Chemoenzymatic Synthesis of Apremilast: A Study Using Ketoreductases and Lipases

The key step in the chemoenzymatic synthesis of apremilast was to produce the chiral alcohol (R)-1-(3-ethoxy-4-methoxyphenyl)-2-(methylsulfonyl)ethanol, (R)-3. Two enzymatic approaches were evaluated to obtain (R)-3, one using ketoreductases and the other lipases. Bioreduction of 1-(3-ethoxy-4-methoxyphenyl)-2-(methylsulfonyl)ethanone (2), using ketoreductase KRED-P2-D12, led to (R)-3 with 48% conversion and 93% enantiomeric excess (ee). Kinetic resolution of rac-1-(3-ethoxy-4-methoxyphenyl)-2-(methylsulfonyl)ethyl acetate (rac-4), via hydrolysis reaction, with 20% of n-butanol, catalyzed by lipase from Aspergillus niger yielded (R)-3 with > 99% ee, 50% conversion and E-value (enantiomeric ratio) > 200. The reaction between enantiomerically pure (R)-3 and 4-acetylamino-isoindol-1,3-dione (8) afforded apremilast in 65% yield and 67% ee.


Chemoenzymatic Synthesis of Apremilast: A Study Using Ketoreductases and Lipases
The latter was obtained with 99% yield and 96% ee, through asymmetric hydrogenation of N-(1-(3-ethoxy-4-methoxyphenyl)-2-(methylthio)vinyl)acetamide, in the presence of the rhodium catalyst Rh(NBD) 4 BF 4 NBD: norbornadiene, and the ligand (S c ,R p )-DuanPhos. After that, the chiral β-acetylamino sulfide was oxidized, in the presence of 30% H 2 O 2 and TaCl 5 catalyst, to produce the corresponding β-acetylamino sulfone with 50% yield and 94% ee. 12 More recently, a new approach was reported 13 for the synthesis of apremilast, which consisted of the reaction between 4-amino-2-(2-methylsulfonyl)vinyl)isoindoline-1,3-dione and (3-ethoxy-4-methoxyphenyl)boronic acid, using the rhodium catalyst [RhCl(C 2 H 4 ) 2 ] 2 and a chiral ligand, (1S,4S)-1,7,7-trimethyl-2,5-diphenylbicyclo[2.2.1] hepta-2,5-diene or (1S,4S)-2,5-diphenylbicyclo[2.2.1] hepta-2,5-diene. After acetylation of the amino group attached to the benzenic ring, apremilast was obtained with 94% yield and 97% ee. 13 Biocatalysis is a powerful alternative tool to mainstream chemical synthesis to produce chiral drugs in enantioriched/enantiopure forms. This methodology has evolved to industrial application due to the ease in complementing conventional methods of synthesis of active pharmaceutical intermediates (APIs). The differential of a biocatalytic processes lies in the fact that an enzyme is obtained from renewable sources and it is biodegradable and non-hazardous, therefore being environmentally friendly. 14 Conversely, the use of conventional synthetic methodologies is faced with the need to remove traces of metal catalysts such as platinum, rhodium, ruthenium, and palladium in the final stages of the process, increasing the cost of producing APIs. Biocatalytic processes can be economically more attractive, as they are generally conducted under mild conditions that include atmospheric pressure, room temperature and aqueous reaction medium, reducing waste generation. 14 Moreover, the advent of direct protein evolution has enabled to tailor an enzyme to accept a specific unnatural substrate, improving catalytic performance, thermostability and stereospecificity. 15 Consequently, biocatalysis is one of the main technologies endorsed by the pharmaceutical industry to produce APIs, a fact that can be proven by the large number of patents that have been granted over the last five years. Currently, approximately 300 biocatalytic processes are operating on an industrial scale. [15][16][17][18][19][20][21][22] Ketoreductases (KREDs), also called alcohol dehydrogenases (ADHs) or carbonyl reductases (CRs), together with lipases, have been the most used biocatalysts at industrial scale. 