Chiral phosphoric acid-catalyzed stereodivergent synthesis of trisubstituted allenes and computational mechanistic studies

Chiral molecules with multiple stereocenters are widely present in natural products and pharmaceuticals, whose absolute and relative configurations are both critically important for their physiological activities. In spite of the fact that a series of ingenious strategies have been developed for asymmetric diastereodivergent catalysis, most of these methods are limited to the divergent construction of point chirality. Here we report an enantioselective and diastereodivergent synthesis of trisubstituted allenes by asymmetric additions of oxazolones to activated 1,3-enynes enabled by chiral phosphoric acid (CPA) catalysis, where the divergence of the allenic axial stereogenicity is realized by modifications of CPA catalysts. Density functional theory (DFT) calculations are performed to elucidate the origin of diastereodivergence by the stacking- and stagger-form in the transition state (TS) of allene formation step, as well as to disclose a Münchnone-type activation mode of oxazolones under Brønsted acid catalysis.

Chiral molecules with multiple stereocenters are widely present in natural products and pharmaceuticals, whose absolute and relative configurations are both critically important for their physiological activities. In the past few decades, numerous highly enantioselective and diastereoselective reactions have been developed. However, modulation the sense of diastereoselectivity in an asymmetric catalytic reaction is still challenging, because the diastereochemical preference is largely governed by the inherent structure and stereoelectronic nature of the substrates 32 . To address this intrinsic problem, a series of ingenious strategies have been developed for asymmetric diastereodivergent catalysis [33][34][35] , such as using distinct catalysts [36][37][38] , change of metal cations 39 , and ligands 40,41 of the catalysts, change of reaction conditions 42,43 , stereodivergent dual catalysis [44][45][46][47][48][49][50] and stepwise control [51][52][53][54] . Nevertheless, achieving asymmetric diastereodivergent catalysis through modifications of one single type of chiral catalysts remains elusive 55,56 .
Despite the fact that a large number of asymmetric stereodivergent catalytic methods have been developed, most of these methods have been limited to the construction of point stereogenicity divergence, while stereodivergent synthesis of axial stereogenicity has been rarely explored, except using stepwise control strategy 53,54 (Fig. 1b). Herein, we report an enantioselective and diastereodivergent synthesis of trisubstituted allenes via asymmetric conjugate additions of activated 1,3-enynes by oxazolones 57-60 enabled by CPA catalysis, in which the diastereodivergent construction of the allenic axial chirality is realized by modifications of CPA catalysts (Fig. 1c). In addition, the origin of the diastereodivergence is well elucidated by DFT calculations, in which a Münchnone-type activation mode of oxazolones under Brønsted acid catalysis is presented.

Results
Reaction optimizations. We commenced our study by selecting α-alkynyl-α,β-enone 1a and 2-para-methoxyphenyl (PMP) substituted oxazolone 2a as model substrates under CPA catalysis ( Table 1). Under the promotion of TRIP catalyst (CPA A1, 10 mol%) in toluene (with 3 Å molecular sieves) at room temperature, only two diastereomeric allene products among the four potential ones were detected ( Table 1, entry 1), albeit with both poor diastereomeric ratio (dr, 3a:4a 1:2.8) and enantiomeric excess (ee). Subsequently, a series of BINOL and H8-BINOL derived CPA catalysts were screened (entries 2-7). Satisfyingly, the 1-(4-Me-naphthyl) substituted H8-BINOL-derived catalyst B2 provided the allene product in 91% yield with both high diastereoselectivity and enantioselectivity (3a:4a 10:1, 87% ee, entry 7). However, surprisingly, switching the chiral scaffold of 1-naphthyl substituted CPA catalyst form H8-BINOL-type to SPINOL-type (CPA C1) led to the reversal of diastereoselectivity (3a:4a 1:11, entry 8) and formation of product 4a with high ee as well, albeit with moderate yield. To obtain better stereoselectivity control and improve the yield, the effect of the R group at the 2-position of oxazolone was exploited (see Supplementary  Table 1), which indicated that 3,5-dimethoxyphenyl group was the optimal one (entry 