Synthesis of amorpha-4,11-diene from dihydroartemisinic acid

Amorphadiene is a natural product involved in the biosynthesis of the antimalarial drug artemisinin. A convenient four-step synthesis of amorphadiene, starting from commercially available dihydroartemisinic acid, is reported. The targeted molecule is isolated with an overall yield of 85% on a multi-gram scale in four steps with only one chromatography.


Introduction
Amorphadiene (AD) is produced in plants by cyclization of farnesyl-pyrophosphate by the enzyme amorphadiene synthase (ADS) (Scheme 1) [1]. AD is a key intermediate in the biosynthesis of the antimalarial drug artemisinin [1,2]. In this context, the synthesis of AD has been described using a fermentation route (Amyris process) [2e4]. However, AD is not yet commercially available. Herein, we report a short and high-yielding gram-scale synthesis of AD starting from the commercially available dihydroartemisinic acid, 1, which is an intermediate in the Sanofi process to prepare artemisinin [5].
With the goal of providing a direct and scalable access to AD from the commercially available natural product dihydroartemisinic acid, 1, a three-step synthetic procedure, relying on the carboxylic acid reduction to the corresponding alcohol 2, followed by an activation/elimination sequence to produce AD, was envisioned (Scheme 2). Bouwmeester et al. described a similar synthetic approach as ours, starting from artemisinic acid, affording AD with an overall yield of 25% [6]. However, only a generic route was reported without detailed procedures, scale, and yield for each individual step.

Results and discussion
Our attempts for the direct reduction of 1 to 2 focused on the use of lithium aluminium hydride (LiAlH 4 ) as a reducing agent. The reaction was first tested using 2.0 equivalents of LiAlH 4 in anhydrous THF at 0 C (Table 1, entry 1) [7]. However, only moderate conversion of 1 was obtained (60%) and alcohol 2 was isolated in only 25% yield after purification. By increasing the amount of LiAlH 4 in freshly distilled Et 2 O and, after stirring at 23 C for 2 h, alcohol 2 was obtained with a good yield of 85% without purification ( Different conditions were then evaluated to attempt direct conversion of 2 into AD. Unfortunately, the direct elimination of the hydroxyl group, using either the Burgess reagent [8] or a one-pot selenide strategy, inspired in the Grieco method [9], failed to give satisfying results even on purified 2 (Scheme 3) [10].
Alternatively, we considered the activation of the alcohol followed by an elimination. Alcohol 2 was quantitatively converted to its corresponding mesylate 3 (MsCl, 1.1 equiv; Et 3 N, 1.5 equiv; in anhydrous CH 2 Cl 2 , 1 M) (Scheme 4) [11]. Bouwmeester et al. reported the same transformation of 2 to 3 using pyridine as solvent and base, which however, required purification by column chromatography to isolate pure mesylate 3 [6]. We replaced pyridine with dichloromethane as solvent and used only 1.5 equivalents of base, which afforded pure 3 without further purification.
The elimination of the leaving group and formation of the C]C double bond turned out to be more challenging than anticipated. Several conditions were attempted based on related literature procedures ( Table 2) [12e14]. As the direct elimination of the mesylate group failed to give satisfying results ( Table 2, entry 1), 3 was treated with sodium iodide (NaI, 5.0 equiv) and 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU, 3.0 equiv) in a one-pot reaction, upon which AD was isolated, after column chromatography, in 35% yield (Table 2, entry 2). Alternatively, DBU was added only after completion of the Finkelstein reaction using NaI (2.0e10.0 equiv) but yields in AD remained moderate (30e40%; Table 2, entries 3e5).
As the one-pot procedure was not high yielding, another method used by Baran et al. on a similar scaffold as 3, [15] consisting of isolating the iodo intermediate 4, after reacting 3 with NaI prior to the addition of base was realized (Scheme 5). The substitution of the mesylate moiety by an iodine proceeded well, affording 4 in yields to 97%. Scheme 1. Semi-synthetic approach for the production of artemisinin (Amyris/Sanofi processes).