17 It is noteworthy that KREDs represent more than a quarter of the commercially available enzymes. This class of enzymes requires a cofactor, such as nicotinamide adenine dinucleotide (NADH) or its phosphorylated form (NADPH), together with a regeneration system. 23 These enzymes have been widely used to produce chiral secondary alcohols, an important building block in the pharmaceutical industry to produce APIs and in fine chemistry. [24][25][26][27] Lipases stand out among the enzymes used to produce APIs, since they can be used without the need of cofactor addition, besides acting on a wide range of substrates. Both factors, high abundance in nature and ease of genetic manipulation, contributed to the existence of a reasonable set of commercially available lipases. Finally, these enzymes are chemo-, regio-and enantioselective, and are generally active in organic solvent, constituting an advantage for solubilizing the organic substrate to be modified. [28][29][30] Herein, we report the study on the chemoenzymatic synthesis of apremilast using as key chiral intermediate a (R)-sulfonylethanol, Figure 1. Attempts to produce the intermediate (R)-sulfonylethanol involved the use of commercial KREDS and lipases from a sulfonylketone and rac-sulfonylethanol acetate, respectively, Figure 1. All investigated approaches focused, especially, on the optimization of the reactional conditions.

Bioreduction of ketone 2 catalyzed by KRED-P2-D12
KRED-P2-D12 (2 mg) and ketone 2 (1.3 mg) were added in a 2 mL Eppendorf. Then, 25 μL of ethanol and 400 μL of a stock solution of KRED Recycle Mix P (composed of 125 mmol L -1 of potassium phosphate, 1.25 mmol L -1 of MgSO 4 , 1 mmol L -1 of NADP + , 80 mmol L -1 , pH 7.0 from 0.15 g of Mix P dispersed in 5 mL of Milli-Q water) were added. Subsequently, 75 μL of IPA was added and the system was stirred at 180 rpm on the orbital shaker at 30 °C for 24 h. The reaction was quenched with the addition of 1 mL of EtOAc, followed by centrifugation for 5 min at 4500 rpm. Soon after, the organic phase was treated with anhydrous Na 2 SO 4 and after filtration, the sample was analyzed in GC-MS and HPLC. After analysis, the formation of (R)-3 was confirmed with a conversion of 48% and 93% ee.
Synthesis of (R)-3 and (S)-4 via kinetic resolution of rac-4 catalyzed by lipase from Aspergillus niger A solution of acetate rac-4 (72.8 mg, 0.229 mmol) in 2.3 mL of 0.1 M phosphate buffer pH 7/n-butanol (8:2) (v/v) was prepared. Subsequently, lipase from Aspergillus niger (3:1 m/m) was added to the solution and the reaction mixture was stirred at 250 rpm in the orbital shaker at 45 °C for 6 h. After this period, 3 mL of EtOAc was added to the reaction mixture, followed by centrifugation for 5 min at 1000 rpm. Then, the organic phase was treated with anhydrous Na 2 SO 4 . After filtration, the solvent was evaporated under reduced pressure and the crude product was purified by chromatographic column with flash silica gel using CH 2 Cl 2 :MeOH (9.7:0.3) as eluent, to give 32 mg of (R)-3 ( Synthesis of 4-nitroisoindoline-1,3-dione (6) In a 5 mL flask, 4-nitroisobenzofuran-1,3-dione (5) (1.55 mmol, 300 mg) and 540 μL ammonia hydroxide were added. Then, a reflux condenser was adapted to the system (without water flow) and the reaction system was heated to the boiling temperature until all the water had evaporated. Thereafter, the temperature was increased to 230 °C and heating was continued for 2 h. After this time, a yellow solid, identified as 4-nitroisoindoline-1,3-dione (6) (278 mg, 93%), was obtained with Rf = 0.61 (hexane: EtOAc (1:1)) and mp 213-215 °C.