9 and 10). Finally, a variety of solvents were also screened, and CCl 4 was chosen as the optimal solvent (see Supplementary Table 1 for details), in which the chiral allene 3a was obtained in 98% yield, 20:1 dr (3a:4a) with 91% ee in the presence of CPA B2, while diastereomeric chiral allene 4a was generated in 85% yield, 12:1 dr (4a:3a) with 98% ee under the catalysis of CPA C1 (entries [11][12]. Substrate scope. Having established the optimal conditions for stereodivergent synthesis of chiral trisubstituted allenes via modifications of CPA catalysts, the substrate scope under the catalysis of CPA (S)-B2 was firstly investigated (Fig. 2). Various substituted phenylacetylenyl groups were well tolerated under the optimal conditions, regardless of the electronic nature and positions of the substitutions, affording the allene products 3a-3h with high diastereoselectivities (>7:1) and enantioselectivities. In addition, substitutions of the R 1 group with heteroaryl, alkenyl and alkyl groups were also amenable, which yielded the products

Me
Ph Ph X-ray strcuture of (±)-3m with high enantioselectivities, albeit with moderate to high diastereoselectivity control (3i-3k). Next, a range of R 2 groups at the β-positions of the enones were explored, which suggested that various substituted aryl groups (3l-3p) were well tolerated, as well as an alkyl group, albeit with moderate dr value (3q). The relative configurations of the allene products 3 were assigned by analogy to 3m, whose relative structure was confirmed by X-ray crystallography. Subsequently, a range of groups at the ketone site (R 3 ) were also investigated, where the Et-substituted, nPr-substituted, nBu-substituted, and iBu-substituted substrates all generated the chiral allenes with both excellent enantioselectivitie and diastereoselectivitie (>9.4:1 dr, >98% ee, 3r-3u). Finally, a series of substitutions (including some functional group-containing substituents) at the 4-position of oxazolones were also exploited under the optimal conditions, which indicated that the chiral allene products could be produced with high enantioselectivities, albeit with a bit erosive diastereoselectivities (3.7:1-5.6:1, 3v-3y).
After investigation of the scope under (S)-B2 catalysis, the substrate scope generality with catalyst (R)-C1 was also studied (Fig. 3). All the R 1 , R 2 and R 4 substituted substrates explored in Fig. 2 were subjected into investigation under the catalysis of (R)-C1 catalyst, which generated the diastereomeric allenes 4 as the major products with both high dr and ee values for most cases (4a-4y). The absolute structures of the chiral allenes 4 were assigned by analogy to product 4c, whose absolute configuration was unambiguously confirmed by X-ray crystallography. Unfortunately, the variations of the R 3 groups of α-alkynyl-α,β-enone 1 under standard conditions were not well tolerated, which led to  Reaction mechanism and origin of diastereodivergence. To investigate the origin of diastereodivergence in construction of the allenic axial chirality, a control experiment was performed. In the presence of CPA catalysts (S)-B2 and (R)-C1, respectively, the asymmetric isomerization of racemic α-alkynyl ketone 23 5a proceeded efficiently to give the chiral trisubstituted allene 6a in the opposite enantiomeric bias (Fig. 4a). In addition, some control experiments of the stabilities of the chiral allene products under various conditions were performed (see Supplementary Fig. 2), which indicated that the conversion from a kinetic epimer to a thermodynamic epimer under the reaction conditions is probably not likely. Based on these results and previous reports [16][17][18] , a preliminary stepwise mechanism was proposed: (1) oxazolones were activated by CPA catalyst via tautomerism to form active enol intermediates; (2) the Michael addition of α-alkynyl enones by the activated oxazolone intermediates generated another enol intermediates INT A stereoselectively under the guidance of CPA catalysts; (3) the CPA catalyst mediated the proton transfer 61,62 of INT A to generate products with the allenic axial stereogenicity (Fig. 4b). For the stereoselectivities of these reactions, we presumed that both (S)-H8-BINOL-B2 and (R)-SPINOL-C1 catalyst generated the same (S,R)-syn-configuration in the Michael addition step, which was followed by the proton transfer step to afford the distinct axial chirality of (S,R,R)-product or (S,R,S)-product, respectively.
