PS-IM. Furthermore, we decided to choose rac-4a for the subsequent optimization studies since the chloro substrate gave better results in terms of enantioselectivity and was obtained in one step less than iodo derivative rac-4b, thus simplifying the whole synthetic pathway toward tenofovir [(R)-(−) -11].
Once the suitable enzyme and substrate had been identified, optimization studies were continued to evaluate the effect of the organic cosolvents on Amano PS-IM as the rate and stereoselectivity of the enzymatic reactions as well as the thermal stability of the enzyme largely depend on the reaction medium 17 (Table 2). For this purpose, 11 organic solvents of varying polarity (log P from −0.31 to 2.52) were investigated, including water-miscible polar solvents, such as 1,4-dioxane, acetonitrile (CH 3 CN), and acetone, or nonpolar solvents, such as toluene (PhCH 3 ).
The results showed that the enzymatic reaction gave higher conversions of the substrate (45−47%) in MTBE and PhCH 3 than in other solvents (12−38%). In this regard, we observed the following rate trend: CH 3 CN ≈ acetone ≈ tert-amyl alcohol (2methyl-2-butanol) < 1,4-dioxane ≈ THF < EtOAc ≈ CH 2 Cl 2 < CHCl 3 < vinyl acetate ≪ MTBE ≈ PhCH 3 . In turn, another order was observed for enantioselectivity: 1,4-dioxane ≈ CH 3 CN ≈ acetone ≈ EtOAc ≈ THF ≈ MTBE ≈ CH 2 Cl 2 ≈ tert-amyl alcohol (E = 200−300) < CHCl 3 ≈ vinyl acetate (E = 300−400) ≪ PhCH 3 (E > 500). As can be noticed, no clear correlation between the log P value and the biocatalytic activity and the enantioselectivity was discerned. Considering the reaction rates and the values of the E factor, it was clear that the most promising results were obtained when Amano PS-IM was suspended in PhCH 3 , which is in line with the common rule that enzymes' catalytic activity, stability, and enantioselectivity are generally improved in water-immiscible solvents with log P ≥ 2 due to negligible enzyme distortion in such media, suitable hydration of the protein molecule, and preservation of its active conformation. 18 Therefore, toluene was chosen as a cosolvent for further optimization studies.
In the next step, the effect of the reaction time on the outcome of (Amano PS-IM)-catalyzed KR of rac-4a with vinyl acetate in PhCH 3 was evaluated (for details, see Table S3 in Supporting Information). The KR assays were arbitrarily terminated at periodic time intervals between 4 and 30 h. Following the time course revealed that the most optimal compromise between the values of percent conversion and percent ee for the desired enantiomer (R)-(−)-5a is evident when the enzymatic reactions were terminated after 24 h. Although the elongation of time to 30 h led to improved conversion, the enantiomeric excess for the isolated acetate (R)-(−)-5a slightly decreased from >99% to 99%.
The temperature was studied to shorten the reaction times of the KRs as well as to investigate whether the immobilized enzyme can retain its enantioselectivity and stability at elevated Conditions: rac-4a 25 mg (0.12 mmol), Amano PS-IM 5 mg (42 mg/mmol), solvent 2 mL, vinyl acetate 1 mL (92 equiv), 24 h at 40°C, 800 rpm (magnetic stirrer). b Logarithm of the partition coefficient of a given solvent between n-octanol and water calculated using ChemBioDraw Ultra 13.0 software. c Conversion values (%) (i.e., consumption of substrate rac-4a) were determined by GC analyses after derivatization of a crude mixture with BSA as a silylating reagent; for confirmation, the percent conversion was calculated from the enantiomeric excess of the unreacted alcohol (ee s ) and the formed acetate (ee p ) according to the formula conv. = ee s /(ee s + ee p ). d Determined by HPLC analyses using columns packed with a chiral stationary phases. e Calculated according to Chen et al. 16 The Journal of Organic Chemistry pubs.acs.org/joc Article temperatures (for details, see Table S4 in Supporting Information). The reactions were carried out at 40, 50, and 60°C for 8 and/or 24 h for each temperature. When the lipasecatalyzed KR of rac-4a was performed at temperatures > 50°C, the E values dropped to 226−371, which resulted in the isolation of the formed acetate (R)-(−)-5a in lower optical purities (96− 98% ee).
