Main

Polyene cyclizations are one of the most complex and challenging reactions in nature and assemble complex molecular architectures from structurally simple precursors1,2,3. The conversion of the acyclic triterpenoid squalene to pentacyclic (+)-hopene, for example, is catalysed by a single enzyme, squalene–hopene cyclase11,12,13 (SHC; Fig. 1a,b).

Fig. 1: Origin of the ambergris odorant (−)-ambrox and its synthesis by polyene cyclizations.
figure 1

a, Polyene cyclization of squalene to (+)-hopene and (+)-ambrein catalysed by class II terpene cyclases and formation of the odorant (−)-ambrox from (+)-ambrein, the main constituent of ambergris. b, Active site of SHC in complex with its inhibitor 2-azasqualene (PDB code: 1UMP) (ref. 13). The conserved DXDD motif required for substrate protonation is highlighted. c, Mechanistic considerations for the polyene cyclization of (3E,7E)-homofarnesol to (−)-ambrox. d, Stork–Eschenmoser hypothesis for the stereospecific antiparallel addition of carbenium ion to an alkene in 1,5-polyolefin cyclizations. e, Catalytic asymmetric concerted polyene cyclization of (3E,7E)-homofarnesol to (−)-ambrox realized in this work. Depictions of SHC and the substrate/catalyst model have been rendered using UCSF Chimera X (refs. 58,59).

The cyclase promotes substrate preorganization and conformational selection according to the Stork–Eschenmoser hypothesis8,9,10,14 (Fig. 1c,d), stabilizes transient charges and specifically selects the initial protonation site11,15. Out of 512 theoretically possible stereoisomers, (+)-hopene is produced as a single all-equatorial stereoisomer. Squalene is also considered to be the biosynthetic precursor of the triterpene (+)-ambrein (Fig. 1a), the main constituent of ambergris, a grey waxy substance formed in the gastrointestinal tract of the sperm whale (Physeter macrocephalus)16. The use of ambergris as fragrance, condiment and medicine dates back to several centuries5,7. Until now, it is highly valued for its distinctive scent. The most important odorous component in ambergris is the rare naturally occurring terpenoid (−)-ambrox (2a), formed on photo-oxidation of (+)-ambrein17,18. Producing (−)-ambrox with high diastereo- and enantioselectivity is of particular importance, as each diastereomer and enantiomer differs in its odour threshold and exhibits a different and sometimes even disagreeable odour sensation19,20. Solely all-trans configured ambrox and other diastereomers that fulfil the triaxial rule established by Ohloff deliver the desired odour impression with warm amber tonalities18,21,22.

Until now, the most efficient syntheses of (−)-ambrox by polyene cyclization harness genetically engineered SHCs23,24,25,26,27,28,29,30,31. Inspired by SHC and other class II terpene cyclases3, chemists have long sought for small-molecule catalysts that can effect polyene cyclizations with similar efficiencies1,6,14,32,33,34,35. Pioneering work mediated by Lewis-acid-assisted Brønsted acids4,36 illustrates the difficulty of this transformation. In a stepwise process, (3E,7E)-homofarnesyl triethylsilyl ether is cyclized to cyclohomofarnesols and homodrimenols in the presence of two equivalents of SnCl4-coordinated (R)-2-(o-fluorobenzyloxy)-2′-hydroxy-1,1′-binaphthyl. A subsequent diastereoselective cyclization using excess trifluoroacetic acid (TFA, 10 equiv.) and SnCl4 (2 equiv.) affords (−)-ambrox (54% yield, 74:26 diastereomeric ratio (d.r.) and 87.5:12.5 enantiomeric ratio (e.r.)). The complexity becomes increasingly evident on consideration of potential intermediate steps en route to the desired fully cyclized product (Fig. 1c). First, regioselective protonation of the distal double bond of the polyene substrate is crucial to avoid the formation of unproductive side products 6 and 7. Second, in the case of a stepwise process, undesirable diastereomers resulting from protonation of cyclohomofarnesols 3 are formed, as shown in previous comprehensive studies37. Third, exerting enantiocontrol over the entire cyclization sequence to produce the desired antipode of ambrox as a single enantiomer is extremely challenging. To the best of our knowledge, simultaneously achieving high chemo-, diastereo- and enantioselectivity in a catalytic asymmetric polyene cyclization towards (−)-ambrox has not been realized with small-molecule catalysts so far. Ideally, a suitable Brønsted-acid catalyst should not only exhibit sufficient acidity to protonate olefins but also provide a confined and enzyme-like pocket to effect the desired transformation with high stereoselectivity. Encouraged by our previous studies on catalytic asymmetric functionalizations of unactivated olefins using IDPi catalysts38,39,40,41,42, we proposed that this particular class of strong and confined chiral Brønsted acids43,44,45 would be ideally suited to enable the asymmetric polyene cyclization of homofarnesol. Here we describe the development and realization of a catalytic asymmetric synthesis of (−)-ambrox that proceeds with unprecedented stereoselectivities for small-molecule catalysts (Fig. 1e).