Entry
LiAlH 4 (equiv) The reaction was conducted with non-distilled solvent.
After isolation, the iodo-compound 4 was reacted with DBU (5.0 equiv) in boiling acetone and AD was isolated in 45% yield (Table 3, entry 1). It is worth noting that 4-hydroxy-4-methylpentan-2-one was also isolated from this reaction, which presumably indicates that a self-condensation of acetone occurred and thus, acetone was not considered as a relevant solvent to carry out this transformation. In THF, a longer reaction time (24 h instead of 5 h) was required to achieve full conversion of 4 (Table 3, entry 2) but the yield in AD remained modest (45%). With these results in hand, we postulated that DBU could promote undesirable side reactions and we decided to use alternative bases. No conversion was observed with Et 3 N (5.0 equiv) ( Table 3, entry 3), whereas a moderate conversion was recorded with 5.0 equivalents of t-BuOK after 40 h at 65 C (55%) ( Table 3, entry 4). This result was improved by using microwave irradiation and gratifyingly, a complete conversion of 4 was obtained after 1 h producing AD in 89% yield (Table 1, entry 5). Reducing the amount of t-BuOK to 3.0 equiv did not give full conversion of 4 after 1 h in THF (Table 3, entry 6). By using a freshly prepared solution of t-BuOK in THF (1 M), the excess of base was reduced to 1.3 equivalents and the reaction time to 0.5 h, affording 1 in an excellent yield of 96% (Table 3, entry 8). More importantly, it allowed us to increase the concentration in 4 to 0.75 M, which also enabled to scale-up the reaction to 4 g of 4 per batch using a standard 20 mL sealed tube.
Because thermal heating is usually preferred over microwaveassisted reaction for large scale synthesis, we also investigated the preparation of AD using conventional heating. A moderate conversion (50%) of 4 was achieved when the reaction was performed in tert-butanol (t-BuOH) at 65 C (Table 3, entry 9), whereas the iodo intermediate 4 was fully converted to AD when the reaction was carried out in boiling t-BuOH (Table 3, entry 10). Ultimately, we were able to perform the elimination with only 2.0 equivalents of t-BuOK (on a 6.5 g scale) and to isolate AD in 63% yield (Table 3, entry 11). We found that, in this case, 10% of the alcohol 2 was formed along with other unidentified side products (estimated yield based on 1 H NMR).

Conclusion
In summary, a screening of conditions allowed us to secure a straightforward access to the important terpene, amorphadiene (AD), from commercially available dihydroartemisinic acid 1 (Scheme 6). The most challenging step was the elimination step, which required considerable optimization. Eventually, 1.3 equiv of t-BuOK (1 M in THF) were sufficient to achieve full conversion of 4 in 0.5 h and to produce AD in 96% yield. Alternatively, we also developed a thermal approach in boiling t-BuOH affording AD in 63% yield. It is noteworthy to mention that alcohol 2 could be transformed in one step to its corresponding bromo derivative 5 by employing an Appel bromination (see Experimental Section) [16], however, a lower yield (35%) was obtained for the elimination step.
A reliable and straightforward multi-gram scale synthesis of amorphadiene from dihydroartemisinic acid has been established  with an overall yield of 85%. This will enable future scale-up and access to this key molecule for studying its transformation into artemisinic acid and artemisinin derivatives.

General information
Reagents (Aldrich) were purchased as reagent grade and used without further purification. Reactions in the absence of air and moisture were performed in oven-dried glassware under Ar atmosphere. Flash column chromatography was performed using SiO 2 (60 Å, 230e400 mesh, particle size 0.040e0.063 mm, Merck) at 23 C with a head pressure of 0.0e0.5 bar. The solvent compositions are reported individually in parentheses. Analytical thin layer chromatography (TLC) was performed on aluminium sheets coated with silica gel 60 F254 (Merck, Macherey-Nagel) or with silica gel 60 RP-18 F 254s (Merck, Macherey-Nagel). Visualization was achieved using an alkaline aqueous solution of potassium permanganate (KMnO 4 ). Evaporation in vacuo was performed at 25e35 C and 900e10 mbar. Reported yields refer to spectroscopically and chromatographically pure compounds that were dried under high vacuum (0.1e0.05 mbar) before analytical characterization. 1  constants J are given in Hz and the resonance multiplicity is described as s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), and br (broad). All spectra were recorded at 298 K. Infrared (IR) spectra were recorded on a Bruker Tensor 27 FT-IR spectrometer and are reported as wavenumbersñ(cm À1 ). Highresolution mass spectrometry (HRMS) was performed by the Laboratoire de Spectrom etrie de Masse from Sorbonne Universit e, Paris. Gas Chromatography coupled to Mass Spectrometry (GC/MS) analysis was performed on a Shimadzu GCMS-QP2010S using an electronic impact (EI) spectrometer. Low-resolution mass spectra (LRMS) result from ionization by electronic impact (EI-LRMS). The abundance indicated for each mass number (m/z values) is given in percentage relative to the strongest peak of 100% abundance (base peak). Melting points were determined using a Büchi melting point apparatus in open capillaries. Nomenclature follows the suggestions proposed by the software ChemDraw Professional 2016.