To achieve chemodivergent molecular editing of N-unprotected indoles with fluoroalkyl carbenes, the site- and chemoselectivity control issues between the relatively low reactive C3 − H bond and innate reactivity of nucleophilic nitrogen atom resulting in a mixture of C2-, C3-, and/or N1-functionalization and cyclopropanation products must be overcome (Fig. 2A) (50, 51). We began by concentrating on the development of a selective skeletal editing reaction (see Table S1 for details). The initial reaction of 5-bromo-1H-indole (1, 2.0 equiv) and trifluoromethyl phenyl N-triftosylhydrazone (2, 1.0 equiv) with Rh2(OAc)4 and NaH in trifluorotoluene at 60°C produced the carbon insertion product, 3,4-dihydroquinoline (3) in 52% yield, along with C3 − H gem-difluorination product (4) in 10% yield (entry 1, Fig. 2B). The structure of 3 was explicitly confirmed by X-ray crystallography. Further optimizations reveal that TpBr3Ag(thf) was the most effective, delivering the desired carbon insertion product (4) in 96% isolated yield with three equivalents of 1 at 80°C (entry 3). Changing the ratio of 1 and 2 from 3:1 to 1:3 shut down the skeletal editing pathway while providing a 90% yield of the peripheral (i.e., a tandem cyclopropanation and N1 − H gem-difluoroolefination) functionalization product 5 (entry 4). The product yield was increased to 98% by running the reaction in DCE (1,2-dichloroethane) at 60°C (entry 5). Switching the NaH base with N,N-diisopropylethylamine (DIPEA) led to the selective C3 − H gem-difluoroolefination, producing product 4 in 40% yield (entry 6). A systematic survey of reaction parameters disclosed that the reaction of 1 (1.0 equiv) and 2 (2.0 equiv) with the Rh2(OAc)4/DIPEA catalytic system in trifluorotoluene at 25°C improved the product yield to 92% (entry 8). In this way, we were able to perform C3 − H gem-difluoroolefination of N-unprotected and N-aryl indoles, which was not possible by Koenigs palladium-catalyzed reaction using diazoalkanes as carbene precursors (52).
Using these optimized conditions, the substrate scope of the carbon insertion reaction was then investigated (Fig. 3A). Trifluoromethyl phenyl N-triftosylhydrazones featuring electron-donating (e.g., methyl, tert-butyl, methoxy, trifluoromethoxy), electron-withdrawing (e.g., trifluoromethyl, ester, vinyl, and nitro), and electon-nutral (e.g., phenyl and halogens) groups provided good to excellent yields of carbon insertion products (6–24). The presence of o-substituents (18–21) and a 3-nitro substituent (24) on the phenyl ring of N-triftosylhydrazone resulted in lower yields, likely due to increased steric hindrance. The disubstituted phenyl substrates were also well tolerated (25 and 26). 3,4-Dihydroquinoline products containing naphthyl (29) and heterocycles such as 1,3-benzodioxole (27), 2,3-dihydrobenzo[b][1,4]dioxine (28), furan (31), thiophene (32), dibenzothiophene (30), benzothiophene (33), and N-methyl indole (34), were obtained in moderate to good yields from the corresponding N-triftosylhydrazones. We also noticed that trifluoromethyl alkyl N-triftosylhydrazones reacted smoothly (35 and 36), albeit with lower yields due to competitive self-coupling of fluoroalkyl carbene. This carbon insertion methodology was also extended to electron-poor (e.g., esters, halogens, phenyl, formyl, acetyl, tosylate, and nitrile) and electron-rich (e.g., ethers, methyl, and tert-butyldimethylsilyl ethers) indoles and obtained the desired 3,4-dihydroquinoline products (37–54, 57) in moderate to good yields. Here, the reaction was shown to be sensitive to the position of substituents, with 4-substituted indoles giving lower yields than 5- or 6- or 7-substituted indoles. Indoles with an unsaturated unit performed well in the skeletal ring expansion chemistry, providing an additional functional handle for post-ring expansion functionalizations (55 and 56). Disubstituted indoles also proved to be suitable (58–60). Finally, we applied this skeletal ring expansion strategy in a one-pot, two-step reaction. Using commercially available indoles as starting materials, a tandem carbon insertion and reduction (with NaBH4) resulted in high yields of the corresponding tetrahydroquinoline products (61–63) without the isolation of intermediates (Fig. 3B).
