Asymmetric synthesis of N-bridged [3.3.1] ring systems by phosphonium salt/Lewis acid relay catalysis

Optically pure pseudo-natural products (PNPs), particularly exemplified by azabicyclo[3.3.1]nonane molecules and their analogs provide an attractive platform for structure−activity relationship studies, and also lead new compound discovery in drug development. However, there are currently no examples of guiding catalytic asymmetric strategies available to construct such important PN-scaffolds, thus limiting their broad use. Here, we report a general and modular method for constructing these pseudo-natural N-bridged [3.3.1] ring systems via cascade process by bifunctional phosphonium salt/Lewis acid relay catalysis. A wide variety of substrates bearing an assortment of functional groups (59 examples) are compatible with this protocol. Other features include a [3 + 2] cyclization/ring-opening/Friedel-Crafts cascade pathway, excellent reactivities and stereoselectivities, easily available starting materials, step economy and scalability. The obtained enantioenriched products showed potential of preliminary anticancer activities. Insights gained from our studies are expected to advance general efforts towards the catalytic synthesis of challenging even unprecedented chiral PNPs, offering new opportunities for bioactive small-molecule discovery. Optically pure pseudo-natural products (PNPs) bearing N-bridged [3.3.1] ring systerm provide a platform for structure−activity relationship studies, but currently no examples of catalytic asymmetric strategies are available to construct such scaffolds. Here the authors show a general method for constructing such N-bridged [3.3.1] ring systems via a bifunctional phosphonium salt/Lewis acid relay catalysis.

C hiral azabicyclo [3.3.1]nonane-containing molecules, especially these aryl-fused N-bridged [3.3.1] ring systems that have highly functionalized and complex architectures represent an important and unique family of (pseudo-)natural products, which have arguably been subject to extensive structural, pharmacological, biosynthetic, and synthetic investigations owing to their multitudinous bioactive and medicinal properties [1][2][3] . Most of their family members as exemplified by Sarpagine, Macroline, Suaveoline, Trinervine, etc. (Fig. 1a), which belong to the group of monoterpenoid indole alkaloids [4][5][6][7] , are mainly isolated from infrequent natural plants and particularly have antiproliferative activity and antitumor effect, thus playing key roles in the treatment of Alzheimer's disease (AD), Parkinson's disease, etc. [8][9][10] . Consequently, as an important subject of biology-oriented synthesis (BIOS) 11,12 , these Nbridged [3.3.1] ring systems are arguably one of the most versatile classes of target scaffolds towards developing and discovering smallmolecule drugs and/or their leading compounds, which have been fascinated and pursued by both synthetic and pharmaceutical chemists for decades 7,13,14 . Synthetic access to these (pseudo-)natural compounds was pioneered by Cook and co-workers, who have accomplished very elegant total synthesis of several indole alkaloids with this structural feature by using Dieckmann condensation to construct the N-bridged framework [13][14][15][16] . Besides, efficient strategies also include biomimetic synthesis and olefin metathesis by the Martin group 17,18 , a [5 + 2] cycloaddition/ring enlargement by the Gaich group 19,20 , and other methods [21][22][23] . However, despite these advances, most of the existing methods require tediously long steps including functional-group pre-installations and transformations, harsh reaction conditions, and complicated operations. Moreover, only sporadic examples that disclosed the synthesis of chiral azabicyclo[3.3.1]nonane compounds have been available so far (Fig. 1b), which utterly relied on utilizing chiral auxiliary 24 or employing chiral starting materials [14][15][16][17][18][19][20][21][22][23] , respectively. To the best of our knowledge, the catalytic asymmetric protocol for constructing such ring systems has never been reported to date, which remains a tremendous challenge and long-standing assignment in modern synthetic chemistry. Of note, these eight-membered N-bridged ring systems are particularly difficult to prepare due to their high strain energies, lacking of effective catalytic systems to overcome the negative enthalpic and entropic challenges of such organic reactions.
