Vicinal stereocenters are prevalent in naturally occurring products and pharmaceutical agents (Fig. 1a), and therefore, the stereoselective construction of these structural units represents a subject of a longstanding interest in chemistry1. Traditional approaches usually establish the two stereocenters sequentially, which thus require multistep manipulations. A conceptually straightforward and efficient approach to access vicinal stereocenters is by doubly stereoconvergent2,3,4 C–C bond-forming reactions, such as the ones depicted in Fig. 1b. In this case, racemic compounds 1 and 2 are each converted to prochiral intermediates, whose subsequent combination affords the stereochemically defined product 3 with two stereocenters. In practice, however, the requirement to simultaneously control the relative and absolute configurations during the key C–C bond-forming event presents significant difficulties. This task becomes more challenging if an all-carbon quaternary center is involved, as a result of the increased steric repulsion. The challenge further escalates if none of the two stereocenters is embedded in a cycle, due to the enhanced conformational flexibility5,6,7.

Fig. 1: Enantioselective construction of vicinal tertiary and all-carbon quaternary stereocenters.
figure 1

a Vicinal 3° and all-carbon 4° stereocenters in pharmaceutical agents. b Doubly stereoconvergent approaches to make vicinal stereocenters. c Cu/Mg-dual catalytic APS reaction to make vicinal 3° and 4° stereocenters (this work).

In spite of these challenges, extensive efforts have been devoted to developing reactions shown in Fig. 1b, with remarkable progress achieved8. Among the methods reported to access these stereodiads, the transition metal-catalyzed asymmetric allylic substitution (AAS) reaction9,10,11,12,13,14,15 (3, R5 = vinyl, Fig. 1b) has attracted the most attention and shown tremendous power. For instance, the Trost group developed a Mo-catalyzed AAS of cyanoesters using cinnamyl carbonates, affording products with vicinal tertiary and quaternary carbons16. The Stoltz group employed the Ir-catalyzed AAS reaction in accessing this type of stereodiads17,18. The Carreira group reported an elegant allylic alkylation of aldehydes/ketones by their Ir/organocatalyst dual catalytic system19,20. These methods are flexible since they employ readily available starting materials and generate synthetically versatile, terminal-alkene-containing products (cf. 3). Notably, however, for most of the products 3 accessed by the above AAS-based methods, the R4 substituent (carried from electrophile 2) is limited to aryl or alkenyl groups. These advancements notwithstanding, general methods for building vicinal acyclic, congested stereodiads remain highly needed in synthesis. Ideally, a reaction may employ economical and simple catalysts and at the same time accommodate the broad scope of nucleophiles and electrophiles.

Our group21,22,23,24 has been interested in the copper-catalyzed asymmetric propargylic substitution (APS) reactions25,26,27,28,29,30,31,32,33,34,35,36,37,38. In terms of bond disconnection, the metal-catalyzed APS reactions are analogous to the AAS reactions. In terms of synthetic utility, the terminal alkyne group installed by the APS reaction is arguably the most versatile39,40 functionality in chemistry. However, the scope of nucleophiles that can be employed in the Cu-catalyzed APS reactions remains limited compared with the AAS reactions. To understand such a limitation, we have shown through kinetic studies an important mechanistic distinction of the Cu-catalyzed APS reactions: the generation of electrophilic intermediate (see 9 in Fig. 1c) is likely the rate-limiting step, and this intermediate would undergo polymerization in the absence of a potent nucleophile24. This feature may have rendered expanding the scope of nucleophiles a difficult task. Regardless, through developing new catalytic systems, remarkable progress has been achieved in this field. For example, we21,22,24 and the Nishibayashi group41 have realized asymmetric propargylation of aliphatic alcohols, a class of nucleophiles previously considered inactive in this transformation.

