Strigolactones (SLs) constitute an important new class of plant hormones. Their isolation from natural resources, such as root exudates, is laborious and difficult. Therefore, synthetic SLs are needed to discover their (biological) properties. Such syntheses involve many steps. When repeating a published procedure for the synthesis of orobanchol, we noticed that the structure of the synthesized material was ambiguous. This structure was secured by means of X-ray analysis. An essential step in the synthesis, namely an allylic oxidation of the ABC scaffold, was significantly improved by using Pd/C and tert-butyl hydroperoxide (Corey's method). The second issue deals with the structure of the four stereoisomers of 5-deoxystrigol. The stereochemistry of these compounds was based on the use of Welzel's empirical rules for CD spectra. By means of X-ray analysis the stereochemistry of one of the stereoisomers was established unambiguously, thereby securing the configuration of all four isomers.
The main challenges associated with the synthesis of avenaol include the construction of a bicyclo[4.1.0]heptanone skeleton containing an all-cis-substituted cyclopropane, controlling the stereochemistry at the C8 position of the C ring, and the introduction of a C3 hydroxyl group on the A ring. The construction of bicyclo[4.1.0]heptanone skeletons has mainly been investigated in the context of constructing caged structures26,27,28. For non-caged structure, the direct synthesis of these systems has been limited to the 1,4-addition of a suitable anion of a trans-chloroallylphosphonamide29 or Ir-catalysed or Rh-catalysed cis-selective cyclopropanation reactions30, 31. However, preliminary work in our own group has shown that these methods are unsuitable for the synthesis of avenaol (Supplementary Fig. 1). Furthermore, cyclopropane rings bearing an electron-withdrawing group can readily undergo a ring-opening reaction32,33,34, further highlighting the difficulties of this approach. On the basis of these issues, we envisioned that the use of alkylidenecyclopropane35 as an appropriate intermediate would avoid an unwanted ring-opening reaction and the formation of a caged structure. We also envisioned that avenaol could be obtained from 2 by the dihydroxylation of its convex face and the introduction of the D ring (Fig. 2). The C ring lactone could be constructed by the diastereoselective transformation of the diol based on the reactivity of the cyclopropyl group in 3, which could be obtained by introduction of a hydroxymethyl group to the all-cis-substituted cyclopropane 4. Compound 4 could be synthesised by the intramolecular cyclopropanation of allene 6, which could be readily prepared from aldehyde 7, followed by double-bond isomerisation of alkylidenecyclopropane 5. The intramolecular cyclopropanation of an allene to form a six-membered carbocycle has not been reported, indicating that development work would be required to allow for the construction of the bicyclo[4.1.0]heptanone core.
Construction of an all-cis-substituted cyclopropane
Our synthesis began with the preparation of the cyclisation precursor 6 (Fig. 3). The treatment of known aldehyde 736 with 2-(prop-2-yn-1-yloxy)tetrahydro-2H-pyran and BnMe3NOH37, followed by methylation and acidic treatment gave 8. The subsequent hydroalumination of 8, followed by the treatment of the resulting intermediate with iodine gave 938, which was converted to carboxylic acid 10a via sequential protection as a triisopropylsilyl (TIPS) ether, hydroboration and oxidation by 9-azanoradamantane N-oxyl (nor-AZADO)39. The cyclisation precursor α-diazo-β-ketonitrile 6a was synthesised by sequential β-ketonitrile formation and diazo transfer reactions40. A similar sequence was used to prepare the benzyl-protected methyl diazoketone 6b. The α-diazo-β-ketoester 6c and ketonitrile 6d were also synthesised via 10b (Supplementary Fig. 2).
We subsequently investigated the formation of the alkylidenecyclopropane intermediates via the intramolecular cyclopropanation of allenes 6a–d using a Rh or Cu catalyst41. We initially investigated the cyclisation of methyl diazoketone 6b with Rh2(OAc)4, but this reaction failed to afford the desired cyclised product (Table 1, entry 1). Substrates bearing a β-ketoester or ketonitrile instead of a methyl group were also evaluated in an attempt to stabilise the metal carbene. The reaction of 6c with Rh2(OAc)4 or Cu(CH3CN)4PF6 did not give the desired product (Table 1, entries 2 and 3). In sharp contrast, the cyclisation reactions of the α-diazo-β-ketonitriles 6a and 6d with Rh2(OAc)4 proceeded smoothly to give the alkylidenecyclopropanes 11a and 11d, respectively, in excellent yields, most likely because of the more electrophilic nature of the metal carbene (Table 1, entries 4 and 5)42. It is noteworthy that this reaction only afforded the E isomer, because the metal carbene only approached from the less hindered face of the allene. For further transformation toward the all-cis-substituted cyclopropane, we used compound 11a because of the ease with which this compound could undergo protecting group manipulation.