The integration of the results obtained from the optimization studies showed that the most efficient biocatalyst was Amano PS-IM when used at 42 mg/mmol of rac-4a (2.5 mg/mL) and suspended in a solution of rac-4a (0.06 mmol/mL) in a mixture of PhCH 3 /vinyl acetate (3 mL, 2:1 v/v) as the reaction medium at 40°C. In the next step, to evaluate the general applicability and versatility of the elaborated method on a preparative laboratory scale, enantioselective transesterification of rac-4a catalyzed by Amano PS-IM was scaled up to 500 mg of the racemic substrate (Table 3). Noteworthy, rac-4a was subjected to the optimized biocatalytic conditions with the preservation of the linear enlargement of all of the KR parameters, including substrate amount, enzyme concentration, volume of the acyl donor and solvent, etc. Arresting (Amano PS-IM)-catalyzed KR of rac-4a after 26 h (at 44% conversion) and purifying the crude reaction mixture using preparative silica-gel column chromatography afforded (S)-(−)-4a in 54% yield with 79% ee and (R)-(−)-5a in 47% yield with 99% ee (Table 3, entry 1).
Since we found that the KR products may conveniently be isolated by exploiting the solubility properties of the alcohol (S)-(−)-4a and acetate (R)-(−)-5a in the water−toluene biphasic Conversion values (%) (i.e., consumption of substrate rac-4a) were determined by GC analyses after derivatization of crude mixture with BSA as a silylating reagent; for confirmation, the percent conversion was calculated from the enantiomeric excess of the unreacted alcohol (ee s ) and the product (ee p ) according to the formula conv. = ee s /(ee s + ee p ). c Determined by HPLC analyses using columns packed with a chiral stationary phases. d Isolated yield after column chromatography or liquid−liquid extractive workup. e Chemical purity established using GC and/or HPLC methods. f Calculated according to Chen et al. 16   mixture at room temperature, the enzymatic reaction was performed on a larger scale to obtain sufficient amount for handling ( Table 3, entry 2). In this case, termination of KR of rac-4a after the same time as before furnished (S)-(−)-4a in 48% yield with 82% ee and (R)-(−)-5a in 31% yield with 99% ee. However, when employing liquid−liquid extractive (LLE) workup as a purification step, the yield of (R)-(−)-5a decreased due to a partial loss of the acetate, which was trapped in the water phase. Consequently, the recovered optically active alcohol (S)-(−)-4a contained 8% of acetate (R)-(−)-5a. The absolute configuration of the KR products was determined by comparison with the reported optical rotations (for details, see Table S2 in Supporting Information).

E. coli/ADHs-Catalyzed
Bioreduction of Ketones 3a and 3b. As the kinetic resolution methodology is limited by 50% reaction yields for the desired enantiomer and a cumbersome purification step, an alternative stereoselective transfer−biohydrogenation of prochiral ketones 3a and 3b was investigated to obtain (R)-(−)-4a and -4b in a more efficient manner (Table 4).
Prior to alcohol dehydrogenase-catalyzed bioreduction studies, silylation of the polar hydroxyl groups in alcohols rac-4a and -4b using the standard derivatization protocol employing N,O-bis(trimethylsilyl) acetamide (BSA) in CH 2 Cl 2 had to be applied to obtain more volatile compounds characterized by much lower values of the retention times (t R ) than ketones 3a and 3b. As we needed trimethylsilyl ethers rac-6a and -6b in more significant amounts to proceed with calibration curves for GC analyses, the preparative scale for the silylation reaction was also performed, furnishing both derivatives in the range of 83− 87% yield (after column chromatography) (Scheme 1).
An initial biocatalyst screening was attempted using lyophilized Escherichia coli cells harboring recombinantly overexpressed alcohol dehydrogenases (denoted as E. coli/ ADHs) originating from various microorganisms, including The E. coli/ADH-catalyzed bioreductions of both prochiral ketones 3a and 3b were carried out under standard biocatalytic conditions using a 10 mM concentration of the respective substrate and 10 mg of E. coli cells of each biocatalyst suspended in 50 mM Tris-HCl buffer (pH 7.5) in the presence of 1.0 mM of the external nicotinamide cofactor [NADH or its phosphate counterpart NADPH] for 24 h at 30°C. In addition, to ensure complete solubility of the organic substrates 3a and 3b in the reaction media and enhance the in situ regeneration of the NAD(P)H cofactors, we have supplemented the reaction mixtures with 2.5% v/v of dimethyl sulfoxide (DMSO) and 10% v/v of propan-2-ol (2-PrOH), respectively. Notably, due to the high solubility of the products 4a and 4b in an aqueous medium, all of the enzymatic reactions were terminated by removal of the cells under suction and subsequent azeotropic evaporation of the water with toluene from permeate to eliminate an undesired drop in the product yield (for details, see Supporting Information).
