Sequential ring-closing metathesis–vinyl halide Heck cyclization reactions: access to the tetracyclic ring system of ervitsine

Abstract A chemoselective indole-templated ring-closing metathesis is used to assemble the cyclohepta[ b ]indole substructure of the indole alkaloid ervitsine. A subsequent intramolecular Heck coupling of the resulting alkene functionality with an amino-tethered vinyl halide accomplishes the closure of the unique 2-azabicyclo[4.3.1]decane framework of the alkaloid with concomitant incorporation of the exocyclic E -ethylidene substituent.


Introduction
Ervitsine 1 is a minor indole alkaloid isolated in 1977 from Pandaca boiteaui (Apocynaceae) 2 with an unique tetracyclic framework comprising a 2-azabicyclo[4.3.1]decane system fused to the indole ring and two exocyclic alkylidene (16-methylene and 20E-ethylidene) substituents. This complex architecture attracted the synthetic interest of research groups in the eighties and early nineties, resulting in a few approaches to the core structure 3,4 and a total synthesis based on biomimetic considerations. 5 Despite the variety of strategies used, all routes have in common the formation of the central carbocyclic ring in the last synthetic steps, either by cyclization of an iminium-type ion upon the indole 3-position (bond formed C 5 eC 7 , a) 3a,d,4,5  Our long-standing interest in the development of indole annulation methodologies led us to envisage a straightforward synthetic approach to the bridged ervitsine framework relying on an indoletemplated ring-closing metathesis (RCM) 6 to first construct the central seven-membered ring and a vinyl halide Heck cyclization 7 to close the piperidine ring and at the same time install the requisite 20E-ethylidene substituent (bond formed C 15 eC 20, c). 8 As shown in Scheme 1, the metathetic ring closure of 2,3dialkenylindoles of general structure A would provide cyclohepta [b]indoles B, with the appropriate double bond functionality for the subsequent intramolecular Heck reaction with the amino-tethered vinyl halide. 9 It should be noted that similar Heck couplings of vinyl halides and elaborated cyclohexenes 10 or cycloheptenes 11 have proved to be useful for the assembly of the bridged core of several indole alkaloids. In this context, we have successfully explored vinyl halide Heck reactions upon azocine and azonine rings for the total synthesis of apparicine 12 and cleavamines. 13

