Organic & Biomolecular Chemistry

A route to enantiopure ( R )-(+)-3-methyl-6-isopropenyl-cyclohept-3-enone-1, an intermediate for terpenoids, has been developed and includes a highly chemo- and regioselective Ti ﬀ eneau – Demjanov reaction. Starting from readily available ( R )-( − )-carvone, this robust sequence is available on a deca-gram scale and uses ﬂ ow chemistry for the initial epoxidation reaction. The stereochemistry of the addition of two nucleophiles to the carbonyl group of ( R )-( − )-carvone has been determined by X-ray di ﬀ raction studies and chemical correlation.

Synthetic approaches to the perhydroazulene structures from para-menthanes have involved the contraction of the sixmembered ring to cyclopentanoids followed by heptanyl-annulation, 30,31 or much less frequently expansion to cycloheptanoids and pentanyl-annulation. 32 This last strategy has been studied in our laboratory for the synthesis of cycloheptanoids, 33 and thus access to perhydroazulene terpenoids and alkaloids. 34The ring expansions occur by two different methods: cyclopropanation of a suitable para-menthene-1 to an overbred bicyclic system followed by cleavage of the common C-C bond. 35The second sequence requires an appropriate nucleophile addition to a para-menthanone-2 followed by a regioselective rearrangement, as presented in Scheme 2.
We were pleasantly surprised to discover the complete chemo-and regioselectivity in this rearrangement, as can be seen in Scheme 4, where the epoxide 6 and the regio-isomeric   cycloheptenone 7 were not observed.Similarly, the Nozaki ring expansion with addition of the dibromomethyl carbanion to carvone, and subsequent rearrangement of 5 are also highly effective (Scheme 3). 36ncouraged by the total chemo-and regioselectivity observed in the Tiffeneau-Demjanov rearrangement, we have now examined an attractive alternative by Corey-Chaykovsky epoxidation of (R)-(−)-carvone (1) and N-nucleophile ring opening to the same intermediate amino-alcohols 4a/4b obtained in Scheme 3. We now demonstrate significant improvements over the previous synthetic routes, which allow a deca-gram scale-up, in less bench time, with very simple purifications, thus reducing substantially problems of synthesis logistics.This route also avoids the practical problems of using KCN and LiAlH 4 , or the preparation and use of the CHBr 2 carbanion, on a large scale.We also present a structural assignment of 4a by X-ray diffraction and a chemical correlation, and thus the stereochemistry of 1,2 nucleophilic additions to the carbonyl group of (R)-(−)-carvone (1).

Results and discussion
Our initial experiments with the Corey-Chaykovsky epoxidation of (R)-(−)-carvone (1), 37 using the original NaH procedure for generating the sulfur ylide, were not reproducible on a large scale, besides presenting some practical difficulties with the manipulation of large quantities of NaH.This led us to try the modification using methyl lithium or n-butyl lithium hexane solutions with DMSO, 38 and in these conditions, the epoxidation of (R)-(−)-carvone (1) led to a 90 : 10 ratio (by GC and 1 H-NMR) of 6a and 6b in 90-95% yields (Scheme 5), and this mixture was used as such in the following reaction.
To test the quality of our commercial Me 3 S + I − , we prepared this reagent freshly by reaction of Me 2 S with MeI, 39 and after recrystallisation the Me 3 S + I − turned out to be slightly more efficient in our reaction.We conclude, however, that the commercial product is perfectly adequate for our purposes, not justifying the time spent on its preparation and purification.
The base used to form the ylide and reaction conditions were also studied, and the results are shown in Table 1.A solution of the alkyl-lithium was added to DMSO at room temperature, a biphasic mixture was formed, the dimsyl-Li (heavier phase) was transferred, via cannula, to a previously prepared solution of Me 3 S + I − in THF/DMSO at −10 °C.The (R)-(−)carvone solution in THF was added, and after 3 h a crude oil containing the mixture of epoxides 6a/6b was obtained, to be used as such in the following reaction.
The epoxides 6a/6b were formed on a 23 gram scale from 18 g (120 mmol) of carvone under the conditions shown in Table 1 (entry 9), and used as such in the subsequent reaction.
However, a significant improvement was found while conducting this reaction under continuous flow conditions. 40We used a Vapourtec E-series 41,42 flow equipment, with 3 peristaltic pumps (A, B and C), a 10 mL coil reactor and 1.0 mm i.d.PTFE tubing.The equipment schematics with the best results are shown in Scheme 6.
n-BuLi in hexanes was pumped through C (we used a red end crimped fluoropolymer in this pump, compatible with organometallic solutions) and then mixed with a solution of Me 3 S + I − in DMSO pumped by B. The ylide then meets the (R)-(−)-carvone stream from A, and the reaction takes place in the 10 mL reactor coil kept at room temperature.After the continuous off-line aqueous work-up the desired epoxides 6a/6b were obtained.DMSO (99%+, Alfa-Aesar) was used directly from the bottle, without any further treatment in the flow experiments.
In the initial experiments, our attempts to optimise the procedure were accompanied by blockages within 15 min of processing.The blockages came from the precipitation of the Me 3 S + I − and/or LiOH from n-BuLi hydrolysis.This problem was solved using a wider bore T-piece and tubing (2.0 mm i.d.)  as shown in Scheme 6.
After the standard parameters had been optimised, the preparation was continuously run for over 2 hours, to generate the epoxides 6a/6b (9.4 g, 57 mmol) in 95% isolated yield, from 9.0 g (60 mmol) of (R)-(−)-carvone, 18.4 g (90 mmol) of Me 3 S + I − and 40 mL (90 mmol) of n-BuLi 2.25 M in hexanes.The crude colourless oil was used in the next step without any need of further purification.
In the batch process using NaH as the base, we obtained the epoxides 6a/6b in 87% yield, whereas the batch process using n-BuLi as the base and a continuous flow process led to 95% yield without the need for purification, and ready for the next synthetic step.The continuous flow procedure was also better due to the ease of operation at much larger scales with minimum manual interactions, resulting in an increase in bench time and simplicity.
Initially the ring opening of the epoxides 6a/6b (Scheme 5) was studied with the obvious nucleophiles ammonia 43  sodium amide, 44 but with limited success.For example, we used a commercial ammonia solution in H 2 O (25-30%, Fisher Scientific) and prepared other ammonia solutions by bubbling ammonia gas, at room temperature, into the desired solvent for 2 h.The solutions were titrated with a 0.12 M aqueous HCl standard solution containing bromocresol green.
We obtained ammonia solutions in MeOH (8.0 N), isopropanol (4.0 M), dimethoxyethane (2.9 M).These ammonia solutions were then reacted in screw cap sealed pressure tubes with the epoxide mixture 6a/6b, and led to the formation of the desired amino-alcohols 4a/4b, and unidentified by-products.The corresponding diols were also formed when we used aqueous ammonia solutions, due to the presence of water (in a 26 : 10 ratio) (Table 2).The separation of the aminoalcohols on a multi-gram scale we deemed to be impracticable.
On the other hand, sodium amide led to complex mixtures as shown by TLC analysis, probably due to its reactivity as a base.
Ring opening of the epoxides 6a/6b was easily accomplished by potassium phthalimide and phthalimide in dimethylacetamide (DMA) or DMF at 155-160 °C, according to the classic Gabriel procedure. 45Table 3 summarizes our epoxide ring opening results under conventional or microwave (MW) heating, with DMA presenting better results.
Microwave heating at 150-160 °C in DMA gave the best results, but due to the volume limitation of the microwave tubes (∼14 mL), the large scale reactions (2 flasks with 8.2 g, 50 mmol each and 23.1 g, 120 mmol; Table 3 entries 15 and  16) were performed with conventional heating.The crude product obtained as a brown oil was used directly in the next step.
The phthalimido-alcohol 9a/9b reduction described by Ganem et al. 46 was also examined using NaBH 4 in isopropanol : H 2 O at room temperature for 24 h.A yellow oil was obtained containing the amino-alcohols 4a/4b together with a complex mixture of by-products.As the product 4a/4b    The assignment of the stereochemistry of 3a/3b and thus 6a/6b was made by chemical correlation with 4a, from the mono-crystal X-ray diffraction studies (Fig. 3) and nOe irradiation (see ESI †) of the major amino-alcohol 4a, obtained by reduction of the TMS-cyanohydrin mixture 3a/3b.Other nucleophilic addition reactions to the carbonyl group of carvone have been studied previously, 49 and the stereochemistry of major approaches has been shown to be preferentially anti-to the isopropenyl group. 50We have now confirmed that this is the correct stereochemistry for the major isomers 3a and 6a.The structure of 4a is shown in Fig. 3, and the chemical correlation of the amino-alcohols 4a/4b obtained in both sequences, establishes the same 90 : 10 diastereomeric preference of addition of cyanide and the sulfonium ylide nucleophiles.