Biocatalytic approach to produce the chiral (R)-sulfonylethanol intermediate (R)-3: via KREDs and lipases
Initially, the ketone (2) was obtained from the commercially available 3-ethoxy-4-methoxybenzonitrile (1) in two steps. The first step consisted of the reaction of 1 with the dimethylsulfone carbanium formed in the presence of n-butyllithium, yielding the corresponding enamine. This latter was treated one-pot with a 2.5 M HCl solution to produce the ketone (2) in 85% yield (Figure 2). Two alternative routes were investigated to produce (R)-3, one involving the reduction of ketone 2 catalyzed by KREDs (Figure 2, path A) and other by lipase-mediated hydrolysis of the rac-sulfonylethanol acetate (rac-4), Figure 2, path B.

Reduction of ketone 2 catalyzed by KREDs
Initially, ketone 2 was subjected to chemical reduction in the presence of sodium borohydride in methanol and dichloromethane, leading to hydroxysulfone rac-3 in 98% yield (Figure 2, path B). Then, a method using GC-MS was developed to separate ketone 2 from hydroxysulfone Vol. 32, No. 5, 2021 rac-3, aiming to determine the conversion in the reduction reaction catalyzed by KREDs.
A screening was performed with the 24 ketoreductases from Codexis ® (Codex KRED Screening Kit) using previously reported conditions. 32 Ketone 2 was subjected to a reduction reaction at 30 °C, 24 h, 180 rpm in the presence of a KRED with its respective cofactor recycling system and using DMSO as co-solvent (85.7% of KRED Recycle Mix, 9.5% of IPA and 4.8% of DMSO; v:v:v). In this case, DMSO was used to solubilize ketone 2 in the reaction medium. After 24 h of reaction, the conversion value was determined by GC-MS analysis. Among the 24 KREDs assessed, only 8 were active. The results are summarized in Table 1. The table containing all KREDs assessed for ketone 2 reduction is found in Supplementary Information (SI) section.
Among the eight KREDs active in reducing ketone 2, only two showed anti-Prelog selectivity, leading to hydroxysulfone (R)-3 ( Table 1, entries 1 and 2). Both KRED-P2-D03 and KRED-P2-D12 produced (R)-3 with values of ee > 90%, the latter leading to a higher conversion value (25%). Due to this, KRED-P2-D12 was chosen for further tests with the aim of increasing the conversion of bioreduction reaction of ketone 2. It is known that one of the advantages of the KRED-P2-D12 is the tolerance to a high amount of IPA used in the recycling of the cofactor, in the same way as the nineteen first KREDs listed in Table S1 (SI section). 33 The reduction of the carbonyl group takes place under thermodynamic control and, generally, requires a large excess of IPA to favor high conversion values. 34 In this way, we decided to evaluate the conversion behavior in the bioreduction of ketone 2, progressively increasing the concentration of IPA in the reaction medium with a concomitant decrease in the Recycle Mix and keeping the concentration of DMSO in 4.8%. The results are summarized in Table S2 (SI section).  With the increase in IPA up to 14.3%, there was an increase in conversion from 25 to 33%, but concentrations above this value led to a decrease in conversion. This behavior is in line with results published in the literature. 35 In addition to the DMSO, eleven more organic co-solvents were evaluated, which can modify the three-dimensional structure of an enzyme and its catalytic activity. 35,36 In this case, ethanol was the most efficient co-solvent, increasing the conversion from 33 (DMSO) to 48%, acting as a cosolvent, as well as a regenerating cofactor. 37 Once the conditions for the reduction of ketone 2 in the presence of KRED-P2-D12 were optimized, we turned our attention to the biocatalytic approach consisted of obtaining alcohol (R)-3 via kinetic resolution of acetate rac-4 by hydrolysis reaction catalyzed by lipases (Figure 2, path B).