Reaction development

At the onset of our study, extensive screening of solvents showed a substantial increase in reactivity in fluorinated alcohols, which was recently exploited in numerous transformations involving carbocationic intermediates46,47,48. Extensive studies on the role of perfluorinated solvents in diastereoselective polyene cyclizations suggest that perfluorinated alcohols assemble a catalytically active H-bonding network that controls substrate conformation and stabilizes cationic intermediates46 in line with previous studies49,50,51,52. Based on this initial result, we aimed to improve the selectivity using different chiral Brønsted acids in 1,1,1,3,3,3-hexafluoropropan-2-ol (HFIP). We observed that IDPi catalysts converted homofarnesol to ambrox and its minor diastereomer 5β,8α,9β-ambrox 2c as opposed to other classes of previously developed chiral Brønsted acids (see Supplementary Information for full optimization studies). Subsequently, a screening of structurally diverse IDPi catalysts was conducted (Fig. 2a).

Fig. 2: Reaction optimization and scale-up.
figure 2

a, Catalyst screening and optimization of the polyene cyclization of (3E,7E)-homofarnesol to (−)-ambrox. See Supplementary Information for details. b, Stacked gas chromatography traces of the crude reaction mixture obtained with PADI 9 (top) and IDPi 8g (middle) as catalysts. Enantiopure (−)-ambrox is shown as a reference (bottom). c, Scale-up of the catalytic asymmetric polyene cyclization in the presence of IDPi catalyst 8g in PFTB affording (−)-ambrox. d, Technical synthesis of (3E,7E)-homofarnesic acid 11 from (E)-nerolidol and catalytic asymmetric polyene cyclization of 11 to (+)-sclareolide. r.t., room temperature; d.r., diastereomeric ratio; e.r., enantiomeric ratio; e.e., enantiomeric excess.

Combining the C8F17-perfluoroalkyl chain in the imidodiphosphorimidate (IDPi) core (catalyst 8d) with a 4-isoamyl-C6H4 group in the 3,3′-position of the 1′-binaphthyl-2,2′-diol backbone delivered the desired product in 40% yield with good enantioselectivity (7:93 e.r.) alongside remaining cyclohomofarnesols 3. The addition of 1H,1H-perfluoro-1-octanol enabled the reaction to run at lower reaction temperatures that provided ambrox with improved d.r. (90:10) and e.r. (4:96). Changing the solvent to perfluoro-tert-butanol (PFTB)46 gave similarly high reactivity and selectivity compared with HFIP. After extensive optimization, spirocyclohexyl-2-fluorenyl-substituted IDPi 8g emerged as an optimal catalyst. (−)-Ambrox was obtained in 54% yield and with excellent diastereo- and enantioselectivity (d.r. >20:1; e.r. = 95:5). The IDPi catalyst thus effects the formation of two C–C bonds and one C–O bond in a single reaction step while simultaneously exerting stereocontrol over four stereogenic centres, including one quaternary stereogenic centre. Accordingly, two catalyst motifs under two different conditions (IDPi 8d, HFIP; and IDPi 8g, PFTB) emerged as optimal systems to provide ambrox 2a in high yield, diastereo- and enantioselectivity. A pronounced effect of the 3,3′-substitutent on the diastereo- and enantioselectivity suggests that narrowing the active site of the IDPi catalyst presumably enables preferential binding of (3E,7E)-homofarnesol in its all-trans conformation in line with the Stork–Eschenmoser hypothesis8,9,10. The importance of a confined active site is further showcased by comparison of the IDPi catalyst 8g with highly acidic but structurally less confined N,N′-bistriflylphosphoramidimidate (PADI) catalyst 9 (ref. 53), which gave a complex product mixture (Fig. 2b).