Experimental
General protocol for preparation of 6a/6b in flow The continuous flow preparation of the epoxides 6a/6b was carried out using a three-stream reactor assembly.The Vapourtec E-Series machine was charged with a 0.5 M solution of (R)-(−)-carvone in DMSO ( pump A) at the rate of 1.0 mL min −1 , a 0.4 M solution of Me 3 S + I − in DMSO ( pump B) at the rate of 1.88 mL min −1 and a solution of n-BuLi (2.25 M in hexanes) pumped directly from the bottle through C at the rate of 0.340 mL min −1 .The DMSO was used directly without any purification.The desired flow rates were set and all pumps were started.DMSO and hexane were pumped for 5 min.Pump C was timed to switch to pumping n-BuLi for 5 min before switching pumps A and B simultaneously to (R)-(−)-carvone and Me 3 S + I − at the rates as determined above.The streams of pumps B and C were mixed through a T-piece generating the sulfur ylide which was mixed with a stream of (R)-(−)-carvone from pump A. A PTFE tubing of 2.00 mm i.d. was used between pump C and the second T-piece.The resulting stream was driven to a 10 mL coil reactor at room temperature with a residence time of 3.1 min at these flow rates.The quenching was achieved by continuously collecting the output in a conical flask with cold water for 2 h.

Procedure for hydrazinolysis of phthalimido-alcohols
To a suspension of the crude products 9a/9b in EtOH was added NH 2 NH 2 •H 2 O.The reaction system was heated at 80-85 °C for 3 h, cooled down to room temperature, and the white solid formed ( phthalyl hydrazide) was filtered off in a sintered funnel and washed with EtOH.The ethanolic filtrate afforded after evaporation in vacuo a yellowish oil mixed with a solid characterized as 4a/4b.The crude mixture of amino-alcohols 4a/4b can be used directly in the next step.

Scheme 1
Scheme 1 Synthetic methods to obtain cycloheptane rings, with target terpenoid carbon skeletons.

a
Conversion based on the 1 H-NMR signal ratios of H-11 of 6a/6b, 9a/9b and 10. b Performed simultaneously in two flasks containing the same quantities (8.21 g, 50.0 mmol) of epoxides 6a/6b.

Fig. 3
Fig. 3 Structure of the major amino-alcohol 4a, obtained by X-ray diffraction studies, indicating observed nOe interactions.(Ellipsoids shown at 40% probability level).

Table 2
The opening of epoxides 6a/6b with ammonia solutions a Conversion based on 1 H-NMR signal ratios of H-2 of 4a/4b, 8 and by-products.