Palladium-catalyzed regio- and enantioselective migratory allylic C(sp3)-H functionalization

Transition metal-catalyzed asymmetric allylic substitution with a suitably pre-stored leaving group in the substrate is widely used in organic synthesis. In contrast, the enantioselective allylic C(sp3)-H functionalization is more straightforward but far less explored. Here we report a catalytic protocol for the long-standing challenging enantioselective allylic C(sp3)-H functionalization. Through palladium hydride-catalyzed chain-walking and allylic substitution, allylic C-H functionalization of a wide range of acyclic nonconjugated dienes is achieved in high yields (up to 93% yield), high enantioselectivities (up to 98:2 er), and with 100% atom efficiency. Exploring the reactivity of substrates with varying pKa values uncovers a reasonable scope of nucleophiles and potential factors controlling the reaction. A set of efficient downstream transformations to enantiopure skeletons showcase the practical value of the methodology. Mechanistic experiments corroborate the PdH-catalyzed asymmetric migratory allylic substitution process.

In contrast, the corresponding allylic C−H functionalization ought to be more straightforward and economic but is far less developed 9,10 . In general, the key ƞ 3 -π-allyl metal intermediate is generated through direct allylic C−H cleavage (Fig. 1a) 10 . Elegant studies from Rainey 11 , Trost 12,13 , Gong [14][15][16][17][18][19] and White 20,21 groups et al. have realized the C−H functionalization of the allylic carbon center that usually requires further activation by another vicinal sp 2 carbon or heteroatom unit. A few studies have been performed on unactivated allylic C−H bond but generally with moderate enantioselectivities [22][23][24][25] . Therefore, efficient strategies are still highly desired for enantioselective allylic C−H functionalization, especially for functionalization with inert allylic C−H bonds.
Here, we show that the merger of PdH-catalyzed chain-walking and ƞ 3 -allylation serves as an efficient route to the stereoselective activation of the inert allylic C(sp 3 )−H bond. A series of flexible acyclic nonconjugated dienes undergo the C(sp 3 )−H allylation in high yields and enantioselectivities (up to 93% yield, generally 90−96% ee) and 100% atom-efficiency. Mechanistic studies further provide evidence for the designed migratory allylation strategy.