In continuation of our effort in this field, we wonder if we could employ the simple, linear 1,3-dicarbonyl based nucleophiles (e.g., 4) in the Cu-catalyzed APS reactions (cf. Fig. 1c, 4 + 5 to 6, via 8 and 9). Once realized, this would represent a general tool to construct acyclic, vicinal tertiary, and all-carbon quaternary centers. Furthermore, by facile functional group interconversions of the terminal alkyne group installed, a broad scope of derivatives could be readily accessed. We reasoned that the skinny nature of the alkyne substituent in 5 (and 9) could, to some extent, ameliorate the steric repulsion between both reactants42. If 4 could be deprotonated under mild conditions, the resulting soft nucleophile 8 should be able to trap 9 effectively. However, a preeminent challenge lies in how to control the configuration of the acyclic, all-carbon quaternary center in 6: the prochiral nucleophilic center in 8 is quite distant from the chiral environment exerted by the ligands in 9. Previous related studies had to employ cyclic nucleophiles such as 2-indolones to ensure high diastereoselectivities43.

In this work, we develop a Cu/Mg-dual catalytic system for a highly diastereoselective and enantioselective propargylation of simple, linear 1,3-dicarbonyl compounds (Fig. 1c). We establish that the use of a bulky ligand for Cu-catalyst and Mg(OtBu)2 as a base is vital for the efficiency and selectivity of this process. As key advantages of this method, it employs non-precious metal catalysts, and readily available reagents/substrates, and demonstrates significant scope with respect to both reaction partners. Importantly, the R group in electrophile 5 could be either aryl or alkyl in nature. The terminal alkyne-containing products 6 could be smoothly converted to other classes of compounds (67), further adding to its synthetic utility.


Reaction optimization

To commence our investigation, we used propargyl carbonate 10a and β-keto ester 11a as model substrates (Fig. 2). Not surprisingly, under the “classical” Cu-catalyzed APS reaction conditions [Cu(CH3CN)4BF4 as the catalyst precursor, L1 as ligand, and iPr2NEt as base], product 12 was formed, but with low yield and poor diastereoselectivity (entry 1). We then focused our efforts to improve the efficiency and selectivity of this process. We first considered strategies to restrict the rotational flexibility of the nucleophile 11a. Along this line, we investigated the use of various Brønsted bases that contain Lewis acidic metal counterions44,45,46, with the expectation that they would deprotonate 11a to form cyclic metal enolate intermediates such as 13. Among the bases screened (entries 2–6), Mg(OtBu)2 afforded 12 in best yield (84%) but still with moderate diastereoselectivity (entry 6). The use of other Cu sources (entries 7–8) instead of Cu(CH3CN)4BF4 led to inferior selectivities and efficiency. We next examined the impact of ligand structures on the reaction performance (entries 9–12). The use of Hu’s tridentate P, N, N- ligand28 or bis-oxazoline ligand47 L3 gave no detectable amount of product. Interestingly, the use of a bulkier Pybox-type ligand L4 slightly improved the diastereoselectivity. Along this line, we synthesized various sterically more demanding ligands, and identified pyridinebisimidazoline48 (PyBim) L5 to give 12 in excellent enantio- and diastereoselectivities (entry 12). These results suggest that using large ligands may be conducive to acyclic stereocontrol for this transformation. Under the above conditions, we further showed either 10a or 11a could serve as the limiting reagent, giving almost identical results (entry 13). Lastly, when the ratio of ligand to Cu was reduced to 1:1, the product was obtained in almost identical stereoselectivitities (entry 14). We used a slightly excess amount of ligand in the following studies to ensure efficient ligation of Cu salts and to minimize background nonstereoselective propargylation processes.

Fig. 2: Condition optimization.
figure 2

Reactions in this table were performed on a 0.1 mmol scale. Yields and d.r. values were determined by 1H NMR spectroscopy of the crude reaction mixture with 1,1,2,2-tetrachloroethane as an internal standard. The ee values were determined by HPLC analysis. a10a (1.5 equiv.) and 11a (1.0 equiv.) were used. bThe ratio of ligand to Cu is 1:1.

