Condition optimization. We began our study by optimizing the reaction conditions using enynamide 1a and α-diazoketone 2a as the model substrates. To avoid the possible decomposition of diazo compounds and chiral N-oxide catalysts by the Au catalyst, a rational stepwise, one-pot procedure was chosen for this cascade reaction (please see the details on the effects of varying the procedure in Supporting Information). Upon irradiation with 6 W blue light-emitting diodes (LEDs), the designed cascade reaction catalyzed by 5 mol% of cationic gold and 20 mol% of chiral N-oxide OC1 proceeded very well, affording the desired product, 3aa, smoothly (Table 1, entry 1). To further improve the stereoselectivity, other chiral catalysts, OC2-OC4, were examined, although the results were not better than that obtained using OC1 (Table 1, entries 2-4). Routine screening of solvent and temperature were carried out (Table 1, entries 5-11), and the reaction performed in CHCl3 at -60 °C gave much improved enantio- and diastereoselectivity (Table 1, entry 11: 93% yield, 86% ee, and 15:1 dr). Furthermore, introducing an ethyl group to the 2-position of pyridine in chiral N-oxide catalyst OC1 afforded superior results in terms of both reaction efficiency and stereoselectivity (Table 1, entry 12: 94% yield, 93% ee, and >19:1 dr). The structure of chiral catalyst OC5 and absolute configuration of product 3aa were previously established via single-crystal X-ray diffraction analysis. For comparison, several classic chiral tertiary amines, NHC, and phosphine catalysts were also examined, but no better results were not obtained for this asymmetric cascade reaction (please see the details in Supporting Information).
Substrate scope. After establishing the optimized conditions, we examined the generality of this biomimetic cascade reaction. Remarkably, various conjugate enynamides 1 were compatible with the combined photoactivation and relay catalysis system (Table 2). Regardless of the electron-donating and electron- withdrawing substituents on the phenyl group at the alkynyl unit (R1), enynamides 1a-h smoothly reacted with α-diazoketone 2a, yielding the corresponding chiral annulated furans, 3aa-3ah, in 90-94% yield, 92-95% ee, and >19:1 dr. 2-Naphthyl- (1i) and 3-thienyl-substituted (1j) enynamides were also suitable, affording the corresponding adducts in high yields with high stereoselectivities (3ai: 92% yield, 94% ee, and >19:1 dr; 3aj: 88% yield, 94% ee, and >19:1 dr). In addition to enynamides with aryl or heteroaryl substituents, those with alkyl substituents (i.e., n-Bu, c-hexane, and t-Bu), 1k-m, were also suitable for this combined photoactivation and relay catalysis system, delivering chiral bicyclic lactams 3ak-3am with good results (82-87% yield, 91-95% ee, and 17:1 dr). In parallel, we examined the effect of varying the alkenyl unit in enynamide 1 (R2). A wide range of enynamides bearing electron-donating (3an, 3ao, 3as, and 3at) or electron-withdrawing groups (3ap-3ar) at the para position of the benzene ring were suitable for this transformation (89-94% yield, 90-95% ee, and >19:1 dr). Substitution at the meta position was also tolerated, having no significant effect on the reaction efficiency and stereoselectivity (3at: 90% yield, 95% ee, and >19:1 dr; 3au: 93% yield, 90% ee, and >19:1 dr). Enynamides having a 2-naphthyl, 2-thienyl, or cyclohexyl substituent at the alkenyl unit were also suitable and were converted to the corresponding products with good outcome (3av-3ax: 84-93% yield, up to 94% ee and >19:1 dr).
Subsequently, we probed the scope of α-diazoketones 2 for the present transformation. As summarized in Table 3, a variety of α-diazoketones could participate in this biomimetic cascade reaction well. Electronically varied substituents (e.g., Me, Cl, Br, and CO2Me) at the para position of the benzene ring were compatible, providing chiral products 3ba-3ea in 91-95% yield, 92-98% ee, and >19:1 dr. Varying the substitution position had a minimal effect on the reaction yield and enantio- and diastereoselectivity (3fa–3ia: 85-91% yield, 90-93% ee, and >19:1 dr). Replacement of the methyl group with an ethyl or i-Bu group delivered the corresponding products, 3ja (83% yield, 93% ee, and >19:1 dr) and 3ka (87% yield, 92% ee, and >19:1 dr) successfully. Functional groups at the alkyl side, such as OBn and alkynyl, were also well tolerated in this combined photoactivation and relay catalysis system, delivering highly functionalized bicyclic products 3la and 3ma in 86% yield, 92% ee, and >19:1 dr and 78% yield, 95% ee, and >19:1 dr, respectively.