Kinetic resolution of rac-1-(3-ethoxy-4-methoxyphenyl)-2-(methylsulfonyl)ethyl acetate (rac-4) via lipase-mediated hydrolysis
First, the previously prepared hydroxysulfone rac-3 was subjected to acetylation reaction in the presence of acetic anhydride, DMAP and triethylamine in dichloromethane, providing acetate rac-4 in 85% yield. Chiral HPLC methods were developed for both rac-3 and rac-4 to reliably measure the enantiomeric excesses of both the remaining substrate and the final product in the lipase-catalyzed resolution of rac-4.
Fifteen commercially available lipases (listed in "Enzymes" sub-section) were evaluated in the kinetic resolution of rac-1-(3-ethoxy-4-methoxyphenyl)-2-(methylsulfonyl)ethyl acetate (rac-4) via hydrolysis reaction, Figure 2 (path B). The reactions were conducted using conditions previously reported by our research group, 38 such as 1:2 ratio (m:m) of substrate/lipase, phosphate buffer 0.1 M (pH 7.0), acetonitrile as co-solvent (buffer/co-solvent 8:2, v:v), 30 °C, 250 rpm and 24 h. In these conditions, only the lipase from Aspergillus niger was active, leading to alcohol (R)-3 with 93% ee and acetate (S)-4 with 98% ee, E-value (enantiomeric ratio) of 127 and 51% conversion (Figure 2, path B). Although we obtained promising results in the production of (R)-3, we decided to evaluate the behavior of this kinetic resolution in the absence of a co-solvent and in the presence of several other co-solvents. The results are summarized in Table 2.
Notably, the kinetic resolution of rac-4 in the absence of co-solvent, in just 13 h of reaction, reached the maximum conversion value (ca. 50%) albeit with a low enantioselectivity value ( Table 2, entry 1). It is known that water participates directly or indirectly in noncovalent interactions to maintain the conformation of an active enzyme. 39 In a totally aqueous microenvironment, the conformational change of the active site is highly flexible, 40,41 which could provide a favorable adjustment of the two enantiomers of rac-4, leading to low enantioselectivity. The kinetic resolution of rac-4 in the presence of the most varied co-solvents, polar or nonpolar (with different log P and dielectric constant),  reached conversion values close to 50% in reaction times that varied from 9 to 24 h with E-values ranging from 15 to > 200 (Table 2). Although there is no general rule that indicates the physical/chemical properties of a co-solvent that act in the alteration of enzyme activity and enantiomeric ratio, the most evaluated are log P and dielectric constant. [42][43][44] In fact, two distinct enantiomeric ratio values were obtained in co-solvents with log P < 0, since in acetonitrile the E-value was 127 and in 1,4-dioxane it was 60 ( Table 2, entries 2 and 3). This difference in enantiomeric ratio could be attributed to the ability of acetonitrile to act in the stabilization of enzyme charges, since it presents a higher dielectric constant value (38) in relation to the 1,4-dioxane dielectric constant (2.21). Among co-solvents with log P > 0, the only one that provided a high E-value (E > 200) was n-butanol ( Table 2, entry 6). Apparently, n-butanol would be the co-solvent that presents an ideal balance between the values of log P (0.88) and dielectric constant (18), enabling a high enantiomeric ratio in the performance of lipase from A. niger. Isopropanol has a dielectric constant value (18.3) like n-butanol (18), but a much lower log P (0.05) ( Table 2, entries 4 and 6, respectively). Although the analysis of log P and dielectric constant are not sufficient to fully explain the behavior of co-solvents in the kinetic resolution of rac-4, it is evident that both acetonitrile ( Table 2, entry 2) and n-butanol ( Table 2, entry 6) promote a change in the microenvironment of the catalytic site of lipase from A. niger, inducing an ideal conformation for high enantiomeric ratio, especially in the case of n-butanol which provided an E-value > 200. In a few more attempts to optimize the kinetic resolution of rac-4, we decided to increase the temperature from 30 to 45 and 50 °C ( Table 2, entries 7 and 8). Actually, at a higher temperature, the resolution of rac-4 occurred in a shorter reaction time