To unveil the different stereochemical control of CPA catalysts in Michael addition and proton transfer steps, density functional theory (DFT) calculations were performed at M06-2X/6-311+ +G(d,p)//M06-2X//6-31G(d) level of theory [63][64][65] (see Supplementary Data 1 for Cartesian coordinates of the optimized structures). Surprisingly, based on a well-accepted enol-type mode (mode A) of oxazolones 57,58 , initial calculations of the Michael addition step using an achiral phosphoric acid catalyst (dimethyl phosphate, DMP) provided the anti-addition predication of stereochemical outcomes, which were not in agreement with the experimental syn-addition results (Fig. 4c). Further computational evaluations using chiral (S)-B2 and (R)-C1 catalysts remain the wrong (R,R)-anti-addition prediction as well, suggesting that the diastereoselectivity of the Michael addition step did not rely on the chiral scaffolds of CPA catalysts (see Supplementary Fig. 4). Inspired by the basicity of related functional groups of oxazolone, the imine part of oxazolones acting as proton accepter in hydrogen bonding with CPA catalyst is more accessible rather than the carbonyl group (see Supplementary Fig. 4a). Therefore, a Münchnone-type mode 59,60,66 (mode B) was proposed and the correct syn-addition outcomes were achieved depended on the achiral DMP catalyst (Fig. 4c). The calculation results suggested that transition state (TS) of enol-type requires 6.1 kcal/mol more activation energy than that of the Münchnone-type by achiral DMP catalyst in the prototropic activation step. For the Michael addition step, Münchnone-type model is also superior to enoltype one. And among a number of our calculated Newman conformations (see Supplementary Table 2), the most favored conformation of syn-addition-TS in Münchnone-type is more stable than the most favored conformation of anti-addition-TS in enol-type by 4.2 kcal/mol, indicating this diastereoselectivity is highly model dependent. The calculations, based on chiral (S)-B2 and (R)-C1 catalysts, provided similar results that phosphoric acid catalysts tend to activate oxazolones in a fashion of Münchnonetype mechanism in these reactions, leading to syn-addition products for the diastereoselectivities (see Supplementary Figs. 3  and 4).
The energy profile of Münchnone-type mechanism was performed in Fig. 4d. Because only the Michael addition and proton transfer steps are responsible for the chirality control in these reactions, the prototropic activation steps forming the key precursor INT2 were not shown here (for the details of these steps see Supplementary Fig. 3). From the related INT2-C/-B, the activated 1,3-enyne 1a undergoes a nucleophilic attack by Münchnone-type intermediate of oxazolone via TS2 accompanying the proton delivery of CPA catalyst. There are four major transition states with different chiral features located, namely synisomers (TS2-RS, TS2-SR), anti-isomers (TS2-RR, TS2-SS). Although the CPA catalysts of (S)-B2 and (R)-C1 provide quite similar chiral cavities, they surprisingly display distinct modes of stereoselectivity control in both Michael-addition and proton transfer steps depended on the flexible BINOL backbone and rigid SPINOL backbone, respectively. For (S)-H8-BINOL-B2 catalyst, the final configuration of product Pro-SRR is determined by the proton transfer step with 1.7 kcal/mol free energy difference after the facile and reversible Michael-addition step. Due to sterically repulsive interactions between the phenyl group (R 2 ) at the β-position of 1a and the ethyl group of 2a, the TS2-B-RS is unstable for 1.9 kcal/mol comparing to the favorable TS2-B-SR that is supported by the distortion interaction analysis in Supplementary Fig. 5b. By using the (R)-SPINOL-C1 catalyst, the Michael-addition step mainly determined the reaction rate and led to the more stable INT3-C-SR intermediate readily for proton transfer step forming Pro-SRS. From Supplementary Fig. 5b, the interaction energy of (R)-C1 catalyst and substrate dominates the energy difference of TS2-C-SR and TS2-C-RS probably because of the steric repulsion between 3,3'-substitutent of the C1 catalyst and the ethyl group of 2a, indicating the rigid conformation of SPINOL-C1 catalyst may encounter non-negligible interactions between catalyst and substrate, compared with the relatively flexible conformation of H8-BINOL-B2 catalyst. The stereoisomerization of enol intermediate INT 3 via low barrier tautomerization and single bond rotation was omitted for clarity. Finally, our calculation predicted the products with 95% ee, 17:1 dr (exp. 91% ee, 20:1 dr) and 98% ee, 15:1 dr (exp. 97% ee, 12:1 dr) under (S)-B2 and (R)-C1 catalysis, respectively, which are in agreement with the experimental results.