Reaction scope

With the optimized reaction conditions established, we next examined the substrate scope of this transformation. As shown in Fig. 3, a variety of propargyl carbonates could be employed as electrophiles. For example, the aryl group at the propargylic position could be either electron-rich (14e, 14l) or electron-deficient (14a, 14g). Those bearing a substituent at the ortho position (14f) or containing bicyclic ring systems (14i, 14l–m) partook in the reaction smoothly. Functional groups such as aryl halides (14b–c, 14h), boronic acid esters (14d), and cyano (14g) are compatible with the reaction conditions. These functional groups in products are useful handles for further derivatizations. Importantly, pharmaceutically relevant heterocycles, including furans (14j), thiophenes (14k), thiazoles (14n), benzodioxanes (14l), pyridines (14o), and indoles (14m) could also be incorporated. Lastly, we show that the substituent at the propargylic center is not limited to aryls, and those bearing alkyl groups can be employed (14p–q) under slightly modified conditions.

Fig. 3: Substrate scope with respect to electrophiles.
figure 3

Reactions were performed on a 0.2 mmol scale in THF (2 mL) with 5 mol% Cu(I), 10 mol% L5, and 20 mol% Mg(OtBu)2. Yield is that of an isolated product. The ee values were determined by HPLC analysis, the d.r. values were determined by 1H NMR analysis of the crude reaction mixture. aThe pentafluorobenzoate group was used as leaving group. bCu(CH3CN)4BF4 (20 mol%), L5 (40 mol%), and Mg(OtBu)2 (80 mol%) were used. See SI for experimental details.

Fig. 4: Substrate scope with respect to nucleophiles.
figure 4

Reactions were performed on a 0.2 mmol scale in THF (2 mL) with 5 mol% Cu(I), 10 mol% L5, and 20 mol% Mg(OtBu)2. Yield is that of an isolated product. The ee values were determined by HPLC analysis, the d.r. values were determined by 1H NMR analysis of the crude reaction mixture. aCu(CH3CN)4BF4 (15 mol%), L5 (30 mol%), and Mg(OtBu)2 (50 mol%) were used. See SI for experimental details.

We then examined the scope with respect to the nucleophiles (Fig. 4). To our delight, both β-ketoesters (15a–m) and malonate esters (15o–u) are tolerated by this reaction. Again, various (hetero)aryl groups of different electronic properties and substitution patterns can be incorporated. Notably, pyridine rings (15f) and chromones (15g) are well accommodated. Moreover, various alkyl substituents are allowed. In addition, alkene groups (15e, 15r) are also compatible with our conditions. Even the allyl ester group in 15t, a potent reacting functionality in the TM-catalyzed AAS reaction, is completely inert under our conditions, highlighting the mechanistic distinction between the AAS and the APS reactions. Particularly noteworthy is the selective formation of product 15k, which contains a tertiary fluoride: methods to prepare fluorine-containing stereocenters remain underdeveloped but highly needed49. Nucleophiles bearing three different carbonyl groups (15l, 15m) were also examined. Interestingly, while product 15l was obtained with excellent diastereo- and enantioselectivities, 15m was formed in poor selectivities. Reactions employing 1,3-diketone nucleophiles proceeded with good efficiency but low diastereoselectivites (15n). Regardless, we show this reaction could be carried out on a gram-scale under slightly modified conditions, with no notable erosion of yield and stereoselectivities (15o).

Product derivatization

To further demonstrate the synthetic utility of this method, we performed product derivatization reactions shown in Fig. 5. First, treatment of 12 with pinacolborane (H-BPin) in the presence of HZrCp2Cl led to vinyl boronic acid ester 16a50. Alkyne 12 could also undergo [3 + 2] cycloadditions with the in situ generated nitrile oxides, to afford 1,2-oxazoles 16b51. Further, alkyne 15o partook in an annulation reaction to give the indole 16c52. Treating 15o with basic conditions generated allene 16d53. Moreover, alkyne 15o could participate in a Sonagashira coupling with methyl 4-iodobenzoate, leading to 16e54. The CuAAC reaction40 of 15o with anti-HIV drug Zidovudine resulted in 16f with excellent efficiency. Notably, selective amination55 of 12 could be achieved using methylurea with NaOMe employed as a base. Finally, treatment of 15t with pyrrolidine in the presence of Pd(PPh3)4 triggered a deallylation and a decarboxylation cascade56, yielding product 18 that bear contiguous tertiary stereocenters in good diastereo- and enantioselectivities. The synthetic versatility of these products adds to the utility of our current method.

