Main

Bridged bicyclic cores are widely utilized in medicines and drug carriers owing to their high bioaffinity1,2; however, the diversification of such scaffolds remains challenging because of limited access to variously substituted cyclic 1,3-dienes (Fig. 1a). In synthesis and catalysis, such strained bicyclic molecules often display unique reactivity; for example, various norbornenes mediate the Catellani reaction3 and many transition-metal-catalysed meta-C–H functionalization reactions4. Moreover, they serve as effective monomers for bicyclic polymers, finding applications in optical materials and shape-memory materials5. Diels–Alder [4 + 2] cycloaddition has been recognized as one of the most effective methods for synthesizing bridged bicyclic compounds since its discovery in 19286,7,8. Hence, developing different methods to access diverse cyclic 1,3-dienes for Diels–Alder reactions has attracted significant interest9 (Fig. 1b). While a previous report indicated that allylic C–H activation could convert mono-alkenes to 1,3-dienes for in situ Diels–Alder reactions (Fig. 1c), the lack of powerful catalysts capable of controlling selectivity limited its scope to acyclic terminal alkenes10. Deploying multiple C–H functionalizations to transform saturated four-carbon units into 1,3-dienes for formal Diels–Alder reactions remains a challenging goal. Unlocking the reactivity of the multiple C–H bonds in aliphatic substrates in one system is required to reduce this approach to practice. The chemo-, regio- and stereoselectivity challenges generated by this complex multi-step cascade reaction system demand powerful catalytic systems. Recent successes of developing [3 + 2] annulation11,12 and [2 + 2] annulation13 reactions via stitching two C(sp3)–H bonds with olefins or aryl halides14 to construct cyclic motifs prompted us to explore the possibility of stitching four C(sp3)–H bonds with olefins to forge six-membered carbocycles. Here, we report a Pd(II)-catalysed formal [4 + 2] Diels–Alder reaction of saturated cyclic carboxylic acids and olefins. With more than 576,000 cyclic carboxylic acids commercially available across different ring sizes and substituent patterns, this method significantly expands the scope of current Diels–Alder reactions for constructing bridged bicyclic scaffolds. Our bifunctional pyridine–pyridone ligand plays a crucial role in enabling C–H functionalization reactivity while controlling both the chemo- and regioselectivity of this multi-step cascade reaction (Fig. 1d). The ligand-promoted allylic decarboxylation pathway outcompeted the common vinyl C–H and allylic C–H activations of the olefin intermediate (side products B and C), delivering the decarboxylated allyl–Pd intermediates10,15. Density functional theory (DFT) calculations suggest that the regioselectivity of subsequent β-H elimination of the allylic palladium is also controlled by the ligand, affording the desired diene species for the Diels–Alder reaction. Notably, overall oxidation of the C–H bonds is furnished by the Pd(II) catalyst using a practical oxidant Cu2(OH)2CO3.

Fig. 1: C–H activation enabled Diels–Alder reaction of saturated carboxylic acid.
figure 1

a, Bicyclic molecules in research and industry. b, Unsaturated substrates for Diels–Alder reaction. c, Transition-metal-catalysed C(sp3)–H functionalization. d, C–H activation enabled Diels–Alder reaction of saturated carboxylic acid (this work). DG, directing group; FG, functional group.

Full size image

Although dehydrogenation of aliphatic acids has been demonstrated with acyclic chains to produce dienes, forming dienes—especially cyclic dienes—in the presence of an interfering dienophile remains a significant challenge. Our exploratory studies began by testing the reaction of commercially available α-methyl-cyclohexanecarboxylic acid as a potential diene surrogate with maleimide as the dienophile. Pd(II) catalyst coordinated with a six-membered pyridone-oxazoline ligand enabled the decarboxylative Diels–Alder product in 44% isolated yield (Supplementary Table 1). The resulting product, bicyclic succinimide, has wide applications in biochemistry and as a medicinal carrier (Fig. 1a). Building on recent advancements in Pd-catalysed methylene C–H activation16,17, we evaluated a series of bidentate pyridone ligands. Both six-membered and five-membered bidentate pyridone ligands were effective for this transformation, with the five-membered ligands demonstrating higher activity than their six-membered counterparts (Supplementary Table 1). During oxidant screening, we found that the classical silver additives such as Ag2CO3 were consistently effective for this reaction. Surprisingly, copper oxidants such as Cu2(OH)2CO3 are also found to effectively replace expensive silver complexes as oxidants, delivering the Diels–Alder product. After extensive optimization of reaction parameters, the five-membered bidentate chelating pyridone–pyridine ligand L10 was identified as the most effective, yielding the bicyclic product 2a in 76% (Tables 1 and 2).