From the perspective of efficiency and practicability of chemical synthesis, one essential objective is to convert readily available and inexpensive starting materials into structurally complex and multifunctional molecules with a perfect stereocontrol 25 . In this context, asymmetric organocatalysis that greatly enhances the synthetic toolbox by complementing metal-based and enzymatic methodologies offers powerful solutions to this endeavor 26 , which is conducive to expanding the diversity of chiral molecules for modulating biological targets, thus providing more opportunities for drug discovery and development 27 . Among the well-known asymmetric organocatalytic systems, asymmetric phase-transfer catalysis (PTC) 28 particularly involving phosphonium salts as catalysts provide a powerful and versatile tool for the enantioselective synthesis of diverse chiral molecules [29][30][31][32] . The groups of Maruoka, Ooi, and Zhao have pioneered to make significant contributions in this field [33][34][35][36][37] . Recently, our group has developed a series of multi-/bifunctional phosphonium salt catalysts and demonstrated their applications in a variety of asymmetric synthetic reactions [38][39][40][41] . In connection to the above synthetic challenge and our disclosure of phosphonium salts and their successful applications in asymmetric synthesis, we envisioned that the employment of highly tunable bifunctional phosphonium salts together with metal-free Lewis acid catalysts may result in a practical cascade transformation towards constructing these synthetically challenging azabicyclo[3.3.1]nonane scaffolds.
Herein, we describe an asymmetric synthesis of pseudo-natural N-bridged [3.3.1] ring systems via a novel cascade approach by chiral phosphonium salt/Lewis acid relay catalysis. Starting from commonly available indole-based α,β-unsaturated ketones, and cyclic azomethine ylides, this cascade reaction proceeded along a [3 + 2] cyclization/ring-opening/Friedel-Crafts pathway. The stereocontrol in the phosphonium salt-promoted cyclization step is realized via the synergistic effects of ion-pair and hydrogen-bonding between both reactants and cationic catalysts, respectively. DFT investigations provided insights into the Lewis acid-promoted ringopening/ring-expansion process. Moreover, preliminary biological activity investigations indicated that these azabicyclo[3.3.1]nonanebased molecules are potential antitumor agents (Fig. 1c).

Results
Reaction development. Our general approach to address the above issue is depicted in Fig. 1c. We initially designed and synthesized indole-based α,β-unsaturated ketones 1, and cyclic azomethine ylides 2 as starting materials. We hypothesized that an asymmetric 1,3-dipolar cyclization between these two types of reactants would proceed by phosphonium salt catalysis, generating chiral indole-fused tropane intermediates 3 41 . Next, metal-free Lewis acid catalysts such as BF 3 would promote a ringopening process of this key intermediate via coordination of the Lewis acid to the carbonyl group. Subsequently, an intramolecular Friedel-Crafts-type reaction could occur, in which the electronrich indole moiety serves as the nucleophile to attack the iminium unit, thus furnishing the desired N-bridged [3.3.1] ring systems 4. While 2-alkenylindole compounds have been well utilized in organic synthesis 42 , the employment of 2-indolylethylene ketones in the asymmetric cascade process begins with a dipolar cyclization, in which these substrates initially serve as C2-sythons, is virtually unexplored 43 . Of note, the crucial asymmetric induction in this cascade should come from the first cyclization step, and thus would be expected to originate from the ion-pairing and H-bonding capability of the chiral phosphonium catalyst. Whereas the idea of constructing optically pure pseudo-natural N-bridged [3.3.1] rings by metal-free relay catalysis that involves a chiral phosphonium salt catalyst cooperated with a Lewis acid was very appealing to us, we were mindful that the catalytic asymmetric cascade between 2-alkenylindoles and cyclic azomethine ylides for producing these rigid azabicyclo[3.3.1]nonane scaffolds is extremely challenging, and moreover the catalytic asymmetric protocol towards synthesizing such pseudo-natural products has never been accomplished so far.