To our delight, the biocatalytic reduction of 3a and 3b revealed that from the tested biocatalysts, those E. coli cells which contained the ADH variants derived from L. kef ir (E. coli/ Lk-ADH-Lica and E. coli/Lk-ADH Prince) allowed obtaining the desired optically pure (R)-(−)-4a and -4b (>99% ee) with >99% conversion. In turn, E. coli/ADH-A led to the formation of enantiomerically pure antipodes (S)-(+)-4a and 4b (>99% ee) with quantitative conversions (>99%). Noteworthy, due to the opposite stereopreference of the examined whole-cell biocatalysts toward prochiral ketones 3a and 3b, both enantiomers of a key intermediate could be synthesized in simultaneous attempts.
After successfully screening the ADH biocatalysts, our next task was to scale up the asymmetric bioreduction of ketones 3a and 3b utilizing the stereocomplementary ADH preparations exhibiting either the Prelog selectivity (E. coli/ADH-A) or the anti-Prelog selectivity (E. coli/Lk-ADH Prince) (Scheme 2). Taking advantage of the chloro derivative 3a over iodo derivative 3b (i.e., the higher yield and simplicity of its synthesis due to a lesser number of steps and commercial availability of the reagents), we decided to use 3a as the model substrate exclusively. The biotransformations were performed with 0.4 mmol of 3a in 4 mL of the final volume of the reaction mixture, which resulted in a 10-fold higher substrate concentration (100 mM) than on an analytical scale. By utilizing E. coli/ADH-A, the asymmetric bioreduction of 3a furnished (S)-(+)-4a in 70% yield and with >99% ee. In contrast, when E. coli/Lk-ADH Prince was used as a biocatalyst, (R)-(−)-4a was isolated in 86% yield and with >99% ee. Once again, to avoid drawbacks with the high solubility of the products (R)-(−)-4a and (S)-(+)-4a in an aqueous buffer solution, we employed an alternative workup procedure, which excluded direct liquid−liquid extraction of the Scheme 2. Stereocomplementary Bioreductions of 3a Catalyzed by E. coli/ADH-A or E. coli/Lk-ADH Prince The Journal of Organic Chemistry pubs.acs.org/joc Article water phase with organic solvents. This modification not only allowed us to obtain the appropriate products (R)-(−)-4a and (S)-(+)-4a in higher yields but also generated a more sustainable protocol for their isolation by decreasing the amount of the used solvents during the workup. It was clear that the elaborated biocatalytic methodology can be amplified to a larger scale and thus outperformed the abovepresented lipase-catalyzed KR of rac-4a and -4b (in terms of the product yield) as well as other literature reports 14 (in terms of the enantiomeric purity of (R)-(−)-4a), requiring for asymmetric transfer hydrogenation of prochiral ketones 3a and 3b ruthenium catalysts. In this regard, ADH-catalyzed transformation, utilizing more benign catalysts and sustainable reaction conditions, might be relevant for industry, paving the way for more green technologies.

Molecular Docking.
In order to rationalize the stereoselectivity of the ADH-catalyzed biotranshydrogenation of 1-(6-chloro-9H-purin-9-yl) propan-2-one (3a), comprehensive in silico enzyme−substrate docking calculations were performed. For this purpose, prochiral ketone 3a was docked with receptor molecules prepared based on the crystal structures of both stereocomplementary ADHs, namely, ADH-A from Rhodococcus ruber DSM 44541 (PDB code: 2XAA) 29 and Lk-ADH from L. kef ir (PDB code: 4RF2), 30 which were retrieved from the Protein Data Bank (PDB) database (http://www.rcsb. org/pdb/). As a result of molecular docking experiments, nine of the most energetically favorable binding modes for the ligand− protein complexes for ligand 3a molecule and target ADH-A and Lk-ADH proteins were generated. The visualization of the representative docking poses of 3a to ADH-A and Lk-ADH (taken from the lowest energy docking clusters) with close contacts to amino acid residues located in the corresponding active sites of the studied ADHs is presented in Figure 2.