Results and discussion
To explore the feasibility of the double annulation RCMeHeck methodology for the ervitsine construction, we initially focused on indolic precursors unfunctionalized at the benzylic a-position (Scheme 1, Y¼H, H), knowing that this methylene group could be oxidized at a later stage of the synthesis. 14 Thus, cyclohepta [b]indoles 7e9 and 13 were selected as substrates for the key Heck reaction bearing different (carbamate, amine, and amide) exocyclic nitrogen atoms (Scheme 2). The synthetic route began with 2-allyl-3-indolecarbaldehyde 1, 15 which was equipped with a strong electron-withdrawing group at the nitrogen to guarantee the stability of the gramine [3-(aminomethyl)indole] moiety of the intermediates. From this compound, an aminationeimine allylation sequence was devised to install the homoallylic amine moiety required for the RCM step. We chose a direct route without using protecting groups and incorporated the additional haloalkenyl appendage either at the amination step (for 7e9) or at the final acylation step (for 13), with the hope that it would be sufficiently inert under the RCM conditions. Reaction of aldehyde 1 with (Z)-2bromo-2-butenylamine (2a), followed by alkylation of the resulting imine with allylmagnesium bromide led to the unstable secondary amine 3a (not isolated), which was subsequently acylated with ClCO 2 Me or alkylated with formaldehyde and NaCNBH 3 to give carbamate 4 or tertiary amine 6 in 60% and 50% overall yield, respectively. Starting from 1 and (Z)-2-iodo-2-butenylamine (2b), carbamate 5 was similarly prepared in 65% overall yield through secondary amine 3b. On the other hand, reaction of aldehyde 1 with methylamine and allylation of the resulting imine as above gave the unstable secondary amine 10, which was converted into amide 12 in 60% overall yield by acylation with (Z)-2-bromo-2-butenoic acid (11) in the presence of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC).
At this point we proceeded to study the RCM reaction. Considering the different substitution and electronic nature of the double bonds of the trienic substrates, we expected the preferred RCM event to be the indole-templated cyclization leading to a fused seven-membered ring. Our expectations were confirmed when carbamates 4 and 5 as well as amide 12, on exposure to the second generation Grubbs catalyst in refluxing CH 2 Cl 2 , gave the desired cycloheptenes 7, 8, and 13 as the only products in 80%, 78%, and 87% yield, respectively. The tertiary amine 6 was a worse RCM substrate, requiring the previous conversion into the corresponding hydrochloride to afford 9 in a slightly lower yield (65%).
We also sought to elaborate functionalized tricyclic ervitsine substructures (B, Y¼H, OH or O, Scheme 1) but all our efforts met with failure. The simple extension of the chemistry outlined above to an O-protected 2-(1-hydroxyallyl)indole such as 15 15 (Scheme 3) proved impractical as the formyl group required for the aminationeimine allylation step could not be introduced, only complex mixtures being obtained when 14 was subjected to the FriedeleCrafts protocol (Cl 2 CHOMe, TiCl 4  We then planned to install the homoallylic amine moiety on an indole-3-carbaldehyde such as 16 16 or 17 17 before functionalizing the 2-position either by direct metalation or metalehalogen exchange followed by electrophilic trapping (Scheme 4). Given that the base-sensitive halovinyl chain would probably be incompatible with the latter reaction, it would be introduced after the RCM step. Aldehyde 16 was uneventfully converted into carbamate 18 by successive treatment with methylamine, allylmagnesium bromide, and di-tert-butyl dicarbonate, but treatment of this substrate with LDA or alkyl lithium derivatives (s-BuLi, t-BuLi) in THF under a variety of experimental conditions, followed by addition of DMF, HCO 2 Me, or acrolein, only led to the recovery of the starting material. More satisfactorily, the desired functionalization took place by lithiumehalogen exchange from carbamate 19, which was prepared as above from aldehyde 17. Treatment with t-BuLi in THF at À78 C followed by quenching with acrolein led to an unstable alcohol, which was immediately oxidized with MnO 2 to give ketone 20 in 45% overall yield. Unfortunately, while RCM of 20 took place in the presence of the second generation Grubbs catalyst in refluxing CH 2 Cl 2 to give the expected cyclopentenone 21 in 65% yield (not optimized), all attempts to remove the Boc protecting group for the We turned our attention to the intramolecular Heck coupling to complete the bridged framework of ervitsine. Table 1 summarizes the survey of experimental conditions, including palladium precatalysts, ligands, and additives, using carbamates 7 and 8 as substrates. As can be observed in entry 1, only the starting product was recovered when vinyl bromide 7 was subjected to classical polar conditions 10a (Pd(OAc) 2 , PPh 3 , Et 3 N, CH 3 CN). On the other hand, the use of ligand-free conditions introduced by Jeffery 18 (Pd(OAc) 2 , K 2 CO 3 , TBACl, DMF, entry 2), which had proven successful for the synthesis of related azapolycyclic structures, 10b,c,e resulted in the total decomposition of the material. More successfully, the desired cyclization proceeded upon treatment of 7 under non-polar conditions 11 (palladium catalyst, PPh 3 , proton sponge, K 2 CO 3 , toluene, entries 3 and 4). However, although the conversion yields were good as evidenced by the NMR analysis of the crude reaction mixtures, the isolated yields of the (E)-ethylidene tetracycle 23 after column chromatography were only moderate (30%), the starting product being invariably recovered even under longer reaction times.
It should be mentioned that the analogous N-methyl derivative 9 (Scheme 5) led to complex reaction mixtures under any of the above Heck conditions. This result seemed to indicate that the presence of a basic nitrogen in the halobutene chain is not compatible with the harsh cyclization conditions, probably due to a competitive dealkylation process. So, unsurprisingly, amide 13 proved to be a more robust substrate leading to tetracyclic lactam 25 in 50% yield. We then focused on the more reactive vinyl iodide 8. When subjected to the same non-polar protocol (entry 5), tetracycle 23 was obtained only in a slightly better yield (45%) along with minor amounts of recovered starting product. To increase the efficiency of the process we examined the reaction in the presence of other additives, such as phenol or Ag 2 CO 3 . We were pleased to find that the addition of 20 mol % phenol in combination with K 3 PO 4 resulted in a cleaner cyclization, giving the ervitsine tetracycle 23 as the only product in 65% yield (entry 6). As far as we know, the use of phenol as a catalytic additive in the Heck reaction is unprecedented, although its positive role in some palladiumcatalyzed arylations of ketone enolates has been previously observed. 19,20 According to these reports, 19 the intermediacy of a palladium phenoxide (e.g., C), which would stabilize an otherwise unstable intermediate, could account for the beneficial effect of the added phenol.

N
On the other hand, although the starting material was rapidly consumed in the presence of Ag 2 CO 3 (entries 7e9), the cyclization followed a different course as it led to mixtures of tetracycles 23 and 24, the latter coming from an apparent 7-endo cyclization with inversion of the ethylidene configuration. 21 Tetracycle 23 was the major product when the reaction was carried out in refluxing toluene (entry 7) while the formation of the abnormal product 24 was enhanced by working at lower temperatures (entry 8) or changing the ligand from Ph 3 P to dppe (entry 9).
The formation of unusual Heck cyclization products like 24 has been previously observed 12b,22 and rationalized 23 by considering that the initial 6-exo cyclization is not followed by the expected b-hydride elimination (which would lead to 23) but by an intramolecular carbopalladation on the exocyclic alkene. The resulting cyclopropane intermediate would undergo rearrangement, with concomitant inversion of the alkene geometry, and final b-hydride elimination. In our case, the competitive formation of 24 is only observed in the presence of Ag 2 CO 3 , probably because under these cationic conditions the benzenesulfonyl group is able to weakly coordinate with the initially formed cationic s-alkyl palladium intermediate to give a sevenmembered palladacycle 24 (Scheme 6). The b-hydride elimination would thus be partially prevented and the intramolecular cyclopropanation route favored, in particular when the reaction is performed at a relatively low temperature or in the presence of a chelating phosphine such as dppe.