Synthetic transformation. We then performed a series of experiments to prove the utility of this new methodology using chiral bicyclic lactam 3aa as an example (Fig. 3). First, treatment of 3aa with m-chloroperoxybenzoic acid (m-CPBA) in dichloromethane at 0 °C yielded a new chiral lactam, 4, in 90% yield through a furan epoxidation and ring opening cascade52. Interestingly, the oxidative rearrangement of 3aa was observed under similar oxidization conditions followed by base treatment, producing an interesting spirocyclic product, 5, in 81% yield36. Subsequently, the N-tosyl group could be removed by treatment with SmI2/hexamethylphosphoramide (HMPA) in tetrahydrofuran at -78 °C to give product 6 in 87% yield53-54. Finally, the amide unit of 3aa could be reduced with LiAlH4 to give the corresponding amine 7 in 95% yield with the enantiomeric excess of the product retained55. Notably, almost no loss of optical purity was observed during all four transformations.
Density functional theory (DFT) calculations. To understand how the chiral N-oxide catalyst, OC5, facilitates the annulation between ketene int. I and aza-o-quinone int. II and to probe the origin of stereoselectivity, DFT calculations at the B3LYP-D3(BJ)/6-311+G(d,p)-SMD(CHCl3) level were carried out based on the optimal experimental results (see SI for computational details). We first investigated the concerted Diels-Alder-like pathway (path b) depicted in Fig. 2c. However, only the C–C bond was formed at the transition state (see Fig. S4 for the structure of transition state), indicating that the reaction instead proceeds via a stepwise mechanism (path a, Fig. 2c). The molecular orbital analysis showed that the nucleophilic attack of OC5 activates the ketene int. I. Its highest occupied molecular orbital is not only raised in energy, but also polarized with a larger contribution from the p orbital of the tri-substituted C atom, which promotes C–C bond formation (see Fig. S4 for details). The calculated relative free energy profile of the stepwise mechanism is shown in Fig. 4a. Catalyst OC5 first coordinates to int. I via TS0 to generate activated nucleophile int. III which further undergoes Michael addition, and the C–C bond is formed in the resulting intermediate, int. IV. The reaction further progresses through the concerted addition–elimination of the carbonyl group, comprising simultaneous C–N bond formation and C–O bond cleavage, to release the product and catalyst OC5. Interestingly, a hydrogen bond between the sulphonyl O atom of the imine and amide H atom of OC5 was observed during the reaction, hinting at the vital role of the amide group of the catalyst in controlling stereoselectivity. Unlike the pathway leading to the experimentally observed major product with RR configuration, the other three pathways had relatively high-lying transition states in the last step. The difference between the energy barriers of pathways leading to products with different configurations (ΔΔG‡ = 3.8 kcal/mol between TS0-A and TS2SS and ΔΔG‡ = 2.2 kcal/mol between TS0-A and TS2RS) was in good accordance with the excellent enantioselectivity and diastereoselectivity achieved in the experiments.
Because the origin of stereoselectivity is of primary interest to the mechanistic analysis, further distortion–interaction analysis was performed to shed light on the factors destabilizing the addition–elimination transition states (for detailed distortion–interaction analysis, see Table S6). Given that the products of TS2RR and TS2SS were the same in energy, we analyzed these transition states using the same energy reference. As shown in Fig. 4b, OC5 enters a less hindered space to interact with the carbonyl group in TS2RR, as suggested by the Felkin–Ahn model. On the other hand, it approaches the carbonyl group via a much more congested trajectory in TS2SS. This causes the elongation of the C–C bonds in the more strained six-membered ring (see Fig. S5 for details), leading to a larger distortion energy of the substrate fragment (ΔΔE(sub-dist) = 7.1 kcal/mol) and eventually, the increase in energy of TS2SS. Careful comparison of the geometries of the substrate fragments (Fig. 4c) showed that in TS2RR, the methyl group points toward the phenyl ring of the sulphonyl group, indicating a C-H–π interaction. In contrast, this weak interaction is absent in TS2RS, which has a larger substrate distortion energy (ΔΔE(sub-dist) = 2.7 kcal/mol). In summary, the relatively rigid chiral catalyst, OC5, confines the substrate via O coordination and hydrogen bond to the RR configuration with not only less ring strain, but also weak interaction, which assures good stereoselectivity.