With this consideration in mind, we initially evaluated the feasibility of this hypothesis by choosing the cascade reaction between N-methylated 2-alkenylindole 1a and benzyl-fused cyclic azomethine ylide 2a in the presence of racemic dimethyldiphenylphosphonium iodide catalyst P0 under standard PTC conditions as the model reaction. As expected, the dipolar [3 + 2] cyclization product rac-3a that was identified by X-ray crystallographic analysis (CCDC 2089968) was readily generated and isolated in 94% yield (Fig. 2 favored since the seven-membered intermediate 3a is less stable than the final eight-membered product 4a. Condition optimization. Encouraged by the initial results, we next explored the asymmetric version of this cascade reaction. For the development of effective catalysts, given our recent progress in developing phosphonium-involved bifunctional catalytic systems 38-41 , we chose amide-, dipeptide-and thiourea-derived phosphonium salts as candidate catalysts because of their representative hydrogen-bonding and ion-pairing features. Pleasingly, all tested phosphonium salt catalysts were effective in promoting this reaction, furnishing the desired N-bridged azabicyclo[3.3.1] nonane product in high yields and excellent diastereoselectivities (all > 20:1 d.r.) (Fig. 3). When phosphonium salts P1 bearing an amide framework were employed, the reaction proceeded smoothly, affording the desired product in good yield and diastereoselectivity, albeit with low enantioselectivity. Delightedly, while dipeptide-based phosphonium salts P2-3 that particularly bear double H-bonding-donating moieties were used, the ee value was slightly improved. At last, thiourea-based phosphonium salt P4 was found to be more effective in promoting this reaction, affording the product with a slightly higher ee value.  Tables S2-S6 for more details).
Substrates scope exploration. Encouraged by the above exciting results, we proceeded to establish the substrate scope for this novel cascade reaction. Firstly, the scope of 2-vinylindoles was assessed by using cyclic azomethine ylide 2a as the reacting partner under the above optimal conditions. As shown in Fig. 4, various aryl-substituted 2-vinylindoles regardless of the positions and electronic properties of the substituents on the phenyl ring were perfectly compatible with the reaction conditions, providing the corresponding products (4a-n) in high yields (85−95%) with excellent diastereoselectivities (>20:1 d.r.) and good to excellent enantioselectivities (83−99% e.e.) Notably, 2-vinylindoles containing heteroaromatic rings such as pyrrole, furan, thiophene, pyridine, benzofuran, benzothiophene, and benzopyridine were also proven to be ideal substrates, affording the desired products (4o-u) in high yields with satisfying stereoselectivities. Importantly, alkyl-substituted 2-vinylindoles were also well-tolerated in this reaction, producing the corresponding products (4v-aa) with excellent yields and good stereoselectivities excepting for the 1adamantane-substituted product 4aa with a slightly lower ee value, which may be attributed to the steric hindrance of the adamantane group. Also, good yields and excellent stereoselectivities (4ab and 4ac) were obtained from these reactions, in which the protective group on the nitrogen-atom of substrates was changed. In addition, the substituents of the indole rings did not affect the reactivities and stereoselectivities of the cascade reaction, affording the corresponding products 4ad−i in high yields and excellent dr and ee values in all cases. Subsequently, we further evaluated the scope of the cyclic azomethine ylides 2 (Fig. 5). Delightfully, an array of cyclic azomethine ylides with electron-donating or electron-withdrawing substituents at the 6-or 7-position of the phenyl ring were found to be suitable reaction partners, furnishing the desired products 5a-l with good yields (82-95%) and high ee values (82-98%). Alternatively, the azomethine ylide that was fused by the naphthyl group was also proven to be tolerated for this cascade system, providing the corresponding product (5m) in good stereoselectivities. Also, the cyclic azomethine ylides with different esters were suitable substrates for affording the desired products (5n and 5o). It was worth mentioning that the cyclic azomethine ylides bearing Entry Additive (10 equiv.) t (h) Yield of (rac)-4a a Conditions  different alkyl substituents could be compatible with the standard conditions, affording the products (5p-t) in high yields with excellent diastereo-and enantioselectivities.