Results
Investigation of reaction conditions. We initiated the migratory allylic functionalization by studying reactions with an acyclic non-conjugated diene 1a as the pre-electrophile, diethyl malonate 2a as the pronucleophile, and Et 3 N as the base with a palladium catalyst (Fig. 2). A group of chiral ligands were first evaluated for their potential to facilitate both the chain walking and allylic substitution. Trost ligand L1 which was effective in Pd-catalyzed allylation, failed to furnish the target product 3a (entry 1) 67 . Trace amounts of 3a were observed with bisphosphine ligand L2 in low enantioselectivity, and L3 did not provide any product (entries 2−3). Because oxazoline-type ligands have been widely adopted in Pd-catalyzed chain-walking reactions [36][37][38][39][40][41] , we turned to L4 and L5 for the possibility of migratory allylic substitution. However, none of them led to the desired allylation product 3a (entries 4−5). Then a set of JosiPhos-type chiral ligands were further evaluated due to their previous application in Pdcatalyzed asymmetric hydrofunctionalization of conjugated dienes (entries 6−9) 58,59 . Fortunately, around 20% yield of 3a was observed for all cases, and the enantioselectivities were achieved up to 97:3. The reaction yield was finally raised to 82% when Et 3 N was used as both the base and solvent, though many other solvents also benefited the yield (entries 10−13). Thus, the optimal condition for migratory allylic functionalization was identified as the combination of the skipped diene 1a (1.1 equiv), nucleophile 2a (1.0 equiv), [Pd(allyl)Cl] 2 (2.5 mol%), L9 (5 mol%) and NaBAr F 4 (5 mol%) in Et 3 N at 60 o C for 24 h.
Diverse dicarbonyl derivatives were subsequently tested for migratory allylic functionalization (3w−3z). Dimethyl malonate 2b coupled smoothly with different nonconjugated dienes in good yields and high enantioselectivities without regioselectivity issue 60 (3w, 3x). α-Ketoester 2c was found to be an effective nucleophile, delivering 3y in 68% yield and 97:3 er, albeit with low diastereoselectivity because of the facile epimerization of the α-carbon center of the dicarbonyl unit. 1,3-Dione 2d with a lower pK a showed moderate reactivity and yielded allylation product 3z in 84:16 er, suggesting that the migratory allylation might be suitable for nucleophiles with a variety of pK a 57 .
The utility and robustness of the present protocol were also demonstrated by conducting C(sp 3 )−H functionalization with a series of dienes derived from complex molecules. Dienes tethered to drugs, such as naproxen, oxaprozin, isoxepac, and febuxostat, efficiently underwent the migratory allylation process to furnish the corresponding products in 57−90% yield with 97:3−98:2 er (4e, 4f, 4i, 4j). Electrophiles bearing a Mosher ester, bridged-or macrocycles, camphanic acid, or ribofuranose derivatives reacted in moderate to good yields with high enantioselectivities (4b−4d, 4g−4h). Notably, all the aforementioned cases underwent migratory allylation with exclusive regioselectivity at the target allylic carbon center. However, tertiary nucleophiles and a couple of other substrates turned out to be ineffective for this transformation (see Supplementary Methods for details).
To uncover the pK a range of nucleophiles and the factors controlling reactivity, a series of nucleophiles with pK a ranging from 9 to 24 were evaluated for migratory functionalization reactions (Fig. 4a). The secondary carbon nucleophiles were chosen and evaluated in order to exclude the steric hindrance as a potential interference factor. As the present protocol contained two processes, i.e., chain walking and allylic substitution, alkene 5, which could only undergo chain walking, was adopted to test the walking potential of a chosen nucleophile under standard conditions. Using a nucleophile (2e−2h) with a pK a < 13 resulted in olefin 5 smoothly undergoing chain-walking to produce 6 in 46−78% yield, but little allylation was observed. By contrast, gradually increasing the pK a of the nucleophiles from 18 to 24 (2i−2k), resulted in almost no chain-walking or allylation. Therefore, nucleophiles with pK a between 13 and 18 appear to be suitable for target migratory allylation (2a−2d). Our data suggest that the migratory allylation for nucleophiles with pK a < 13 might be limited by insufficient nucleophilicity, whereas that for nucleophiles with pK a > 18 does not lead to chain-walking because of insufficient acidity to generate the palladium hydride catalyst.
The mixtures of stereoisomers and regioisomers of skipped dienes underwent regioconvergent migratory allylation in moderate yield and high enantioselectivity (Fig. 4b). To uncover the factor leading to the decreased reactivity, diene 7 or 8 was used as substrate individually under standard condition. The conjugated diene 8 underwent hydrofunctionalization smoothly, giving 3a in good yield and high er, though 27% of 8 remained unreacted. In contrast, the internal skipped diene 7 showed very low reactivity, providing 3a in only 4% yield but with high er. We proposed that the coordination of palladium catalyst with the diene substrate might be a crucial factor controlling the results. It should be easier for palladium catalyst to initiate the reaction by coordinating with diene substrate 1a bearing a terminal olefin than 7 and 8 with sterically bulky internal olefin units. The higher reactivity of 8 than that of 7 was possibly resulted from the quick consumption of the stable π-allyl-Pd intermediate from the migratory insertion of a conjugated diene with Pd−H catalyst.
Transformations. The synthetic value of the migratory allylation was highlighted by a gram-scale test and a series of downstream transformations (Fig. 4c). A total of 2.2 gram of compound 3w were obtained in high yield and enantioselectivity under standard conditions. This compound was then converted to a set of important skeletons in short steps. A combination of decarboxylation and oxidation was used to efficiently prepare γ-lactones 11 and 12 with multiple stereogenic centers, as demonstrated by X-ray crystal analysis. Enantiopure cyclopentene 14 was easily constructed through a metathesis reaction. Moreover, Au(I)-catalyzed [2 + 2] cycloaddition 68 enabled the enantioselective construction of complex bicyclo-[3.2.0] motif 16, which exists widely in biologically active molecules 69 .
Mechanistic studies. A set of mechanistic studies were conducted to further elucidate the mechanism of the present protocol (Fig. 5). Quenching the model reaction after only 3 h produced compounds 7 and 8 via alkene migration in 6% and 13% yields, respectively, along with a 25% yield of product 3a (Fig. 5a). This fact supports that chain-walking is involved in the reaction, which is further corroborated by the results of deuteration studies showing D labeling of product d-3a at multiple carbon centers (Fig. 5b). Moreover, a crossover experiment using deuterated d-1a and non-deuterated diene 9 as competing electrophiles led to the detection of d-4e (Fig. 5c). The transfer of D atoms from d-1a to d-4e suggests that the PdH catalyst might dissociate from the ligated diene substrate throughout the chain walking process. Considering the very trace deuteration observed in d-4e, we deduced that PdH dissociation from the diene substrate might be a challenging and negligible process 38 .
Kinetic isotope effect (KIE) was determined to be 1.04, implying that the PdH formation or chain walking process involving hydrogen bond formation or breaking might not be the rate-determining step (Fig. 5d). Further kinetic studies uncovered that the migratory allylation reaction was first order in both [cat] and [2a] (Fig. 5e). However, the kinetic data of diene 1a did not suggest a simple first or zero reaction order, but a saturation kinetics, fitting Michaelis−Menten dynamic model which is commonly used to describe enzymatic reaction 70 . This observation was further supported by a linear relationship between rate −1 and [1a] −1 (Fig. 5e). In order to corroborate the aforementioned facts, we hypothesized the final allylic substitution was the ratedetermining step, and deduced the corresponding initial rate equation (see Supplementary Methods for details). This rate law was consistent with all of the observed kinetic data. Thus, the rate-determining step in the catalytic cycle was determined as the allylic substitution.
In the kinetic studies, an obvious induction period was detected. To elucidate the origin of the induction period, compound 17 from the reaction of nucleophile 2a with [Pd(allyl)Cl] 2 was identified (Fig. 5f). The corresponding initial reaction profile of 17 and product 3a clearly showed that the yield of 17 increased along with the slow generation of 3a at the beginning of the reaction. When 17 reached close to the highest 4−5% yield, the reaction rate for 3a arrived at an obviously higher level and maintained stable. Therefore, the induction period derived from the slow formation of active Pd(0) catalyst via offcycle allylation.
Proposed mechanism. Based on the above facts and reported work on the studies of chain walking 36-41 , a plausible mechanism was proposed as shown in Fig. 6. The proton would first oxidatively add to Pd(0) to generate PdH species in situ which then inserted into the less sterically hindered alkene to give an unstable alkylpalladium intermediate Int-1. After iterative β-H elimination and hydrometallation, a thermodynamically stable ƞ 3 -π-allyl Pd species Int-3 was formed. Finally, the regio-and enantioselective outer-sphere nucleophilic substitution was conducted to deliver the allylation product 3a and regenerate the Pd(0) catalyst.