and decreased from 13 h at 30 °C to 9 h at 45 °C. Surprisingly, at the same time there was an increase in the ee of the product that went from 96% (30 °C) to > 99% (45 °C). However, at 50 °C it was possible to observe a deactivation of lipase from A. niger, since in 7 h of reaction the conversion was only 3% ( Table 2, entry 8). A similar behavior was observed in the kinetic resolution of rac-phenylethylamines catalyzed by lipase from A. niger, whose maximum temperature for E-values > 200 was 45 °C. 45 In a glance analysis, it is surprising to see an increase in enzymatic enantioselectivity with increasing temperature, since in most situations the opposite occurs, enantioselectivity decreases with increasing temperature. However, there are several reports in the literature [46][47][48][49][50][51][52][53][54][55][56] in which enzymatic enantioselectivity increases with increasing temperature. In these cases, the enantiomer is favored by entropy and the degree of flexibility at the active site of the enzyme is more evident than electrostatic and stereostatic effects. 47,49,[54][55][56] Since n-butanol showed the best performance as a co-solvent in the kinetic resolution of rac-4, we decided to investigate the influence of the amount of this co-solvent in the reaction medium, since preliminary studies were carried out in a buffer:n-butanol ratio of 8:2 (v:v). In the 9:1 ratio ee of (R)-3 had a slight drop (98.8%) and remained at > 99% in a 7:3 ratio, leading to the conclusion that 8:2 is the ideal buffer:n-butanol ratio ( Table 2, entries 9 and 10, respectively).
In this way, the addition of an ideal amount of n-butanol to the aqueous reaction medium and the increase in temperature from 30 to 45 °C provided the achievement of (R)-3 in enantiomerically pure form. It is known that in a totally aqueous medium, lipase exhibits a high flexibility 40,41 and this was probably the dominate factor for the lack of enantioselectivity in the kinetic resolution of rac-4 in the absence of co-solvents ( Table 2, entry 1). In the totally aqueous medium, the high flexibility of lipase from A. niger was responsible for accommodating the rac-4 enantiomers in a similar way, leading to low enantioselectivity. The addition of a co-solvent to the aqueous reaction medium with a positive log P such as n-butanol was responsible for dramatically increasing the enantiomeric ratio (E), which went from a value of 4 to > 200 ( Table 2, entries 1 and 6, respectively). The addition of n-butanol must have caused a reduction in the flexibility of the lipase, 57 altering the conformation at the active site and providing a more favorable accommodation for the (R)-4 in detriment to its respective enantiomer, leading to a high enantioselectivity. Although the addition of n-butanol was responsible for dramatically increasing enantioselectivity, obtaining (R)-3 in enantiomerically pure form was only possible when, at the same time, the temperature was increased from 30 to 45 °C (Table 2, entry 7). In this case, the high enantioselectivity must have been caused by two distinct effects, one related to the decrease in the polarity of n-butanol at a higher temperature (e (30 °C) 17.6 and e (45 °C) 14.7), which should cause an increase in the conformational rigidity of the enzyme (as observed in preliminary inhouse circular dichroism data, not shown) and the other related to an increase in the conformational flexibility of the enzyme by the temperature increase. These two effects act in the opposite way, but the balance between them is probably responsible for the ideal enantioselectivity observed in the kinetic resolution of rac-4.
Finally, we investigate the influence of the enzyme loading on the kinetic resolution of rac-4. With the increase in the enzyme:substrate ratio from 2:1 to 3:1, there was a decrease in reaction time from 9 to 6 h ( Table 2, entry 11). On the other hand, with the decrease in the enzyme:substrate ratio from 2:1 to 1.5:1, there was an increase in reaction time to 23 h ( Table 2, entry 12). Therefore, the ideal rate of enzymatic loading enzyme:substrate (m:m) for the kinetic resolution of rac-4 to occur in the shortest possible reaction time is 3:1 ( Table 2, entry 11).