With the optimized conditions at hand, the scalability of the polyene cyclization was evaluated on a decagram scale. After quantitative recovery of the catalyst and purification, 5.2 g of (−)-ambrox (52% yield) with 92:8 d.r. and 95:5 e.r. and partially cyclized products 5 and 10 in 47% yield (Fig. 2c) were obtained. This corresponds to a theoretical volumetric productivity of 296 g l−1 in 20 h reaction time. Current optimized biocatalytic production processes of (−)-ambrox achieve full conversion of 300–450 g l−1 (3E,7E)-homofarnesol in 72 h (refs. 25,26,27,28,29). Several features of the biocatalytic process render it inherently sustainable, such as the use of bio-sourced (E)-β-farnesene produced from sugar fermentation, the use of water as a benign solvent and the convenient isolation of (−)-ambrox, which crystallizes directly from the reaction mixture. The theoretical volumetric productivity of the chemical route developed in this work is comparable to the enzymatic process and is achieved in a shorter reaction time. However, the need for a purification step to separate the catalyst and the side products from (−)-ambrox provides opportunities for further improvement. Owing to the environmental concerns about (per)fluorinated solvents, the solvent was recovered by simple distillation and another gram-scale experiment using recycled PFTB, and the quantitatively recovered IDPi catalyst 8g furnished (−)-ambrox without deterioration of yield and diastereo- and enantioselectivity. The recycling experiments demonstrate that both the catalyst and the solvent can be readily recovered and reused, thereby increasing the sustainability of the approach. Finally, the method was also applied in an asymmetric synthesis of the sesquiterpene lactone natural product (+)-sclareolide, which is a key intermediate in several industrial syntheses of ambrox5,7,54. Subjecting (3E,7E)-homofarnesic acid, which is readily available on a technical scale from (E)-nerolidol54, to the reaction conditions in PFTB furnished the desired lactone natural product 12 in 46% yield, >7:1 d.r. and 94:6 e.r. (Fig. 2d).

Mechanistic studies

To gain insights into the mechanism of the developed polyene cyclization and to investigate the origin of the high chemo-, diastereo- and enantioselectivity, several experiments were conducted. Different mechanistic scenarios for the formation of tricycle 2a from polyene 1a are conceivable. Construction of the tricyclic ring system from the linear polyene could proceed either by a concerted cyclization cascade or an entirely stepwise process by the formation of monocyclic and bicyclic intermediates (Fig. 3a).

Fig. 3: Mechanistic studies.
figure 3

a, Mechanistic scenarios for a concerted and a stepwise pathway. b, 13C KIEs observed at natural abundance in recovered 1a. The experiment was performed in duplicate and the standard deviation is shown. c, Deuterium-labelling study in PFTB-d1 at −40 °C using (S,S)-IDPi catalyst 8g or (S)-PADI catalyst 9 with the respective 2H{1H} NMR spectrum of the tricyclic ether fraction. d, Polyene cyclization of all possible homofarnesol diastereomers to the corresponding tricyclic ethers and rationalization of the obtained products by the respective transition states according to the Stork–Eschenmoser hypothesis.

Inspired by the kinetic isotope effect (KIE) studies at natural abundance to probe the concerted nature of the Diels–Alder cycloaddition55,56, we wondered whether a similar experiment could also provide insights into the nature of the polyene cyclization. KIEs would be expected at all double bond positions in case of a concerted polyene cyclization, as each bond formation would be involved in the rate-determining step. Small but statistically significant 13C KIEs were observed at each double bond of the recovered (3E,7E)-homofarnesol in two independent experiments (96% and 91% conversion at −40 °C) suggesting that protonation and C–C bond formation occur simultaneously (Fig. 3b). Next, deuterium-labelling studies were carried out to further clarify the mechanism of the IDPi-catalysed polyene cyclization and to investigate the selectivity of the initial protonation event. A single deuteration site57 would be expected in case of a concerted polyene cyclization, whereas multiple labelled positions would be characteristic of a stepwise process. Essentially, a single deuteration site (position 3eq) was observed by comparison of the 2H{1H} and 1H NMR spectra of the recovered product 2a from the IDPi-catalysed polyene cyclization at −40 °C (Fig. 3c). This suggests that most of the (−)-ambrox formed at low temperature (−40 °C) in the presence of IDPi 8g is generated by a concerted polyene cyclization. The proportion of (−)-ambrox formed by a stepwise pathway from γ-cyclohomofarnesol 3c (as confirmed by deuterium incorporation at C-17) at −40 °C was estimated to be less than 5% according to 2H{1H} NMR spectroscopy (see Supplementary Information for details). The high selectivity for the equatorial position (3eq) is indicative of a stereospecific antiparallel addition onto the internal olefin, as proposed in refs. 8,9,10. Another important observation is the absence of any side products resulting from protonation or deuteronation of either internal or proximal double bond of (3E,7E)-homofarnesol. In comparison, product 2a isolated from the PADI-catalysed polyene cyclization shows deuterium incorporation at multiple sites, some of which are diagnostic of a stepwise process. Furthermore, bicyclic side product 6 resulting from unproductive protonation of the internal double bond co-eluted with the tricyclic ether fraction and was identified by 1H and 2H NMR spectroscopies. This side product was not observed in the IDPi-catalysed polyene cyclization thus corroborating the high selectivity of the IDPi catalyst for productive protonation of the distal double bond.