Discussion
In conclusion, we have established a reliable protocol to realize challenging enantioselective acyclic allylic C(sp 3 )−H functionalization. Hydropalladation is used to trigger stereoselective chainwalking, which is combined with ƞ 3 -allylation to construct a carbon−carbon stereogenic center from an inert C−H bond in high yields and enantioselectivities and 100% atom-efficiency. Studies on the correlation between the pK a of the nucleophiles with the chain-walking and allylation processes showed the origins of the pK a range of nucleophiles (pK a =~13 to~18) and the potential factors controlling the reactivity. Importantly, the products could be derivatized to access diverse valuable enantiopure structures of biologically active compounds. Mechanistic studies showcased the designed merger of chain walking and allylic substitution. This strategy might open new avenues for achieving diverse inert C(sp 3 )−H functionalizations.

Methods
General procedure for the Pd-catalyzed migratory allylic functionalization. To a 4 mL vial in the glovebox under nitrogen were added [Pd(allyl)Cl] 2 (1.8 mg, 0.0050 mmol), L9 (5.6 mg, 0.010 mmol), sodium tetrakis[3,5-bis(trifluoromethyl) phenyl]borate (NaBAr F 4 , 8.8 mg, 0.010 mmol) and dry Et 3 N (0.2 mL). The mixture was stirred at room temperature for 5 min. Then the skipped diene 1 (0.22 mmol) was added to the solution and the reaction continued to stir for 1 min. Finally, the nucleophile 2 (0.20 mmol) was added to the reaction and the resulting mixture was stirred at 60 o C for 24 h. After this time, the crude mixture was cooled to room temperature, condensed, and crude 1 H NMR was obtained with dibromomethane (7 µL, 0.1 mmol) as an internal standard to help determine the regioselectivity and conversion. The reaction was further purified by flash column chromatography to afford the pure allylation product 3−4.

Data availability
For experimental details and procedures, spectra for all unknown compounds, see supplementary files. The X-ray crystallographic data for 11 (CCDC 2076792), have been deposited at the Cambridge Crystallographic Data Center. These data can be obtained free of charge from The Cambridge Crystallographic Data Center via www.ccdc.cam.ac.uk/data_request/cif. Received: 19 May 2021; Accepted: 10 September 2021;