Next, we focused on the construction of the all-cis-substituted cyclopropane structure from alkylidenecyclopropane 11a. We initially investigated the hydrogenation of the alkylidenecyclopropanes (Supplementary Fig. 3). For example, compound 12, which was prepared by reduction of 11a, followed by the PMB protection of the resulting alcohol, was hydrogenated over a Pd on carbon catalyst using H2 gas. Surprisingly, this reaction gave the undesired trans isomer as the major product, most likely because of the steric effect of the nitrile group. We subsequently investigated the transition-metal-catalysed isomerisation of the double bond in this system using a directing group to reverse this selectivity. To determine the best position for the directing groups, we prepared alcohol 13, silyl ether 14 and allyl alcohol 5 by the stepwise reduction of the nitrile group (Fig. 3). Despite our initial concerns regarding the ring-opening of the cyclopropane system during these transformations, the cyclopropane ring remained intact because it was stabilised as an alkylidenecyclopropane. Compound 12 was initially treated with Crabtree’s catalyst43, which was preactivated with H2, but failed to afford the all-cis-substituted cyclopropane (Table 2, entry 1). Use of a substrate having nitrile and hydroxyl groups resulted in no reaction, and recovery of the starting material (Supplementary Fig. 4). These results indicated that the nitrile group deactivated the Ir catalyst rather than acting as a directing group. In contrast, the reaction of 13 bearing a hydroxymethyl group under these conditions, allowed for the successful isomerisation of the olefin under H2 to give the silyl enol ether 15a in 92% yield with excellent stereoselectivity (Table 2, entry 2). The success of this reaction indicated that the Ir catalyst approached from one face after its coordination to the alcohol (i.e. X, Fig. 4a), resulting in the exclusive formation of the all-cis isomer. However, the hydroxymethyl group on 15a could not be converted to a methyl group without opening the cyclopropane ring. Notably, the treatment of silyl ether 14 with Crabtree’s catalyst resulted in very little reaction, because of the lack of a directing group of the substrate and relatively low reactivity of the catalyst (Table 2, entry 3). The reactivity improved when we used Pfaltz’s modified Ir catalyst bearing a non-coordinating counter anion (i.e. BArF)44,45,46, although this catalyst only afforded the trans isomer (Table 2, entry 4). The selectivity of this reaction was attributed to intermediate Y2, where the PMB ether would act as a better directing group rather than the corresponding TIPS ether (Fig. 4b). Allyl alcohol 5 was therefore used for this conversion. The reaction of 5 under the same conditions44, 45 gave aldehyde 4 and alcohol 16b as 2.7:1 isomeric mixtures, respectively (Table 2, entry 5). This selectivity can be explained by the preferential formation of reaction intermediate Z1 over Z2, which would suffer considerable steric hindrance (Fig. 4c). The diastereomeric ratio improved considerably when we used Pfaltz’s modified Ir catalyst to promote the coordination of the substrate to the catalyst, affording an all-cis:trans ratio of 10:1 (Table 2, entry 6). Most notably, the selectivity of this step was found to be highly reproducible, allowing us to generate gram-scale quantities of the all-cis-substituted cyclopropane 4 for further transformations.