Synthetic applications. In order to evaluate the practicality of this cascade protocol, the cyclic azomethine ylide species that have been installed on different drug molecule skeletons were employed as starting materials. All these supported cyclic azomethine ylides could also serve as effective substrates, constructing the corresponding cascade products 6a-d in good yields, satisfying enantioselectivities, and excellent diastereoselectivities (Fig. 6a). Moreover, this synthetic protocol is equally robust for gram-scale synthesis with no erosion of both yields and stereoselectivities (Fig. 6b). In view of the importance of these azabicyclo[3,3,1]nonane scaffolds in natural products and pharmaceutical chemistry, the synthesis of diverse azabicyclo[3,3,1]nonane-containing molecules is of great interest. As illustrated in Fig. 6c, the eight-membered product 4a could be easily transformed into different classes of highly functionalized bioactive compounds (7a-d) that bear intrinsically synthetic challenges.
Biological activity study. Particularly, we were intrigued by the potential biological activities of these chiral pseudo-natural N-    (Fig. 7c). Generally, titration of 2-indole-vinyl ketone substrate 1a to the optimal catalyst P10 led to a distinct change in the position of thiourea-NH 1 signal of catalyst, while titration of cyclic azomethine ylide 2a to this catalyst P10 led to a less pronounced change (Fig. 7c, left, also see Supplementary Fig. S7 for more details). When methylated phosphonium salt P10' was used as the catalyst, the reaction became slightly slower and the enantioselectivities decreased to 0% ee. In addition, when the reaction was performed in methanol, a racemic product was obtained (Fig. 7c, right). These results offered clear clues that the ionpairing and H-bonding effects are essential for enhancing the reactivity and stereoselectivities of this cascade reaction. Furthermore, bifunctional phosphonium salt catalyst P10 particularly possesses a semi-enclosed cavity with an electropositive region as demonstrated by its computed electrostatic potential (ESP) map (Fig. 7d), which would be advantageous to control the stereoselectivities of the initial [3 + 2] cyclization. On the basis of the above results, a plausible reaction mechanism of this reaction was proposed in Fig. 8a, which illuminated that a complex cascade process containing [3 + 2] cyclization, ring-opening, and intramolecular Friedel-Crafts-type reaction was involved in the catalytic cycle. The initial [3 + 2] cyclization in this cascade is likely to follow the general mechanism described in the literature 44  To further understand the mechanism of this cascade reaction particularly on the process of acid-catalyzed ring-opening/ Friedel-Crafts cascade process of intermediate 3a, we performed density functional theory (DFT) calculations (see Supplementary  Information for computational details). In our calculations, seven-membered ring complex 3a was chosen as the model reactant, which can deliver azabicyclo[3.3.1]nonane derivative 4a as observed experimentally, where 3a was set as the relative zero point for computed Gibbs free energies of various intermediates and transition states. As depicted in Fig. 8b, ligand exchange between boron trifluoride ethyl ether complex and substrate 3a gives carbonyl-coordinated complex D with endergonic energy of 3.6 kcal/mol, which is attributed to the slightly weaker coordination ability of the carbonyl group as compared to ethyl ether. In complex D, the coordination with the Lewis acidic boron trifluoride catalyst facilitates electron density flow from the N-H moiety to the carbonyl moiety in substrate 3a, which significantly activates the C−C bond. The heterolytic cleavage of the C-C bond then smoothly occurs via transition state TS1 with an energy barrier of 7.5 kcal/mol (11.1 kcal/mol relative to 3a), which produces iminium species E. In TS1, the length of the breaking C-C bond is 2.13 Å. Subsequently, the intramolecular Friedel-Crafts-type reaction proceeds via transition state TS2 with an energy barrier of 9.3 kcal/mol, generating cationic species F. In TS2, the length of the forming C-C bond is 2.04 Å. Rearomatization and enol-keto tautomerization of F through proton transfers subsequently occur to afford the eight-membered ring complex G. This process is presumably facile and likely to be solvent-assisted, thus the transition states were not computed. Finally, the ligand exchange with diethyl ether solvent yields azabicyclo[3.3.1]nonane product 4a with the regeneration of active catalytic species (BF 3 ·Et 2 O), which is exergonic by 5.1 kcal/ mol, thus completing the catalytic cycle. Of note, the calculated  ARTICLE results show that the free energy difference between 3a and 4a is 5.5 kcal/mol, which indicates that the ring-opening/ring-expansion process is thermodynamically favored since the sevenmembered intermediate 3a is less stable than the eight-membered product 4a. The calculations indicate that the C-C bond-breaking step is likely the rate-determining step in the right catalytic cycle, and the overall activation free energy for this reaction is 11.1 kcal/ mol. Besides, we carried out additional DFT studies on the stereoselectivity-determining transition states ( Supplementary  Fig. S9), and the results listed in Supplementary Fig. S16 suggested that (S,S,S,S)-3a was the favorable cycloaddition product, which was in agreement with the experimental observation (see Supplementary Information for more details).

