Progress Toward a Large-Scale Synthesis of Molnupiravir (MK-4482, EIDD-2801) from Cytidine

Molnupiravir (MK-4482, EIDD-2801) is a promising orally bioavailable drug candidate for the treatment of COVID-19. Herein, we describe a supply-centered and chromatography-free synthesis of molnupiravir from cytidine, consisting of two steps: a selective enzymatic acylation followed by transamination to yield the final drug product. Both steps have been successfully performed on a decagram scale: the first step at 200 g and the second step at 80 g. Overall, molnupiravir has been obtained in a 41% overall isolated yield compared to a maximum 17% isolated yield in the patented route. This route provides many advantages to the initial route described in the patent literature and would decrease the cost of this pharmaceutical should it prove safe and efficacious in ongoing clinical trials.

Since the yield for the final two steps is not reported, this route has a 17% maximum overall yield (Scheme 1). We previously demonstrated the potential of a two-step route from cytidine (8) to molnupiravir (7), which has many advantages over the previously patented route including cost and overall yield ( Figure 1). 5 Figure 1. This work (green) compared to previous work (blue). 5 However, several challenges prevent the previously reported route from being implemented at manufacturing scale. Most notably, both the intermediate and the final API were purified by column chromatography, so we sought to develop alternate workup and crystallization procedures to provide pure material without chromatographic purification. Other opportunities for improvement included examining the surprisingly high cost of acetone oxime and refining process parameters such as reaction concentration, catalyst loading, and stoichiometry. Thus, we set out to re-examine this route to enable this process to be run at an increased scale.

