The forward synthesis was started with protecting (−)-CBD 2 on the two phenolic groups with methyl tosylate¹⁴ to generate dimethoxy CBD 3 in quantitative yield; a subsequent regioselective epoxidation of the endocyclic double bond on the terpene moiety produced epoxide 4 in high yield. The choice of the protecting group on the phenolic hydroxy groups was restricted to methyl to enable the subsequent epoxide ring-opening reaction,¹¹ as also confirmed in our labs. A Yamamoto rearrangement, enabled a concomitant ring-opening of the epoxide with the removal of the methyl hydrogen, smoothly producing cyclohexanol derivative 5. A strong, bulky base was found to be key for high regio- and chemoselectivity in the epoxide ring-opening reaction.¹⁵ LDA proved to be efficient and cost effective for the purpose, providing allylic alcohol 5 in quantitative yield, without any purification needed. An allylic rearrangement of 5, that would directly afford primary alcohol 6, was envisaged; if successful, this would indeed provide an advanced intermediate to 7-OH CBD 1. A conjugated nucleophilic substitution¹⁶ on the mesylated-derivative 5a was attempted using hydroxide as a nucleophile in aqueous conditions; to our delight, alcohol 6 was obtained in good yields in a telescoped procedure. It should be noted that, given the multiple steps, complex, biphasic solvent and reagent mixture, the yield of the reaction is scale dependent. Care must be taken to ensure all vessel and mixing parameters are matched. To our delight, the remaining mass balance was unreacted allylic alcohol 5 that could be recovered and subjected to the allylic rearrangement again. To afford 7-OH CBD 1 as the final product, the removal of the methoxy-protecting groups on the phenolic moieties posed a stimulating challenge. Indeed, strong acids or acidic conditions induce a tertiary carbocation formation on the isopropenyl substituent resulting in a cyclization to give the THC skeleton, while strong bases in oxidizing conditions give hydroxy-1,4-benzoquinone derivatives as byproducts. A number of different approaches that would avoid the aforementioned conditions were tested (Scheme 2). Mild Lewis acids were tested, such as FeCl₃ and AlCl₃ alone or in combination with Et₃N or NaI; di-ether 6 was recovered unreacted or the reaction led to decomposition. Attention was then turned to the use of nucleophilic reagents including HPPh₂ and ^tBuOK, which have been exploited in mild deprotection protocols of aryl methyl/benzyl/allylethers,¹⁷ but unfortunately only a single phenolic oxygen was deprotected, even under forcing conditions or when subjecting the mono-deprotected product 7 to a second round of the deprotection reaction. Additionally, sodium ethanethiolate has been used to induce a nucleophilic cleavage of the C–O bond, leaving the phenolate as a leaving group. Despite the reported treatment of CBD-ether 3 with 4 equivalents of NaSEt at 150 °C for 5 hours leading to CBD 2,¹⁸ in our hands it was unsuccessful yielding only monodeprotection. A similar result was obtained in our trials of NaSEt; the reaction with dimethoxy derivative 6 afforded only mono-deprotected 7, with only trace 7-OH CBD 1 detected by ¹H NMR spectroscopy and MS. This issue suggests that in the presence of nucleophiles such as HPPh₂ or NaSEt the first demethylation provides a methoxy phenolate that increases the electron density of the aromatic ring, rendering the second deprotection more difficult. While we were facing this challenge, Passarella and co-workers reported the successful deprotection on small scale of dimethoxy CBD 3 to CBD 2 by using a large excess of NaSEt, in two sequential steps, isolating the monodeprotected intermediate.¹⁹ Unfortunately, the same protocol did not prove successful in our hands for the deprotection of dimethoxy derivative 6; 7-OH CBD 1 was obtained in very low yield and only detectable by mass spectrometry. Derivatization of phenolate 7 with an easily removable electron-withdrawing group to affect the second deprotection was attempted, along with adjusting the pH, to no avail.

Eventually, we realized that the Piers–Rubinsztajn reaction^20,21 could be the key to a successful deprotection although it had not been reported for the purpose in the context of cannabinoids. We, therefore, investigated whether it could be employed to our aim. The use of tris(pentafluorophenyl) borane (BCF) as a catalyst in the presence of silyl hydrides constitutes a mild and scalable deprotection methodology for aryl ethers.^22,23 Lewis basic oxygen species, such as methyl ethers, can attack the activated Si-center of the Lewis complex formed between BCF and a siloxane. A reductive cleavage of both O-methyl groups on the phenolic oxygens brings about the release of methane and the formation of Ar–O–Si bonds in their place, that can be easily cleaved with TBAF in acid-free conditions. More importantly for the case of deprotecting dimethoxy derivative 6, the instant formation of the silyl ether is a key point; capping the negative charge of the phenolate, thus enabling the intermediate monodeprotected species 7 to undergo a second deprotection via nucleophilic attack. Initial tests using CBD-ether 3 as a model substrate were very encouraging; ¹H NMR analysis showed the complete formation of CBD 2 as the main product. Unfortunately, subjecting dimethoxy derivative 6 to the same conditions resulted in the unexpected recovery of CBD 2. In hindsight, with the reductive cleavage of methyl groups, the unwanted deoxygenation is also promoted by the reaction system. As demonstrated by Yamamoto, primary and allylic alcohols can be reduced to alkanes using BCF and HSi(Et)₃.²⁴ Therefore, a series of protecting groups (see ESI†) were tested on the primary alcohol of an analogue of dimethoxy derivative 6; these had to be stable under the reaction conditions and, ultimately, be removed during the final step of the Piers–Rubinsztajn sequence. TIPS was eventually selected as the protecting group of choice for compound 6. After simple quantitative protection with TIPS, crude 6a is telescoped in the Piers–Rubinsztajn reaction to crude 6b, that is also telescoped to the final deprotection to 1. 7-OH CBD 1 was obtained in a satisfying 54% yield over the three steps (Scheme 3).