The concerted nature of the IDPi-catalysed polyene cyclization was further studied by subjecting all possible diastereomers of homofarnesol 1 to the reaction conditions (Fig. 3d). For each diastereomer, the formation of the respective main isolated product can be rationalized with the Stork–Eschenmoser hypothesis37, in line with a predominantly concerted polyene cyclization. An increase in reaction temperature to −30 °C was necessary to convert (3E,7Z) and (3Z,7Z)-homofarnesol (1c and 1d). We propose that access of sterically congested substrates such as cyclohomofarnesols 3 and (7Z)-isomers of homofarnesol (1c and 1d) to the confined active site is precluded at lower temperatures.

To understand the origins of the different diastereo- and enantioselectivity observed in HFIP and PFTB (Figs. 2a and 4a), the reaction progress, as well as the enantiomeric and diastereomeric ratios, were monitored by gas chromatography and high-performance liquid chromatography (HPLC) over time (Fig. 4b). Whereas the enantioselectivity and diastereoselectivity remained essentially constant (95:5 e.r. and >20:1 d.r. for 2a) in PFTB, an increase of the enantiomeric ratio of product 2a was observed in HFIP on the progression of the reaction (91:9 → 95:5 e.r.). This was accompanied by a decrease in diastereoselectivity (2a:2c > 20:1 → 89:11). Cyclohomofarnesol intermediates 3a and 3c were also detected. The ratio of monocycles 3a:3c changed from approximately 1:2 to 2:1 over the course of the reaction indicating that the kinetic exocyclic double bond product 3c isomerizes to the thermodynamic endocyclic double bond isomer 3a (ref. 42). The enantiomeric ratio of α-isomer 3a was substantially lower (16:84 e.r.) in comparison with that of γ-isomer 3c (1:99), suggesting a kinetic resolution. 5β,8α,9β-Ambrox (2c) was obtained in a similarly high enantiomeric ratio (1.5:98.5) throughout the reaction, indicating its formation by protonation of the axial conformer of 3c. By contrast, the increase in the enantiomeric ratio of ambrox over time can be rationalized by protonation of the equatorial conformer of monocycle 3c and formation of practically enantiopure ambrox, improving the overall enantiomeric ratio. Thus, protonation of both equatorial and axial conformers of essentially enantiopure γ-cyclohomofarnesol 3c is probably responsible for the observed increase in enantioselectivity and concomitant drop in diastereoselectivity. In comparison with the observations in HFIP, the particularly high diastereoselectivity towards 2a observed in PFTB at −40 °C seems to be the consequence of a highly concerted polyene cyclization and suppression of a stepwise pathway. To probe this hypothesis, the reactivity of the monocyclic intermediates was investigated by subjecting the individual isomers (3a, 3b and 3c) to the reaction conditions (Fig. 4c). Solely γ-cyclohomofarnesol 3c showed appreciable conversions towards ambrox isomers 2a and 2c in HFIP, whereas α- and β-isomers 3a and 3b provided less than 5% of tricyclic ethers. When monocycles 3a, 3b and 3c were used as substrates in PFTB at −40 °C, less than 5% conversion was observed. This result is in line with the deuterium-labelling experiment (Fig. 3c), showcasing that protonation of either monocyclic intermediate in a stepwise pathway does not occur to an appreciable extent at −40 °C. At −30 °C, partially cyclized 5 and isomerized 10 were obtained as the main products (see Supplementary Information for details), indicating that the confined active site of IDPi preferentially selects the more accessible double bond in the proximity of the alcohol over the congested cyclohexene. It is important to note that side products 5 and 10 predominantly originate from the protonation of monocyclic intermediates rather than from the undesired protonation of the proximal double bond of (3E,7E)-homofarnesol. The low amounts of 5 and 10 (<5%) observed at −40 °C underline this notion.

Fig. 4: Origin of diastereo- and enantioselectivity, reactivity of monocyclic intermediates and mechanistic proposal.
figure 4

a, Identified intermediates and products in the reaction of 1a to 2a in HFIP or PFTB. b, Yield, enantiomeric excess (e.e.) and diastereomeric excess (d.e.; 2a:2c) over time in HFIP/1H,1H-perfluorooctan-1-ol at −23 °C according to gas chromatography and HPLC analysis. c, Reactivity of cyclohomofarnesols 3 in HFIP and PFTB. d, Mechanistic proposal for the formation of 2a and its diastereomer 2c in HFIP/1H,1H-perfluorooctan-1-ol and PFTB.

On the basis of the mechanistic studies conducted here, we suggest that the key to attaining the observed high diastereo- and enantioselectivity with catalyst 8g is the preference of a concerted reaction pathway over a stepwise process (Fig. 4d). The scalable catalytic asymmetric polyene cyclization of (3E,7E)-homofarnesol to the important ambergris odorant (−)-ambrox reported in this study addresses a longstanding challenge in chemical synthesis and provides the desired product in unprecedented stereoselectivities. We believe that the method will find widespread application in related polyene cyclizations and may expedite the asymmetric synthesis of natural products, odorants and pharmaceuticals from readily available achiral starting materials.