Table 1 Substrate scope of cyclic carboxylic acids
Full size table
Table 2 Substrate scope of bioactive molecules and dienophiles
Full size table

We began exploring the scope of this transformation by applying our optimized conditions to six-membered cyclic carboxylic acids, which transformed cyclohexanecarboxylic acids to bicyclo[2.2.2] derivatives (Table 1). Satisfyingly, this method proved to be compatible with α-alkyl substituted carboxylic acids (2a Me, 2b Et, 2c nPr, 2d iBu), despite the potential reactivity of the methylene C–H bonds on the alkyl chain. Carboxylic acids bearing additional cyclic alkyl rings were also well tolerated, providing the corresponding cycloadducts in good yields (2e and 2f). Furthermore, carboxylic acids with phenyl and benzylic groups at the α-position (2g, 2h and 2i) were converted into their Diels–Alder products in high yields, which is particularly remarkable given the general ease of cleaving C(sp²)–H bonds compared with C(sp³)–H bonds. In addition, carboxylic acids bearing a variety of other functional groups, including halogens (2j and 2k), –OMe (2l), –CF3 (2m), –OPh (2n) and –OTBS (2o), were effectively converted under this methodology. Surprisingly, the α-CF3-substituted cyclohexanecarboxylic acid (2p) performed well in this system, despite the typical preference of Diels–Alder reactions for electron-rich dienes when using electron-deficient dienophiles. We next explored whether this method could be applied to cyclic carboxylic acids substituted at other positions on the ring. Indeed, both β- and γ-substituted (2q and 2r). Cyclic carboxylic acids were smoothly converted to the corresponding adducts. Notably, a δ-substituted cyclohexanecarboxylic acid also yielded a Diels–Alder product with a 45% yield (2s), despite the δ-C–H bond being a methine C(sp3)–H bond rather than a methylene C(sp3)–H bond. Subsequently, we applied this methodology to five-membered cyclic carboxylic acids. In our system, most substituted cyclopentanecarboxylic acids were successfully reacted to give the expected bicyclo [2, 2, 1] products (3a–3k). In the case of 3a, no unreacted cyclopentanecarboxylic acid was detected after the reaction. The α,β-dehydrogenative product may be generated in the first step, consuming the starting material. Moreover, seven-membered cyclic carboxylic acids were also amenable to this method and afforded the corresponding bicyclo [2, 2, 3] adducts in moderate yields (4a–4d).

Succinimide bicyclic rings are recognized as a class of bioactive compounds in biochemistry and medicinal research. So far, research has been primarily restricted to modifying the N-substitution of maleimide to evaluate bioactivity, the modification of bicyclic cores are limited18,19. We thus investigated whether this methodology enables the preparation of a series of bioactive molecules with modifications to both the bicyclic core and the N-substituent (Table 2). For example, derivatives of noreximide (3fa), a sedative approved by the US Food and Drug Administration, could be easily prepared using this approach. In addition, N-aryl succinimides, a known series of androgen receptor antagonists, could also be synthesized through this method (3fc and 3fd). This methodology also tolerates complicated maleimide–drug linkers. Maleimide conjugates with fluorescein (3da), tocainide (3db), mexiletine (3dc), sulfamethoxazole (3dd) and tryptamine (3de) are compatible, yielding the corresponding bicyclic products. We also tested the reactivity of other dienophiles. Fumarate (5), dicarbonyl ethylene (6) and maleic anhydride (7) successfully reacted to give the corresponding Diels–Alder products. Interestingly, when using terminal alkenes such as the more representative vinyl ketone (8) and acrylate (9) as dienophiles, only the single bicyclic regioisomer was formed due to the steric effects. Notably, the optimal reaction conditions for these dienophiles differ significantly from those for maleimide, as is expected given that different dienophiles generally require distinct reaction conditions in Diels–Alder reactions. In some cases (such as 6, 8 and 9), the reaction rate of cyclic conjugated dienes with these dienophiles is slower than the polymerization of the cyclic conjugated diene, which leads to moderate yields of the final Diels–Alder products.