Conclusions
We have succeeded in synthesizing tetracycles 23 and 25, which embody the 4E-ethylidene-2-azabicyclo[4.3.1]decane bridged core of the indole alkaloid ervitsine, using a combination of an indoletemplated RCM and a vinyl halide Heck cyclization. This result highlights the power of the double annulation RCMeHeck methodology for rapidly building up the highly complex structure present in some indole alkaloids.

General
All nonaqueous reactions were carried out under an argon atmosphere. All solvents were dried by standard methods. Reaction courses and product mixtures were routinely monitored by TLC on SiO 2 (silica gel 60 F 254 ) and the spots were located with aqueous potassium permanganate solution. Drying of organic extracts was carried out over anhydrous Na 2 SO 4 . The solvents were evaporated under reduced pressure with a rotary evaporator. Column chromatography was carried out using the flash chromatography technique on SiO 2 (silica gel 60, SDS, 0.04e0.06 mm). NMR spectra were recorded in CDCl 3 using Me 4 Si as an internal reference. HRMS were obtained using an LC/MSD TOF mass spectrometer. 25 (3.12 g, 14.6 mmol) was added dropwise (1 h) to a solution of hexamethylenetetramine (2.25 g, 16 mmol) in CHCl 3 (18 mL) heated at reflux. The resulting mixture was heated at reflux for 4 h and then allowed to stand in the refrigerator overnight. The mixture was cooled in an ice bath and the quaternary salt was collected by filtration. The crude salt was dissolved in a warm solution, prepared from H 2 O (6 mL), EtOH (29 mL), and 37% HCl (8 mL). The mixture was stirred for 4 h and then allowed to stand overnight. A precipitate of NH 4 Cl was formed, which was removed by filtration, washing carefully with ethanol. The filtrate was concentrated to a quarter of the volume and the resulting solid was removed by filtration. The filtrate was concentrated to dryness and the solid residue was carefully dried. The residue was digested with MeOH (15 mL) and the resulting solid was removed by filtration. The filtrate was concentrated to dryness to give 2a hydrochloride (2.75 g, quantitative), which was used in the next reaction without purification. (4). Et 3 N (0.31 mL, 2.23mmol) was added to a solution of amine 2a hydrochloride (0.29 g, 1.5 mmol) in CH 2 Cl 2 (5 mL) and the mixture was stirred at rt for 10 min. Aldehyde 1 15 (0.34 g, 1.0 mmol) in CH 2 Cl 2 (5 mL) and AcOH (0.06 mL, 1.0 mmol) were successively added and the resulting mixture was stirred at rt for 18 h. The reaction mixture was diluted with CH 2 Cl 2 (5 mL), basified with a saturated aqueous Na 2 CO 3 solution (10 mL), and extracted with CH 2 Cl 2 (2Â10 mL). The organic extracts were dried and concentrated to give the crude imine (480 mg). Allylmagnesium bromide (1 M in Et 2 O, 1.6 mL, 1.6 mmol) was added under Ar to a cooled (À78 C) solution of the above imine in anhydrous THF (30 mL), and the resulting mixture was stirred at rt for 2 h. The reaction mixture was quenched with a 10% aqueous NH 4 Cl solution (5 mL) and extracted with Et 2 O (3Â10 mL). The ethereal extracts were dried and concentrated to give the crude amine 3a (342 mg). A solution of the above amine 3a in anhydrous THF (12 mL) was added under Ar to a suspension of NaH (60%, 56 mg, 1.4 mmol) in THF (2 mL) cooled at À20 C, and the mixture was stirred at À20 C for 20 min.   (6). Aldehyde 1 15 (0.17 g, 0.5 mmol) was allowed to react with 2a hydrochloride and allylmagnesium bromide as described for the preparation of carbamate 4. The resulting crude amine 3a (170 mg) was dissolved in CH 3 CN (1.5 mL) and the resulting solution was treated with 37% aqueous formaldehyde (1.7 mmol) and NaBH 3 CN (34 mg, 0.55 mmol) for 45 min at rt. The acidic pH was maintained with regular addition of AcOH. The reaction mixture was basified with 2 N NaOH (5 mL), diluted with H 2 O (10 mL), and extracted with Et 2 O (3Â10 mL). The organic layer was washed with 2 N NaOH (2Â10 mL), dried, and concentrated. The resulting residue was chromatographed (90:10 hexanes/AcOEt) to give 6 as a light brown oil: 134 mg (50%); 1