Fig. 5: Product derivatization.
figure 5

a Derivatization of 12 and 15o. b Derivatization of 12. c Derivatization of 15t. Reactions were performed on a 0.1 mmol scale. Yield is that of an isolated product. The ee values were determined by HPLC analysis. The d.r. values were determined by 1H NMR analysis of the crude reaction mixture. See SI for experimental details.

Mechanistic studies

To explain the observed diastereoselectivities of our reaction, we proposed the stereochemical model in Fig. 6a based on experimental results and literature precedence25,26. We reason that Mg(OtBu)2 activates nucleophiles by deprotonating β-ketoesters to form the cyclic intermediate 20. On the other hand, the copper complex activates electrophiles by forming the copper allenylidene intermediate 19. Attack of 20 to 19 may adopt a geometry shown in Fig. 6 (see 21), in order to minimize the repulsion between the ligated copper moiety and the large substituent of 20. The use of bulky ligand to complex with Cu(I) helped to better differentiate the two substituents (RL vs. RS) in the nucleophiles, leading to higher diastereoselectivities. As shown in Fig. 6b, the reaction employing cyanoester 23 as nucleophile under our standard conditions gave the corresponding product 24 with poor diastereomeric ratio. This is likely because cyanoester 23 could hardly form a cyclic metallic complex (cf. 18) with Mg(OtBu)2, due to the conformational constraints imposed by the linear cyano group. These results provide support for our proposed stereochemical model.

Fig. 6: Mechanistic studies.
figure 6

a Proposed stereochemical model. b Reaction of substrate 23. The ee value was determined by HPLC analysis. The d.r. value was determined by 1H NMR analysis of the crude reaction mixture. See SI for experimental details.


In conclusion, we have established a general, mild, enantio-, and diastereoselective method to construct vicinal tertiary and all-carbon quaternary stereocenters from simple starting materials. The method, which exploits the copper-catalyzed asymmetric propargylic substitution reaction, builds two stereocenters simultaneously by a doubly stereoconvergent process and furnishes a broad array of functional group-rich, terminal alkyne-containing products. Critical to the success of this development are the use of a Cu/Mg-dual catalytic system and the use of a bulky, tridentate PyBim-type ligand. The synthetic versatility of products further adds to the power of this method. We anticipate this method will be rapidly adopted to make the challenging stereodiads that are frequently encountered in natural products and drug candidates, and inspire developments of novel Cu-catalyzed APS reactions.


General procedures for the Cu/Mg-catalyzed propargylation

In an N2-filled glovebox, Cu(CH3CN)4BF4 (3.2 mg, 0.01 mmol, 5 mol%), L5 (11.0 mg, 0.02 mmol, 10 mol%), and THF (2 mL) were sequentially added to an 8 mL screw cap vial containing a magnetic stirring bar. The mixture was stirred at r.t. for 1 h. Then, propargyl carbonate 10 (0.30 mmol, 1.5 equiv.), 1,3-dicarbonyl compound 11 (0.20 mmol, 1.0 equiv.), and Mg(OtBu)2 (0.04 mmol, 0.2 equiv.) were added sequentially, and the resulting mixture was allowed to stir at r.t. for 24 h. After completion, the reaction mixture was quenched by the addition of water. The mixture was extracted with ethyl acetate three times, and the combined organic layers were washed with brine, dried over anhydrous Na2SO4, and concentrated. The crude product was purified by flash chromatography, dried under vacuum, weighed, and sampled for HPLC analysis.