On our way towards a metal-free total synthesis of (−)-Δ⁹-tetrahydrocannabinol (4, Figure 2) we developed a convenient organocatalytic Diels–Alder route to generate the desired tricycle 5 of a model-system that also represents an alternative approach to key-intermediate 6 in Danishefsky’s total synthesis of Dynemicin A (3) [14].

Our strategy, using the [4 + 2]-cycloaddition to obtain the cannabinoid tricycle system, is able to employ a variety of dienophiles and, hence, provides a versatile entry to this significant group of naturally occurring compounds [18], whereas the achieved Diels–Alder products can be further modified to gain a variety of heterocycles. Most published strategies are based on the use of a preexisting cyclohexane moiety in the starting material to form the heterogenic tricycle [19–28].

In former experiments of our group [29–31] it became apparent that without a catalyst no conversion towards the desired product occurred. Only the homodimer of the diene 10 could be isolated. Metal-based Lewis acids (e.g., catechol boronates) proved to be inefficient or too reactive. Due to the fact that other groups have used thiourea-derivatives as successful catalysts [32], we decided to test, on a model system, whether these catalysts could increase the reaction rate based on specific hydrogen bonds. In our initial screening we used the proven Schreiner catalyst 1,3-bis(3,5-bis(trifluoromethyl)phenyl)thiourea, which gave a satisfying 60% yield. In order to optimize the yield and to introduce chiral elements, we screened a number of analogues. These thiourea catalysts 9a–l were easy to obtain from the corresponding isothiocyanates 7a,b and chiral amines 8a–f in a one-step synthesis (Scheme 1) [33]. The yields were good to excellent in all procedures (Table 1).

These thioureas were used in an intermolecular Diels–Alder reaction of diene 10 [36] with acrolein (11) to obtain cannabinoid tricycle 5 shown in Scheme 2.

In all cases we only achieved one cis-diastereomer and the carbonyl function was always in the 10-position. The reaction was carried out at low temperature and high dilution to avoid the dimer formation of diene 10. An increase of the temperature to 40 °C resulted in a higher conversion but unfortunately no enantioselectivity was observed. We obtained good to excellent yields from 68% to 99% depending on the substitution of the thiourea-catalyst 9a–l (Table 2).

When the substitution of the various thioureas 9a–l is further compared with the obtained yields in the Diels–Alder reaction (Table 2), the following tendencies are also observed. A noticeable fact is that, in contrast to the thioureas with a cyclopropyl- and m-halophenyl-moiety (Table 2, entries 7–9), the corresponding bis(trifluoromethyl)thioureas (Table 2, entries 16–18) provide a higher conversion.

Next to the previously described thiourea catalysts, we also analyzed iminium-ion catalysts according to MacMillan on various model systems (Figure 3) [37]. The effect of these catalysts is based on the formation of iminium ions by condensation of the dienophile bearing a carbonyl group, with the sterically hindered imidazole catalyst. In this way, one side is shielded and only the other side can be attacked by the diene. Hence, reaction rate and asymmetric induction are increased.

The synthesis of imidazolidinone catalysts [38] is the premise for the construction of the demanding amide 13, which was afforded by the reaction of (S)-phenylalanine methyl ester hydrochloride (12) with methylamine (Scheme 3). In the second step of the synthesis, amide 13 was cyclized with different aldehydes 14 and addition of catalytic amounts of FeCl₃ at high temperatures into (2S,5S)- or (2R,5S)-imidazolidinone 15 [39–40].

To gain electron-poor catalysts for the degradation of the dienophilic LUMO in the Diels–Alder reaction and to consequently increase the reaction rate, we used aldehydes with electron-withdrawing groups in ortho-, meta- and para-positions (Scheme 3, Table 3). Reaction with p-nitrobenzaldehyde gave compound 16 in 66% yield (Figure 4).

The organocatalytic species 15 were synthesized in moderate yields. A conspicuous feature is that the yields of the (2S,5S)-derivatives, with the exception of the tert-butyl substituted catalyst 15a (Table 3, entries 1–2), are always higher than those of the corresponding (2R,5S)-components (Table 3, entries 3–12). The configuration of the afforded imidazolidinone catalysts 15 could be confirmed by NOESY experiments and from their X-ray structures (Figure 5).

Before testing the catalysts in the Diels–Alder reaction, we analyzed a few co-catalysts with the commercially available imidazolidinone catalyst 15h. A pentyl-substituted tricycle was used as a model system (model system II) 19 (Scheme 4, Table 4).

Hydrochloric acid has been proven to be the best co-catalyst (Table 4, entry 1), providing cis-19 in good yield and with an enantiomeric excess of 96%. Even perchloric acid and trifluoroacetic acid gave high enantiomeric excesses but in much poorer yields (Table 4, entries 2 and 3). Using para-toluenesulfonic acid also afforded cis-19 in good yield, but its enantiomeric excess could not be determined (Table 4, entry 4). Trifluoromethanesulfonic acid as co-catalyst only led to decomposition (Table 4, entry 5).

Under these optimized reaction conditions, we performed a catalyst screening with the previously synthesized imidazolidinone catalysts 15 to study their asymmetric induction (Scheme 5). As starting material we used compound 17 with the THC-typical pentyl side chain. The results are summarized in Table 5.

The used organocatalysts, having a (S)-configured stereocenter in the 5-position, should prefer the formation of opposite enantiomers, compared to the commercially available (2R,5R)-15h. This is indicated by negative enantiomeric excess (Table 5). Due to low yields of diastereomer trans-19 and the less successful determination of their enantiomeric excess, Table 5 only contains results for cis-19. Except for in the case of 15a, we observe tendencies such as the achievement of higher yields through the use of (2S,5S)-configurated imidazolidinones 15 compared to the yields afforded by (2R,5S)-configurated catalysts 15. There is an opposite trend for the enantiomeric excess, which is also related to the steric hindrance of the phenyl substituent, i.e., the smaller the substituent, the lower the enantiomeric excess. Application of the known tert-butylimidazolidinone catalyst 15a provides the highest yield (66%) and enantiomeric excess (98% ee) with its (2S,5S)-derivate (Table 5, entry 3).