Once our objective of obtaining (R)-3 in enantiomerically pure form has been achieved, we focus our attention on the synthesis of apremilast.

Synthesis of apremilast
To perform the synthesis of apremilast, it was necessary to obtain acetylaminophthalimide 8, starting from commercially available nitro anhydride 5 (Figure 3).
The latter was converted to the corresponding phthalimide 6, in the presence of ammonium hydroxide and heated to 230 °C, with 93% yield. Subsequently, nitrophthalimide 6 was reduced to aminophthalimide 7 in the presence of Pt as a catalyst, with 96% yield. In the last step, aminophthalimide 7 underwent an acetylation reaction, leading to the achievement of acetylaminophthalimide 8 with 54% yield.
Finally, apremilast could be obtained via a Mitsunobu reaction between enantiomerically pure alcohol (R)-3 and acetylaminophthalimide 8. When the reaction was carried out using THF as solvent at -5 °C, apremilast was obtained with 65% yield and 51% ee. In this case, a degree of racemization was observed during the reaction process, suggesting the reaction occurred via an S N 1-type mechanism. Even so, a moderate value of enantiomeric excess was probably assured due to the formation of an intimate ion pair between the phosphonium intermediate with ortho-and para-alkoxybenzyl alcohol with a carbocation character. To enhance the formation of intimate ion pairs in the reaction medium, leading to an increase in the ee of apremilast, the reaction was repeated with the addition of a more non-polar solvent, toluene. Thus, when the Mitsunobu reaction was carried out in the presence of a 1:1 (v:v) mixture of toluene/THF, maintaining the temperature of the reaction medium at -5 °C, apremilast was obtained with a yield of 65% and with 67% ee (Figure 3).

Conclusions
In summary, a straightforward chemoenzymatic synthesis from apremilast was developed. Nitrile 1 was an ideal starting material as it provided both intermediates used in the two biocatalytic approaches, ketone 2 and rac-4 acetate, in high yield values. In the first approach studied, among the 24 KREDs evaluated, only two enzymes acted anti-Prelog, with emphasis on KRED-P2-D12 which led to alcohol (R)-3 with 48% conversion and 93% ee, using ethanol as co-solvent. Between the two biocatalytic approaches for obtaining the alcohol (R)-3, the one that used lipases stood out compared to the one that used ketoreductases, since in this case the chiral intermediate was obtained at the maximum conversion of 50% and in the enantiomerically pure form. Lipase from A. niger has been shown to be an effective enzyme in the key stage of apremilast synthesis. It is worth mentioning that the kinetic resolution of rac-4 was only effective (conversion 50% and E > 200) in the presence of lipase from A. niger when the reaction was carried out in phosphate buffer medium containing 20% of n-butanol and when the temperature was increased from 30 to 45 °C. A plausible explanation for the increase in enantioselectivity with increasing temperature is that the preferred enantiomer in enzymatic hydrolysis, acetate (R)-4 is favored by entropy, a phenomenon that involves changes in the conformation of the active enzyme site. Thus, a combination and a balance of factors that include an increase in enzymatic stiffness by n-butanol and an increase in flexibility due to temperature increase must be responsible for the lipase from A. niger to reach an ideal conformation for a high degree of discrimination in relation to (R)-4.
Finally, apremilast was obtained through the Mitsunobu reaction of alcohol (R)-3, in the enantiomerically pure form, with acetylaminophthalimide 8, resulting in 65% yield and 67% ee. Although the reaction is expected to undergo S N 1-type mechanism, our data suggests that the use of a mixture of toluene:THF must have ensured a reasonable percentage of intimate ionic pairs in the reaction medium, providing apremilast with a reasonable value of ee.