Total synthesis of avenaol
Having established a successful route to the all-cis-substituted cyclopropane, we turned our attention to the formation of the C ring. Aldehyde 4 was converted to diol 3 by the introduction of an exo methylene at the α position of the aldehyde, followed by a hydroboration (Fig. 5). We subsequently screened a wide range of conditions to allow for the differentiation of the hydroxymethyl group. Surprisingly, the diastereoselective DDQ-mediated intramolecular cyclisation of 3 to give acetal 17 resulted in the formation of the caged compound 18 in ~30% yield as a single isomer (Supplementary Table 1, entry 1). We envisaged that the selective transformation of 18 to 22 via an oxidative ring cleavage reaction would allow for the two hydroxymethyl groups to be differentiated. Thus, we switched our focus to the unexpected formation of the tetrahydropyran ring. The treatment of this system with Cu(OTf)2 was found to be ineffective, whilst Zn(OTf)2 and Sc(OTf)3 gave the cyclised products 18 and 19 (Supplementary Table 1, entries 2–4). These results indicated that acidic conditions would be important, and that this transformation would proceed via the secondary cation intermediate A. With this in mind, we investigated the addition of BF3·OEt2 and p-toluenesulfonic acid (pTsOH) (Supplementary Table 1, entries 6 and 7). The results revealed that pTsOH gave a best yield, although a large portion of the other hydroxy group also reacted with a by-product derived from PMB group to give mixture of 18 and 19. The addition of thiophenol (PhSH) was effective for trapping this by-product, allowing for the diastereoselective formation of the desired product 18 as a single product. Interestingly, the ring-opening product was not observed under these conditions, most likely because of the stability of the bisected cyclopropylcarbinyl cation intermediate A, where the π-orbitals of the cation would interact with the sp2-like orbitals of the cyclopropane ring47,48,49.
The C–H oxidation of the tetrahydropyran ring in 18 was investigated for the construction of compound 22. The alcohol moiety in compound 18 was initially protected by a benzoyl group. The addition of a stoichiometric charge of CrO3 or a combination of RuCl3 and NaIO4 resulted in the oxidation of the ring to give the undesired lactone 23 and carboxylic acid 24, respectively (Table 3, entries 1 and 2). We also conducted the C–H oxidation according to the procedure reported by White’s group using (2S,2′S-(−)-[N,N′-bis(2-pyridylmethyl)]-2,2′-bipyrrolidinebis(acetonitrile)iron(II) hexafluoroantimonate ((S,S)-Fe(pdp))50, which gave the desired alcohol 22 in 65% yield via 21. It is noteworthy, however, that this reaction required a stoichiometric amount of an iron reagent (Table 3, entry 3). In an attempt to improve the yield, we investigated the use of dimethyldioxirane (DMDO) and methyltrifluromethyldioxirane (TFDO)51. The use of an excess of DMDO resulted in a slow reaction (i.e. starting material remaining after 1 day) (Table 3, entry 4), whereas the use of TFDO at 0 °C gave a mixture of alcohol 22 and lactone 23, presumably because of its high reactivity (Table 3, entry 5). This reaction was subsequently performed at −78 °C to improve the regioselectivity and proceeded smoothly to give 22 in excellent yield (Table 3, entry 6).
The final stage of the total synthesis involved the formation of an α-hydroxyketone and the introduction of the D ring of avenaol. Mesylation of the alcohol moiety in 22, followed by the substitution of the resulting mesylate with cyanide gave the corresponding nitrile (Fig. 5). The subsequent reduction of the ketone moiety, hydrolysis of the nitrile and benzoyl protecting groups, followed by an acidic treatment, resulted in the formation of the lactone ring to give 25. To avoid the possibility of an intramolecular cyclisation between the α position of the lactone and the C2 position of the A ring (Supplementary Fig. 5), we proceeded via the dehydration of the alcohol rather than an oxidative transformation. The stereoselective dihydroxylation of 2, followed by the selective protection of the alcohol at the C3, gave 26 with excellent selectivity. Formylation, followed by the introduction of a butenolide unit 2716 gave a mixture of C2′ epimers. Dess–Martin oxidation gave the protected avenaol 28, which was separated from the corresponding C2′ epi-28 by column chromatography over silica gel. To determine the stereochemistry, the silyl groups in both 28 and C2′ epi-28 were subsequently removed using HF·pyridine to give avenaol (1) and C2′ epi-1, respectively. Despite the low yield for the latter of these two reactions, we were able to obtained a crystal of C2′ epi-1 for X-ray crystallography. The X-ray crystal structure of C2′ epi-1 confirmed that the relative stereochemistry between C2′ and C8 was as shown in Fig 5. Moreover, the spectroscopic data obtained for synthetic avenaol (1) (i.e. 1H, 13C NMR and HRMS) were identical to those of the natural sample 19. These results therefore confirm that the proposed structure is correct.