Results and Discussion:
The oxime ester acylating agent (9) proved uniquely suited to selective enzyme acylation (SI Section 2.1). Unfortunately, the acetone oxime used in the synthesis of the acylating agent (Scheme 2a) proved to be a major cost driver of our initial route investigation (SI Section 1).
Surprised by the high cost of such a simple material, we sought to determine the cost of synthesizing this material ourselves. Gratifyingly, we found that a reported procedure 6 for acetone oxime synthesis provided 64% yield on our first attempt (Scheme 2b). Use of methyl tert-butyl ether (MTBE) to replace diethyl ether in the workup produced a yield of 71%. This method was repeated on 500 g scale with NaOH as base, and 73% yield was observed. We anticipate a significant reduction of the cost of acetone oxime prepared by this procedure compared to current commercial sources.
With a cheaper preparation of acetone oxime in hand, we also sought to improve the synthesis of oxime ester 9. Our previous synthesis (Scheme 2a) used a slight excess of isobutyric acid chloride and triethylamine, and 50 volumes (V) of dichloromethane (DCM) as solvent. We found that we could increase the throughput of the reaction by decreasing the solvent to 30 V, as well as increasing triethylamine addition from 1. Streamlined preparation of acylating agent 9 facilitated investigation of the enzymatic acylation of cytidine ( Figure 1, Step 1). We first sought to investigate environmentally preferable solvents to substitute for 1,4-dioxane (dioxane). 7 Unfortunately, extensive solvent screening ( Figure 2, additional data in SI Section 2.2) did not reveal another solvent that worked as well as dioxane for the reaction. Cyclopentanone appeared promising based on our initial screening, and its high boiling point (131 ºC) opened the possibility for an increased reaction temperature that we hoped would accelerate conversion. Unfortunately, higher temperatures led to lower conversion (SI Section 2.3), possibly due to release of the enzymes from the polymer beads 8 or degradation of the product. The lower conversion with cyclopentanone led us to return to dioxane as the optimal solvent. Last, we did a cursory investigation to ascertain whether a lower temperature or concentration could benefit the reaction; we found that lowering the temperature to 40 ºC lowered conversion (SI Section 2.4).
We then screened other enzymes in dioxane and additional solvents, but the originally identified  We hoped that increased enzyme loading would allow us to decrease solvent usage, which would in turn improve the reaction throughput and process mass intensity (PMI). Unfortunately, our screening revealed that increasing concentration significantly hindered conversion; we hypothesize that this effect is due to the limited solubility of cytidine in dioxane. Ultimately, we chose conditions using 4.0 equivalents of oxime ester, 150 weight % N435, and 50 volumes of dioxane as solvent, due to the highest distribution of product 10 relative to remaining starting material 8 and diacylated product 14 (SI section 3.3).
Additionally, the breadth of data on regenerating and recycling N435, including repeated uses in organic solvents, 8,9a-d suggests that N435 could be recycled for this reaction as well. Reuse of the enzyme would further decrease the raw material costs associated with this route.
Upon scaling up the reaction, we noticed a slowed reaction rate, from 24 hours to completion at small scale to approximately 40 hours at 19 g scale (SI Section 2.8). We believe this change is related to differences in the stirring method used at different scales: we stirred using magnetic stir bars at 750 rpm for scales up to 1 g, while larger-scale reactions used an overhead stirrer set to 260 rpm. Using this same setup, the reaction was scaled ten-fold to 200 g with no change in performance.
Throughout the reaction optimization, the major impurity observed was diacylated product (14), the structure of which was confirmed as the N-acylated product via 15 N-1 H 2-D NMR (SI Section 2.9). However, we also saw a consistent unknown impurity in the HPLC, constituting 5-10 LC area percent (SI Section 2.10). We noticed that this impurity increased in proportion with enzyme loading, leading us to believe that the impurity was leached from the enzyme beads by dioxane. Indeed, a control reaction confirmed that the impurity formed simply from stirring the enzyme beads in dioxane at 60 ºC overnight and was not related to either the cytidine or the acylating agent (SI Section 2.9). Isolation of this impurity yielded a yellow gel that did not ionize on LC-MS. We hypothesize that this material is a macromolecular compound leached from the Lewatit VP OC 1600 solid support; solubilization of the support in studies of N435 catalysis has been reported for some solvents. 8 Fortunately, simply rinsing the enzymes with dioxane before use eliminated this impurity (SI Section 2.9). Thus, the main impurities remaining in the reaction were the diacylated product 14, unreacted cytidine 8, and excess acylating agent 9. Allowing this reaction to cool to room temperature (20 °C) and then filtering enabled the removal of the enzyme beads as well as unreacted starting material (SI Section 3.2); this was due to cytidine's insolubility in room temperature dioxane.
The main challenge for purification was therefore removing excess acylating agent and diacylated product from the desired product 10. Upon completion of the reaction, remaining cytidine starting material as well as the enzyme beads were filtered out and then dioxane was removed via rotary evaporation. We then determined three methods (A, B, and C) for the purification of 10 from compounds 14 and 9 (Table 1). Methods A and B were successful for the purification of desired product 3 up to purity of >99% on smaller scales (10−100 g), while purification C was successful on up to 200 g scale. The transamination reaction of intermediate (10) to MK-8842 (7) was previously reported in 96% yield with 94% purity via column chromatography. 5 Previously, 70% aq. isopropanol was used as the solvent for this reaction; however, we chose to replace this with 70% aq. 1-butanol solution. We hypothesized that the decreased solubility of butanol in water (as compared to the solubility of isopropanol in water) would decrease the amount of hydroxylamine salts present in the organic layer post-reaction, thus leading to a more facile purification. The optimization of this transamination reaction began by using 4.5 equivalents of hydroxylamine sulfate relative to starting material 10, 5 which displayed good conversion of 10 to desired product 7 after 22 hours (SI Section 4.1). In addition to desired product formation, side product N-hydroxycytidine (NHC, 15) was formed in this reaction in small quantities. It was determined that decreasing the stoichiometry of hydroxylamine sulfate from 4.5 equivalents to 2 or 3 equivalents led to very similar reaction conversion (Figure 4, SI Section 4.1) and increased the purity of the crude reaction mixture after workup, which simply involved separating the organic and aqueous layers and then removing the 1-butanol from the organic layer (crude purity, Table 2).   Table 2.
After the development of successful small-scale conditions (5 g) the next goals were to scale up this reaction to 80 g and to develop recrystallization conditions for the crude reaction mixture. We found that increasing the scale from 5 to 10 to 20 g did not significantly affect the reaction time    (Table 3). However, increasing the scale to 80 g required an increased reaction time to 40 hours for similar conversion ( Figure 5).   Table 3.
We found that the pre-crystallization purification of this reaction on 80 g scale worked the same as on a 1 g scale, by simply separating the aqueous and organic layers. In order to increase the purity of the final compound from 80-90% after this initial purification, recrystallization conditions  were developed. Due the difference of solubility of compound 7 in cold versus hot water (Table   4), water was chosen as the recrystallization solvent. Thus, it was determined that after the solvent was removed from organic layer, 7 could be recrystallized in greater than 99% purity from heating to 65 °C and slowly cooling in water (2 V). After the initial recrystallization, the remainder of 7 remained in the aqueous filtrate. We anticipate that further recrystallizations of the filtrate would produce additional molnupiravir.   were then separated, and 1-butanol was removed from the organic layer via rotary evaporation to yield solid white, crude material (5.2 g, 92% purity via qNMR). This crude material was then dissolved in water (2 V) and heated to 65 °C for 30 minutes (60 minutes for 80 g scale). After completely dissolved, the mixture was allowed to cool to room temperature and then 5 °C without stirring. The solid was then filtered and washed with MTBE (2 V) to obtain molnupiravir (5.03 g, 48% yield, >99% purity by qNMR). 1  butanol (20 V) and then hydroxylamine sulfate (134 g, 817 mmol, 3.2 equiv) was added. The mixture was stirred vigorously and heated to 75-80 °C for 40 hours. The mixture was cooled to room temperature, the layers were then separated, and 1-butanol was distilled from the organic layer yielding solid white, crude material (85 g). This crude material was then dissolved in water (2 V) and heated to 60-65 °C for 60 minutes. After completely dissolved, the mixture was allowed to cool to room temperature and stirred for 1 hour, then cooled to 5-10 °C and stirred for 3 hours.
The solid was then filtered and the wet washed with MTBE (3 V) to obtain 56.0 g molnupiravir.
The purification was repeated again for a final yield of 50.0 g (58% corrected) with 97% purity (qNMR).
ASSOCIATED CONTENT Supporting Information.
The following files are available free of charge.