Our synthesis started from commercially available (+)-β-citronellene (10). The ozonolysis of the trisubstituted double bond followed by a reductive treatment with NaBH4 chemoselectively afforded primary alcohol 11 in 78% yield (Scheme 1). Due to easier purification, this alcohol was preferred to the aldehyde in our synthetic route, allowing a key stereoselective Krische allylation [21,22] to be envisaged. Applying reported conditions for this allylation – in presence of allyl acetate (10 equiv), [Ir(COD)Cl]2 (2.5 mol %), (S)-SEGPHOS (5 mol %), 3-nitrobenzoic acid (10 mol %), Cs2CO3 (20 mol %) in THF at 100 °C for 24 hours – we obtained homoallylic alcohol 12 in a good 86% yield, with a diastereomeric ratio (dr) of 93:7 deduced from the NMR analysis of the methyl substituent signals in CD3OD (NMR spectra compared to those of a 50:50 mixture of diastereoisomers, obtained from the addition of allylmagnesium bromide onto the corresponding aldehyde). The stereochemistry of the resulting secondary alcohol was expected to be (R) according to Krische's studies involving (S)- SEGPHOS [22]. This result was secured by the NMR analysis of Mosher's esters made from (R)-(+)- and (S)-(−)-α-methoxy-α-trifluoromethylphenylacetic acid (MTPA) (see Supporting Information File 1) [23-25], confirming the installation of the C-11 stereocenter of latrunculins.

Scheme 1: Synthesis of fragment 15 from (+)-β-citronellene (10).

The next steps consisted in the functionalization of 12, in view of its coupling to 8. We first relied the chemoselective epoxidation of the homoallylic alcohol, done in presence of VO(OiPr)3 (20 mol %) and t-BuOOH to afford epoxide 13, in 86% yield and a dr of 75:25 (measured by NMR, presumably resulting from the major diastereoisomer of 13; minor isomers were not identified), when the reaction was performed at room temperature during 6 hours. This vanadium catalyst superseded VO(acac)2 in terms of yields [26,27]. Additional epoxidation attempts allowed to improve the dr to 82:18 (82% yield) when the reaction was left at −30 °C for 6 days. Unfortunately, it was not possible to set up an appropriate nucleophile through the umpolung of aldehyde 8 to react with this epoxide, which led us to envisage the following aldol strategy through ketone 15. Attempts of Wacker reactions to produce 15 were unsuccessful on 12, presumably due to a competition between the two olefinic parts. After protection of the secondary alcohol as a para-methoxybenzyl (PMB) ether (78% yield of 14), the ketone (15) was installed in two steps from the epoxide (direct rearrangement attempts of the epoxide to form the ketone were unsuccessful). Thus, the epoxide was first reduced on its primary carbon in presence of LiAlH4, and the resulting secondary alcohol was oxidized in presence of Dess–Martin periodinane (DMP), giving ketone 15 in 78% yield over the two steps. This six-step sequence to 15 was performed in a 35% overall yield from starting material 10.

The aldehyde partner (8) for the aldol reaction brings the thiazolidinone heterocycle of the natural product. It was synthesized in four steps from ʟ-cysteine ester derivative 16, first reacting with carbonyldiimidazole (CDI) to afford thiazolidinone 17 in 85% yield (Scheme 2). The nitrogen atom was protected with a PMB group in 72% yield (18), after deprotonation with NaH and reaction with PMBBr. The ester moiety of 18 was then chemoselectively reduced into alcohol 19 in 90% yield, in presence of LiBH4 to avoid the reduction of the thiazolidinone part. Finally, the aldehyde (8) was generated in 78% yield by oxidation in presence of DMP.

Scheme 2: Synthesis of fragment 8 from ʟ-cysteine ethyl ester hydrochloride (16).

The assembly of aldehyde 8 and methyl ketone 15 was envisaged through a stereoselective aldol reaction. After unsuccessful attempts of Mukaiyama aldol reactions with silyl enol ethers [28], we found that dicyclohexylboron enolate 20, made in situ from ketone 15 and Cy2BCl in presence of DIPEA, performed well in the aldol reaction to furnish product 21 in 55% yield with a good dr of 91:9 (Scheme 3).

Scheme 3: Synthesis of fragment 21 through a stereoselective aldol reaction.

The stereocontrol of the reaction could be envisaged through two principal mechanisms. A remote stereocontrol by the nucleophile could first be expected [29], through a 1,5-anti-induction of the aldol stereocenter by β-alkoxy ketone 9, leading to an (S)-configuration [18-20]. This control is supposed to follow a boat transition state A stabilized by a formyl hydrogen bond [30]. It is known to be dependent on the nature of the β-alkoxy substituent, being particularly favoured by the PMB and other aromatic groups, while being disfavoured by silyl protecting groups. Alternatively, an (R)-configuration of C-15 could result from a polar Felkin–Anh model controlled by aldehyde 8 through chair-transition state B [14-16]. To determine the configuration of C-15, we initially relied on the comparative NMR analysis of Mosher's esters [23]. Despite clear 1H NMR spectra, irregular values of ΔδS–R precluded the unambiguous determination of the C-15 stereocenter [24]. These difficulties were attributed to the hindered character of this secondary alcohol, substituted by the thiazolidinone ring, possibly leading to a strong conformational distortion of Mosher's model. The question of the resulting stereoselectivity was thus left open for later resolution.

To complete this study, the 1,3-anti-diastereoselective reduction of β-hydroxyketone 21 was undertaken through the Evans–Tishchenko method [31,32], in presence of SmI2 and an aldehyde (Scheme 4). para-Nitrobenzaldehyde was used [33] to introduce a labile para-nitrobenzoate on the product, planning an easy deprotection of the alcohol. This would also pave the way to an orthogonal manipulation of protecting groups on the stereopentade, in view of designing molecular tools for biological purpose. The reduction took place in 76% yield with complete stereoselectivity. However, a mixture of two inseparable products was obtained, containing the expected but minor alcohol 22 (10%), and more surprisingly the major isomer 23 (66%). This compound results from the transposition of the para-nitrobenzoyl (PNBz) group onto the 13-OH, which could be favoured by the steric hindrance of C-15 and a possible π–π stacking with the OPMB group.

These PNBz esters were readily hydrolyzed to furnished diol 24 in 97% yield. The oxydation of the PMB group, in presence of DDQ under anhydrous conditions [18], gratifyingly afforded acetal 25 in 74% yield, whose stereochemical assignment by NOESY NMR experiment showed the syn stereochemistry of the acetal. By deduction, it was confirmed that the asymmetric boron aldol reaction between 8 and 15 proceeded through a 1,5-anti induction by the ketone to form 21. Most importantly, compounds 22–25 bear the (11R,13R) configuration of latrunculins (1 and 2).

Scheme 4: 1,3-Anti-diastereoselective reduction of 21 with PNBz transposition, and final determination of the relative stereochemistry by NOESY experiment on 25.