Discussion
In summary, we have developed an unprecedented catalytic asymmetric protocol for the synthesis of chiral N-bridged [3.3.1] ring systems, which is an important family of pseudo-natural products that bear an intrinsically synthetic challenge. This is enabled by a fascinating cascade reaction via bifunctional phosphonium salt/Lewis acid relay catalysis. Given the easy availability of both 2-indole-vinyl ketone and cyclic azomethine ylide starting materials together with the simple-operating and transition-metal-free conditions, this protocol provides a general and modular platform to access a great diversity of pseudonatural azabicyclo[3.3.1]nonane compounds and their analogs with high chemical yields and excellent stereoselectivities. Preliminary biological activity studies indicate that these N-bridged [3.3.1] molecules are potential anticancer agents. Experimental and computational mechanistic investigations revealed that it proceeded through a complex [3 + 2] cyclization/ring-opening/ Friedel-Crafts cascade process enabled by phosphonium salt/ Lewis acid metal-free relay catalysis. This study is expected to stimulate a more systematic exploration of the related organic synthesis, particularly toward the synthesis of diverse chiral pseudo-natural products for drug discovery.  Fig. 8 Proposed reaction mechanism and DFT study. a Proposed reaction mechanism. b Free-energy profiles for the BF 3 -catalyzed ring-opening/ Friedel-Crafts cascade process. Energy values are given in kcal/mol. Bond lengths are given in angstrom. Cs 2 CO 3 (195 mg, 0.6 mmol) and catalyst P10 (13.9 mg, 0.01 mol), followed by the addition of Et 2 O (2.0 mL). The reaction mixture was stirred at −20°C for 72 h, and TLC show that the reaction was completed. Then, BF 3 ·Et 2 O (1.0 mmol) was added directly. After being stirred at room temperature for another 12 h, the reaction was quenched by saturated NaHCO 3 (aq) and extracted by ethyl acetate three times, and the combined organic phase was concentrated in vacuo, and the crude residue was purified by column chromatography on silica gel (hexane/ethyl acetate = 3/1) to afford target product 4a.

Methods
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Data relating to the characterization data of materials and products, general methods, optimization studies, mechanistic studies, mass spectrometry, HPLC and NMR spectra, computational studies, and biological activity studies are available in the Supplementary Information. The X-ray crystallographic coordinates for structures of rac-3a, rac-4a, 3a, 4a, ent-4a, 3a-1 reported in this article have been deposited and available free of charge from the Cambridge Crystallographic Data Centre (CCDC), under deposition CCDC numbers 2089968, 2053966, 2015156, 2019560, 2047395, and 2041571, respectively. All other data are available from the corresponding author upon request.
Received: 13 July 2021; Accepted: 16 December 2021; the National Key R&D Program of China (2018YFA0903500), the "1000-Youth Talents Program" (YJ201702), Fundamental Research Funds from Sichuan University (2020SCUNL108), Beijing National Laboratory for Molecular Sciences (BNLMS202101), the Hunan Provincial Natural Science Foundation (2021JJ40150) and the Fundamental Research Funds for the Central Universities. We also acknowledge the comprehensive training platform of the Specialized Laboratory in the College of Chemistry at Sichuan University and the Analysis and Testing Center of Sichuan University for compound testing. Computational work was supported by the Center for Computational Science and Engineering at Southern University of Science and Technology.