For the origin of chirality control in the proton transfer step for construction of the allenic axial chirality, we found two types of models among our 20 calculated TSs structures (see Supplementary  Table 3). In Fig. 5a, stacking-type model tends to form S-axial chirality configuration TSs, where the substitutions of the intermediate in transition state TS3 have strong intramolecular interaction (i.e., π-π interaction supported by the NCIs plots 67 of Fig. 5c) and the resonance stabilization between the phenyl group (R 1 ) and the allenic moiety in 1a (see Supplementary Fig. 6 for details). In contrast, stagger-type model prefer R-axial chirality configuration TS with emphasizing the intermolecular interactions of CPA catalyst and substrates (for the distortion-interaction analysis and NCIs plots of TSs see Supplementary Figs. 6 and 7). Due to the relatively rigid conformation of (R)-SPINOL-C1 catalyst, stacking-type model is favored for the TSs of proton transfer step to fit the cavity of catalyst with certain entropy loss, leading to S-axial chirality configuration. That is consistent with less steric repulsion between catalyst and substrate in the stackingtype model, which is the dominant effect by Energy Decomposition Analysis (EDA) calculations 68 as shown in Fig. 5b. The (S)-H8-BINOL-B2 catalyst with slightly flexible conformation are more likely to adopt stagger-type mode forming R-axial chirality configuration, which is mainly stabilized by the dispersion effect by EDA calculation. As shown in NCIs plots in Fig. 5c     B2 featuring a 1-(4-Me-naphthyl) substitution (over just 1-naphthyl in catalyst B1) could improve the diastereoselectivity significantly (2.7:1 to 10:1, see Table 1). For this catalyst, the ΔΔE orbital , a high proportion for energy, stabilized for the unfavorable TS3-B-SRS, which can be rationalized by the dihedral angle ∠C1-C2-C3-H1 of forming hydrogen bond and allene as shown in Supplementary Fig. 8. The dihedral angle in TS3-B-SRS is 85.0°, which is more close to perpendicular than that in TS3-B-SRR a Interaction modes between CPA catalyst and enol intermediat for proton transfer , where the entropy term is from Gaussian calculations. For the energy decomposition analysis equation follows: ΔΔE 0 ‡ = ΔΔE steric + ΔΔE elstat + ΔΔE orbital + ΔΔE disper . The ΔΔE ‡ is calculated from the electronic energy difference between stacking-type TS and stagger-type TS (ΔΔE ‡ = ΔE stacking-TS ‡ − ΔE stagger-TS ‡ ). Each energy component (ΔΔE steric , ΔΔE elstat , ΔΔE orbital , and ΔΔE disper ) is calculated in the same fashion. Positive ΔΔE values indicate corresponding interactions promote stagger-type, and R-axial chirality TS is more favored. In contrast, negative ΔΔE values indicate S-axial chirality TS with stacking-type is more stable; E steric is the positive term as the steric repulsion, and E disper is the negative term as the dispersion stabilization. c Noncovalent Interactions (NCIs) analysis of transition state for proton transfer (green, dispersion effect; red, steric effect). The gradient isosurfaces (s = 0.3 a.u.) are colored on an BGR scale according to sign(λ 2 )ρ over the range −0.01 to 0.01 au. d Detailed structural information of proton transfer transition states catalyzed by two CPAs. The energies shown in kcal/mol. To confirm validity of the stereo-models in the proton transfer step, we performed additional DFT calculations for understanding substituent effects of R 2 and R 3 group in 1 as shown in the cyclohexyl group (Cy) at the R 2 of 1 was introduced and TS3-B-SRS-Cy would be stabilized due to the CH-π dispersion effect between cyclohexyl and 4-Me-naphthyl group, which therefore decrease the energy gap for diastereoselectivity. And, n-propyl group (nPr) at R 3 in 1 also emphasize the dispersion effect with 4-Me-naphthyl group of B2 catalyst, which mainly stabilize the transition state TS3-B-SRR-nPr and thus increase the diastereoselectivity. In contrast, the relatively rigid conformation and small cavity for C1 catalysis make steric repulsions become important. Introducing the relatively larger cyclohexyl group to substrates would increase energy barriers for both the stacking-forms and stagger-forms of TS3-C, resulting slightly increased energy gap for the diastereoselectivity (see Supplementary Table 5 for details). While switching the R 3 group to n-propyl under C1 catalyst led to very low reactivity by our experiment (see Supplementary Fig. 1), and thus further discussions on the diastereoselectivity is trivial.

