Short, enantioselective, gram-scale synthesis of (−)-zephyranthine†

A reasonable synthesis design by strategically integrating functional group manipulation into the ring system construction resulted in a short, enantioselective, gram-scale total synthesis of (−)-zephyranthine. The concise route includes a catalytic Michael/Michael cascade for the asymmetric synthesis of a penta-substituted cyclohexane with three contiguous stereogenic centers, a remarkable 8-step one-pot operation to easily assemble the zephyranthine tetracyclic skeleton, the regioselective construction of a double bond in the C ring and an asymmetric dihydroxylation. This synthesis is also flexible and paves a potential path to a variety of cyclohexylamine-fused tricyclic or polycyclic alkaloids.


Introduction
Lycorine-type alkaloids (e.g., 1-4; Fig. 1) 1 are members of an Amaryllidaceae alkaloid sub-class 2 and display useful biological properties, 3 including anticholinergic, antiviral, insect antifeedant, and antineoplastic activities, as well as other pharmacological properties. 4 These alkaloids have attracted substantial synthetic attention because of their tetracyclic core structure, multiple chiral centers and bioactivities. As a result, signicant effort has been devoted to assembling the tetracyclic skeleton of such alkaloids 5 and to the syntheses of the natural products themselves. [6][7][8][9][10] Unlike other members of the lycorine family, (À)-zephyranthine (1) 11 has only a limited number of syntheses reported for its fabrication, 6e,12 none of which detail a catalytic asymmetric approach. Herein, we report an efficient, enantioselective, gram-scale protocol for 1 that takes advantage of two one-pot reactions. The rst is a catalytic asymmetric double Michael addition to construct the C ring with three consecutive chiral centers. The second is a novel 8-step procedure involving double deacetalyzation, nitro group reduction to its corresponding amine, tandem double ring-closing reductive amination, and then double ester hydrolysis with subsequent tandem decarboxylation to give the tetracyclic skeleton of 1. Although we have successfully developed a remarkably facile route to 1, we encountered obstacles at a later stage. Unfortunately, the crucial regioselective construction of the C1-C2 double bond in the C ring was hindered by mutable substrates containing nitro groups or amine-type nitrogen atoms. However, this failure was counteracted with the successful, kinetically controlled regioselective enolization of the C ring ketone moiety.