Careful inspection of the productive pose of 3a in ADH-A (Figure 2A and 2B) showed that this ketone forms strong 2.5 Å long metal−acceptor interactions with the catalytic zinc ion and a 3.0 Å long hydrogen bond with the hydroxyl group of the side chain of the Ser40 residue present in the active site. At the same time, the 6-chloro-9H-purine moiety of 3a is located outside the small substrate-binding pocket, which allows ligand 3a to avoid unfavorable steric clashes of the larger substituent with amino acid residues located more profoundly in the catalytic cavity (i.e., Tyr294, Trp295, Asp153, His62, Cys38, Ser40, etc.). In addition, the purine moiety is bound in the broader part of Figure 2. Representative three-dimensional (3D) binding modes of 1-(6-chloro-9H-purin-9-yl) propan-2-one (3a) with stereocomplementary alcohol dehydrogenases, namely, ADH-A (PDB code: 2XAA; A and B) and Lk-ADH (PDB code: 4RF2; C and D), with close contacts to amino acid residues and cofactors located in the active sites. The docked ligand 3a and the cofactors are shown as sticks representation, where 3a is white, NADH is yellow, and NADPH is violet. The overall receptor structures are shown as a semitransparent cartoon diagram (left; A and C), where ADH-A is wheat and Lk-ADH is gray. The most significant amino acid residues contributing to the stabilization of the ligand 3a molecule in the complex with ADH-A or Lk-ADH by polar interactions, alkyl−alkyl (CH−CH), van der Waals (vdW,) and/or π−alkyl (CH−π) interactions are shown in lines representations. Nitrogen atoms are shown in blue, oxygen atoms in red, chlorine atoms in green, and phosphorus atoms in orange. All hydrogens were omitted for clarity. The zinc ion is presented as a semitransparent slate sphere. The formation of intermolecular hydrogen bonds is represented by magenta dashed lines, whereas the plausible trajectory of the hydride transfer from cofactors to a carbon atom of the carbonyl group is shown as red dashed lines. Mutual distances between the amino acid residues and the respective ligand's atoms are given in Ångstroms (right column; B and D). The figure was prepared using the program PyMOL (http://www.pymol.org/). The Journal of Organic Chemistry pubs.acs.org/joc Article the active site cavity pointing toward the exterior amino acids (i.e., Phe43, Phe282, Ile271) within which 3a forms π−π and π− alkyl (CH−π) interactions (see Figure S3A in Supporting Information). The second plausible explanation of the ligand accommodation is that the purine moiety introduces a steric hindrance due to the proximity of the catalytic zinc, which consequently pushes the substrate out of this region. Of note, the additional π−σ interactions between the methyl substituent of 3a and the indole ring of Trp295 stabilize a hypothetical complex of 3a−ADH-A such that the carbonyl oxygen atom of the substrate 3a forces the location of the carbon atom of the carbonyl group to be in close proximity to the C4 atom of the NADH cofactor. As a consequence of this orientation, the hydride is transferred from the cofactor onto the re face of the carbonyl group of 3a following the well-known (S)-stereoselectivity of ADH-A. In turn, a complex of Lk-ADH and the selected top-scoring binding mode of 3a ( Figure 2C and 2D) confirmed that the ligand molecule is accommodated inside the binding cleft with a pose that promotes the (R)-stereoselective ketone reduction pathway. When analyzing in silico simulations, one can see that the confined environment of the Lk-ADH active site orients the 6-chloro-9H-purine moiety perpendicular to the cofactor, forcing the methyl group to locate into the smaller binding pocket in close proximity toward the catalytic triad (Ser143-Tyr156-Lys160). In addition, two strong 2.3−2.9 Å long hydrogen bonds formed between the oxygen atom of the prochiral carbonyl group of 3a and the amino acid residues involved in the proton relay (Ser143 and Tyr156) favored the orientation of the ligand with the si face exposed toward the reactive hydride of the NADPH cofactor. Moreover, the docking of 3a with Lk-ADH revealed that the imidazole ring of the purine moiety formed H-bond interactions with the phenol group of Tyr190. On the other hand, this functionality also establishes π− anion interactions with Glu145. Interestingly, the additional π− alkyl interactions between the purine ring and the side chains of the aliphatic amino acid residues (i.e., Ala202, Val196, Leu153) were found to be beneficial for the stabilization of the ligand− protein complex in such a way that anti-Prelog product (R)-(−)-4a could be obtained within this biocatalyst (see Figure S3B in Supporting  The primary goal of these studies was to develop synthetic procedures that would lead to (R)-(−)-11 in high total yield and excellent enantiomeric excess. Besides the motivations mentioned above, the assumption of employing the shortest possible synthetic pathway leading to less waste generation was also considered.