To showcase the diverse chemical space now easily accessible through this method, we subjected the Diels–Alder products to further transformations, taking advantage of the functional handles retained in the products (Fig. 2). For example, full reduction of the carbonyl groups yields the bicyclic pyrrolidine (10), whereas partial reduction by NaBH₄ produces the hydroxyl γ-lactam as a single isomer (11), a structure commonly found in natural bioactive compounds20. Epoxidation and hydrogenation can also be used to further modify the bicyclic compounds, leading to higher-order multi-cyclic products (12 and 13).

Fig. 2: Synthetic application.
figure 2

Chemoselective reduction of imide moiety and functionalization of double bond via epoxidation or hydrogenation. r.t., room temperature; m-CPBA, meta-chloroperoxybenzoic acid.

Full size image

Exclusive chemo- and regioselectivity observed across all substrates prompted us to further investigate the mechanism and origin of selectivity. Based on previous reports and our control experiments, we propose a three-step cascade reaction pathway, initiated by ligand-enabled β,γ-dehydrogenation of cyclic carboxylic acids to afford Int-A (Fig. 3a), followed by a ligand-enabled concerted decarboxylation process to deliver the diene species Int-C, with exclusive chemo- and regioselective control. The transient diene species engages in a rapid Diels–Alder process smoothly to afford the endoproduct, which is promoted by hexafluoroisopropanol (HFIP) solvent21 (for details of control experiments carried out, see Supplementary Table 5).

Fig. 3: Mechanistic study and DFT calculation.
figure 3

a, Proposed mechanism: palladium-catalysed cascade chemo-/regioselective dehydrogenation–decarboxylation–dehydrogenation for cyclic conjugated diene formation. b, DFT calculation for concerted decarboxylation: energy barriers of concerted decarboxylation step and visualization of the related transition states (bond distance in Å).

Full size image

DFT calculation was carried out to provide insights regarding the control of chemo- and regioselectivity. After the β,γ-dehydrogenation, our ligand could recruit HFIP molecule(s) with hydrogen bonding, which facilitates ligand-to-solvent charge transfer and thus stabilizes the transition state via electrostatic interactions, thereby outcompeting the known allylic or alkenyl C–H activation processes10,13 (Fig. 3b). The η³-allyl Pd complex (Int-B) generated is identified as the key intermediate for controlling regioselectivity. While this η³-allyl Pd intermediate (Int-B) could be transformed into two less-stable η1-allyl Pd intermediates (Int-B1 and Int-B1′), the overall energy barrier of cleaving adjacent C–H bonds could be differentiated by the ligand (23.2 kcal mol−1 compared with 29.6 kcal mol−1), leading to the observed exclusive regioselectivity (Fig. 3a). We propose that the regioselectivity is primarily controlled by steric interactions arising from the Pd coordination and ligand environment (for detailed discussion of how DFT helped us further understand the chemo- and regioselectivity, see Supplementary Section 6 and Supplementary Figs. 1–3).

In conclusion, we have developed a catalytic system using saturated aliphatic acids as surrogates for cyclic dienes, which undergo Diels–Alder reactions in tandem. The entire cascade reaction involves a ligand-enabled triple C(sp3)–H functionalization and a palladium-catalysed concerted decarboxylation process. Mechanistic studies provide a possible explanation for the excellent chemo-, regio- and endoselectivity.

Methods

General procedures of synthesis bicyclic compounds: To an oven-dried 5-ml screw-capped vial equipped with a stirring bar was added the corresponding palladium catalyst (0.01 mmol, 10 mol%), ligand (0.02 mmol, 20 mol%), copper additive (0.11 mmol, 1.1 equiv.), maleimide (0.2 mmol, 2.0 equiv.) and the corresponding cyclic carboxylic acid (0.1 mmol, 1.0 equiv.). Then, 1 ml of solvent was added. Subsequently, the vial was capped and closed tightly. The reaction mixture was then stirred at a rate of 240 rpm at 130 °C temperature for 18 h. After cooling to room temperature, the mixture was passed through a pad of silica gel and Celite using ethyl acetate as the eluent to remove insoluble precipitates. The resulting solution was concentrated, and the residue was purified by preparative thin-layer chromatography to afford the decarboxylative Diels–Alder reaction product.