Results and discussion
The catalytic double Michael addition of g,d-unsaturated-bketoester and nitroolen was previously developed by our group 13d for the asymmetric synthesis of multiple-substituted cyclohexanes bearing 3-5 stereogenic centers, which is expected to develop into the key step of a general method to stereoselectively synthesize a variety of cyclohexylamine-fused alkaloids, including (À)-zephyranthine and lepadiformine-type alkaloids. 14 Our simple retrosynthetic analysis of the target natural product (Scheme 1) revealed that penta-substituted cyclohexane 12, which arose from a catalytic asymmetric double Michael addition of 13 and 14, was likely a key intermediate that would result in the direct formation of tetracyclic ketone 10 via a one-pot operation. Subsequent regioselective construction of a double bond in the C ring, followed by dihydroxylation, would lead to 1.
The synthetic journey commenced with a catalytic asymmetric double Michael addition cascade reaction of 13 and 14, which was promoted by Evans' chiral nickel(II) catalyst (23). 13 Condition screening (Table 1) revealed that the 1st Michael addition, unlike that in the synthesis of (À)-stenine, 13d was very sluggish with the low conversion (<10%) even aer 10 days' reaction at room temperature with THF as solvent or in solventfree conditions; DCM as solvent brought about the fastest reaction that afforded 84% yield of the product with inadequate ee (entry 3) in 48 hours. Finally we found that PhMe as the solvent and Triton B as the base gave both the highest enantioselectivity (90%) and diastereoselectivity (>20 : 1), as well as high yield (85%) of penta-substituted cyclohexane 12, which was identied as the single isomer of an enol ester. Furthermore, NMR analysis showed that three consecutive chiral centers (in the C ring) had been correctly constructed so that 1 would be produced in the subsequent steps.
It can be speculated, as shown in Scheme 3, the steric and electronic effects caused by the g-ethyl group of the product (R 1 ¼ Et, for enantioselective synthesis of stenine) 13d of 1st Michael addition make the intermediate 24 (R 1 ¼ Et) a less active Michael acceptor in the 2nd Michael addition, therefore, a heterogeneous strong base (KOH/SiO 2 ) condition was required to promote this reaction while avoiding damage to the nitro group. Moreover, an isomerisation phenomenon was observed aer the 2nd Michael addition that the keto ester intermediate 25 gradually transformed into its enol ester isomer accompanied by inversion of conguration at the N a -carbon. Evidently, g,d-unsaturated-b-ketoester (R 1 ¼ H) in this work is more active, and the 2nd Michael addition as well as the subsequent isomerisation progressed rapidly and completed in Scheme 1 Retrosynthetic analysis of (À)-zephyranthine (1).
Scheme 2 Synthesis of g,d-unsaturated-b-ketoester 13 and nitroolefin 14. 20 minutes aer addition of Triton B to the reaction mixture upon completion of 1st Michael addition.
To conrm the absolute conguration of compound 12, its enol moiety was either benzyloxycarbonyl (Cbz)-or benzoyl (Bz)protected to prevent unwanted aldol reactions between the acarbon (C1) of the b-keto ester and the aldehydes that would form from the acetal moieties upon their subsequent deprotection. Cbz-protection was achieved by treating 12 with benzyl carbonochloridate (CbzCl) in the presence of NaH to give ester 27 in 90% yield. However, 27 was difficult to purify by recrystallization. By replacing CbzCl with benzoyl chloride (BzCl), similar esterication of 12 afforded benzoate 11 in quantitative yield. Aer recrystallization, the isomeric purity of 11 was greater than 99% ee, as determined by high performance liquid chromatography (HPLC). The absolute conguration of benzoate 11 was conrmed by X-ray crystallography (Scheme 4) with Cu-K a radiation.
Successful construction of the three contiguous stereogenic centers in the newly formed cyclohexane ring allowed us to begin synthesizing 1. Cyclization of 11 to form tetracyclic ketone 10 was accomplished through a multistep one-pot operation (Scheme 5), which began by treating 11 with HBr (1.0 equiv., 33% in HOAc) in HOAc-THF-H 2 O (5 : 1 : 1) at 50 C for 2 h to give dialdehyde 28. Subsequent reaction of 28 with zinc powder at room temperature overnight gave 29. Aer a simple ltration to remove the excess zinc and other solid substances, HCl (8.0 N, 100 equiv.) was added to the reaction mixture to hydrolyze 29 into 30. Tandem decarboxylation of 30 then delivered key intermediate 10 in a total yield of 57% via an eightstep one-pot synthesis.
The reaction of ketone 10 with lithium bis(trimethylsilyl) amide and Comins' reagent 17 at À78 C was a kinetically controlled regioselective enolization, which was followed by triation to afford enol triate 31 in 87% yield. This then underwent a palladium-promoted hydrogenolysis 18 to give 9 in 85% yield. In the last step, an attempt to avoid oxidative damage of the amino nitrogen atom was made by adding some acid to the reaction system; however, this failed owing to the deactivation of AD-mix-b under acidic conditions. Fortunately, the most conventional sharpless asymmetric dihydroxylation 19 of 9 with AD-mix-b under acid-free conditions proceeded smoothly and gave 1 in 67% isolated yield (76% yield of 1 and its diastereoisomer in a ratio of 7.2 : 1). Aer that, amide 32 was synthesized in 78% yield via a PhIO promoted oxidation 20 of 9. Our approach thus provided a formal synthesis of a number of other lycorine-type alkaloids 12a (Scheme 5), such as lycorine (2), 6d,e dihydrolycorine (3) 6d,e and a-dihydrocaranine (4). 6d,e To gain additional insight into the nature of the regioselective enolization of ketone 10, we conducted a theoretical study and the DFT quantum-chemical calculations (Scheme 6, see ESI † for details) revealed that the formation of intermediate 33a is kinetically favored over that of 33b.
Scheme 4 Protection of the enol hydroxy group of 12 and the absolute configuration of compound 11. Scheme 5 Total synthesis of (À)-zephyranthine (1) and synthesis of 32.
We also obtained 36, the double bond positional isomer of 9, from the same intermediate 10 that gave 9. This was accomplished through a 3-step chemical manipulation of the ketone moiety of the C ring (Scheme 7). Intermediate 10 was reduced with sodium borohydride in methanol to give secondary alcohol 34 (dr ¼ 1.2 : 1), which underwent mesylation and then DBUpromoted, thermodynamically controlled methanesulfonic acid elimination to afford 36 as a single regioisomer in 57% overall yield.
To conrm the proposed thermodynamically controlled process, we conducted DFT calculations (see ESI † for details) of elimination reactions of mesylate 35 as indicated in Scheme 8. For both 35a and 35b, the formation of olen 36 is more favorable than formation of 9 according to the free energy changes. 35a is more likely to undergo elimination than 35b to form compound 36 as less energy required. The calculation results supported our conclusion that the formation of 36 by elimination reaction of 35 (both 35a and 35b) is a thermodynamically controlled process.
We had also attempted other routes to regioselectively construct the double bond in the C ring, but these did not proceed as we expected (see ESI † for detailed informations). Ideally, deesterication of 37 could efficiently provide 9 (Scheme 9, top) and ensure that the double bond remained in the correct position (C1-C2); however, this reaction was unsuccessful. We did manage to convert the ester group of 37 into a carboxyl or aldehyde group, but the subsequent decarboxylation or deformylation failed. In addition, transformation of 12 to 38 (Scheme 9, bottom) could not be achieved through direct deesterication, and attempts to obtain 39 from 12 with the same one-pot protocol that gave 10 from 11 (Scheme 5) were also unsuccessful due to unwanted aldol reactions.

Conclusions
The natural product (À)-zephyranthine (1) was synthesized using a highly efficient and practical approach. Strategically integrating functional group manipulation into the ring system construction resulted in two, multi-step, one-pot reactions that greatly simplied the overall operation and improved its efficiency. From readily available 13 and 14, only six steps (18.7% overall isolated yield) were necessary to acquire 1 g of (À)-1. In addition, regioselective construction of the C ring double bond from 10 delivered 9 or 36 through kinetically or thermodynamically controlled pathways, respectively. This, together with the concise synthesis of amide 32, provided a exible and practical synthetic pathway for lycorine-type alkaloids and their analogs. The development of multistep one-pot reactions with greater efficiency and further applications in lepadiforminetype alkaloid syntheses are currently underway.

Data availability
All computational data associated with this article have been inserted in ESI.

Author contributions
H. Z. and J. C. conceived the idea. Y. Z. conducted the most of experiments. Y. Z., G. M., Q. W., S. Y. and X. Z co-synthesized part of substrates. H. Z. and J. C. co-wrote the paper. All the authors discussed the results and commented on the manuscript.

Conflicts of interest
There are no conicts to declare.