In a subsequent attempt, we performed Mg(O t Bu) 2 -mediated O-alkylation of the resulting (R)-(−)-7 with 1.5 equiv of tosylated diethyl (hydroxymethyl) phosphonate (9) in analogy to the method reported by Barral et al. 31 The reaction was carried out in dry dimethylformamide (DMF) for 24 h at 75°C and led to obtaining the respective diethyl phosphonate ester (R)-(−)-10 in 88% yield and with 99% ee. In the original procedure, the authors used room temperature and sodium base NaO t Bu instead of the magnesium one Mg(O t Bu) 2 and obtained (R)-(−)-10 in only a 29% yield. Our result proved the putative role of the Mg 2+ counterion as well as an elevated temperature on the outcome of the alkylation reaction.
The final deprotection of phosphonate ester (R)-(−)-10 was performed according to the procedure reported by Maghami et al. 15a In this regard, the treatment of (R)-(−)-10 with trimethylsilyl bromide (TMSBr) in dry CH 2 Cl 2 for 48 h at 0− 5°C (performed in a pressure tube) allowed us to afford free phosphonic acid (R)-(−)-11 in 91% yield and without loss of optical purity (99% ee).
Taking into account both chemoenzymatic routes, the resulting tenofovir [(R)-(−)-11] was synthesized in 13.3% total yield when following the lipase-catalyzed KR methodology (6 steps) and in 35.9% total yield when following the E. coli/ ADH-catalyzed bioreduction (5 steps) pathway. Compared with the reported methods, 8−12,14 our approach is more straightforward and efficient in terms of the synthesis of optically active

CONCLUSIONS
Here, we report straightforward chemoenzymatic synthesis routes for the facile preparation of tenofovir in an optically pure form, starting from simple and inexpensive commercially available reagents as well as biodegradable enzyme preparations. The elaborated methodology involves a highly stereoselective bioreduction of prochiral 1-(6-chloro-9H-purin-9-yl) propan-2one catalyzed by a (R)-specific ADH variant from L. kef ir (E. coli/Lk-ADH Prince) and/or enantioselective transesterification of racemic 1-(6-chloro-9H-purin-9-yl) propan-2-ol with vinyl acetate in the presence of immobilized lipase from B. cepacia (Amano PS-IM) carried out under kinetically controlled conditions. The ADH-based methodology furnished (R)-1-(6chloro-9H-purin-9-yl) propan-2-ol in high isolated yield (86%) and excellent enantiomeric purity (>99% ee). In contrast, when following the lipase-catalyzed KR strategy, the desired optically active (R)-precursor (99% ee) was obtained in 47% yield after preparative chromatography or in 31% yield after liquid−liquid extractive workup without employing chromatographic purification. The key biocatalytic steps were combined with a convenient "aminolysis−hydrolysis−alkylation−deprotection" reaction sequence to convert the enzymatically generated (R)intermediate into the desired tenofovir active agent in good overall isolated yield (up to 72%) and excellent optical purity (99% ee) after 3 steps. Comparing both investigated chemoenzymatic routes, the one which utilized E. coli/ADH-catalyzed bioreduction of 1-(6-chloro-9H-purin-9-yl) propan-2-one turned out to be more efficient since tenofovir has been synthesized in 35.9% total yield after 5 steps. This approach offers an attractive alternative to the already reported catalytic methods and can potentially serve in tenofovir manufacturing, as well as in the synthesis of other pharmaceutically relevant acyclic nucleoside phosphonates.
4.1.1.2. Method B (Ended with Purification via Liquid−Liquid Extraction). All of the manipulations, except the workup and purification procedures, were carried out by analogy with the above protocol. To isolate the EKR products, the crude reaction mixture was filtered off from the lipase preparation, and the filtrate cake was rinsed with a portion of PhCH 3 (20 mL). After evaporation of the volatiles, the crude oil was dissolved in PhCH 3 (40 mL) and washed with H 2 O (3 × 20 mL). The combined organic layer was dried over MgSO 4 , the drying agent was filtered off, and the permeate was concentrated under reduced pressure to provide (R)-(−)-5a (187 mg, 0.7 mmol, 31% yield, 99% ee, >99% purity) as a white solid. The combined aqueous phase was back-extracted with PhCH 3 (3 × 40 mL) to remove traces of (R)-(−)-5a. Afterward, an aqueous layer was azeotropically condensed with PhCH 3 (100 mL). The resulting oil residue was diluted with AcOEt (40 mL) and dried over MgSO 4 , and after filtering off the drying agent and evaporation of the volatile solvents, the desired (S)-(+)-4a (240 mg, 1.1 mmol, 48% yield, 82% ee, 92% purity) was obtained as a white solid. For details, see Table 3.

■ ASSOCIATED CONTENT Data Availability Statement
The data underlying this study are available in the published article and its Supporting Information.