INT3-B-RS
Derivatizations of the chiral allene products for diversityoriented synthesis. To demonstrate the applicability of these reactions, we devoted our efforts to exploring the derivatizations of the chiral allene products (Fig. 6). Rearrangement of chiral allene 4a into tri-substituted furan 7a was performed under gold (I) catalysis, providing the product in 92% yield. Further alcoholysis of the oxazolone moiety in 7a provided the β,β-di-aryl substituted amino acid derivative 8a, which is a type of important pharmacophore in a series of bioactive small molecules. Analogously, gold(I)-catalyzed rearrangement of 3a afforded the furan derivative 7a with the same bias on a chiral stationary phase, thus confirming the absolute configuration of 3a (Fig. 6a). To achieve diversity-oriented synthesis (DOS) 69 from these chiral allene products, diastereoselective reduction of 3a with L-Selectride followed by in-situ intramolecular transesterification provided the lactone 9a (>25:1 dr) 70 . Aminolysis of lactone 7a with nPrNH 2 yielded the allenic alcohol 10a, which was stereospecifically cyclized in the presence of gold(I) catalyst to give the 2,5-dihydrofuran derivatives 11a, whose absolute structure was confirmed by X-ray crystallography. Analogously, treatment of chiral allene 4a with the same three-steps procedure readily provided the diastereomeric chiral 2,5-dihydrofuran 15a (Fig. 6b). Moreover, derivatizations of the chiral allene products were not limited to stereochemical diversities, but could be extended to skeletal diversities. Electrophilic iodocyclization of lactone 9a in the presence of NIS provided the tricyclic product 12a, in which the phenyl group reacted as the nucleophile. On the other hand, treatment of the diastereomeric lactone 13a with the same electrophilic iodination conditions generated the bridged bicyclic product 16a, in which the amide group reacted as the nucleophile; the structures of these cyclization products were all well confirmed by X-Ray crystallography.

Discussion
In summary, we have disclosed an enantioselective and diastereodivergent synthesis of trisubstituted allenes via asymmetric conjugate additions of oxazolones to activated 1,3-enynes under chiral phosphoric acid catalysis, where the axial stereogenicity of the chiral allene products could be well modulated by modifications of the CPA catalysts. The origin of allenic axial diastereodivergence is well elucidated by the stacking-form and staggerform in transition states from DFT calculations, in which a Münchnone-type model on activation of oxazolones under Brønsted acid catalysis has also been demonstrated with high model dependency of diastereoselectivities. The stereo-specific and chemo-specific transformations of the diastereomeric chiral allenes into more complex stereoisomers and skeletal isomers demonstrate the value of these reactions in organic synthesis, especially in the field of DOS.

Methods
General procedure for asymmetric synthesis of chiral products 3 and 4. To a dried 3 ml vial was added 1 (0.15 mmol), 2 (0.1 mmol), CPA catalyst (0.01 mmol), and activated 3 Å molecular sieves (100 mg). The vial was purged with N 2 for 3 times and then followed by adding CCl 4 (0.5 mL). After stirring for 24 h at room temperature, the reaction mixture was quenched by adding K 2 CO 3. After filtration, the filtrate was concentrated under vacuum to give a residue, which was purified by flash column chromatography to give the allene products 3 or 4. (Toluene is an alternative choice of solvent used in these reactions, if the usage of CCl 4 is restricted). Full experimental details and characterization of new compounds can be found in the Supplementary Information.