The skeletal ring expansion of indoles could be scalable. The carbon insertion reaction on a 5 mmol scale of 2 with 5-bromoindole could still provide the corresponding 3,4-dihydroquinoline product 3 in good yield (1.4 g, 80%). The simple transformations of 3 could yield a variety of synthetically useful scaffolds, amplifying the synthetic utility of product 3 (Fig. 3C). Grignard reactions on the N = C bond in product 3 results in corresponding 2-arylated-, 2-allylated-, and 2-alkylated tetrahydroquinoline products (64–66) in high yields and good diastereomeric ratio (d.r.). Subjecting compound 3 to PINNICK oxidation conditions yielded the 3,4-dihydroquinolin-2(1H)-one (67). The bromo group in product 3 could undergo a variety of cross-coupling reactions. For example, the Buchwald-Hartwig cross-coupling of 3 with aniline furnished the desired N-phenyl-3,4-dihydroquinoline-6-amine (68), whereas the copper-catalyzed amination with aqueous ammonia yielded 6-amino-3,4-dihydroquinoline (69). This amino compound class is an important building block in pharmacology and materials science. The Suzuki-Miyaura coupling of 3 with benzofuran-2-boronic acid produced 6-(benzofuran-2-yl)-3,4-dihydroquinoline (70).
Next, we examined the scope and limitations of the developed methodology in C3 − H gem-difluoroolefination of indoles (Fig. 4A). An array of trifluoromethyl aryl N-triftosylhydrazones with various functional handles reacted well with N-unprotected indoles, affording the corresponding products (71–80) in moderate to good yields. Substrates containing disubstituted phenyl, naphthyl, and heterocyclic frameworks exclusively afforded the corresponding products (81–85) in good yields. N-Unprotected indoles with diverse substituents (e.g., methyl, fluoro, bromo, chloro, and cyano) on any of the positions were also successfully converted to the desired gem-difluoroolefination products (86–94). When compared to methyl and halogen groups, electron-withdrawing (e.g., fluoro and cyano) groups on indoles resulted in slightly lower yields of products (91 and 94). The α-C − F bond was predominantly activated when using N-triftosylhydrazone derived from pentafluoroethyl phenyl ketone, leading to 95 (78% yield), as confirmed by single-crystal X-ray analysis.
The applicability of this protocol with various N-substituted indoles was then assessed. Under the optimized conditions, a variety of aryl, naphthyl, alkyl, benzyl, propargyl, and allyl-protected indoles were also suitable substrates, affording the desired products (96–121) in good to excellent yields. The unsaturated group in the substrates remained intact, demonstrating the chemoselectivity of the process. N-Triftosylhydrazones derived from alkyl trifluoromethyl ketones could also be used as substrates, although low yields were observed (122 and 123). This transformation was compatible with a variety of functional groups. These gem-difluoroolefin products were difficult to obtain by existing methods (47).
Having established the optimal conditions for chemoselective assembling of N1-gem-difluorovinyl tetrahydrocyclopropa[b]indoles, we sought to apply the developed protocol to other substrates (Fig. 4B). Trifluoromethyl phenyl N-triftosylhydrazones with substituents on the 4-position (e.g., halogens, vinyl, trifluoromethyl, trifluoromethoxy) and a 3,5-dimethoxy substituted substrate reacted in moderate to good yields (124–130). Substrates derived from polycyclic rings such as naphthyl- and 2,3-dihydrobenzo[b][1,4]dioxine were found to be compatible in this reaction (141 and 138), paving the way for late-stage diversification of complex molecules. The tolerance of various functional groups on the indole core was then examined. Regardless of positions and electronic effects, indoles with methoxy, esters, and iodo groups underwent this tandem reaction to afford the corresponding products in high yields (132–136). The 5-fluoro-6-bromo indole was also smoothly transformed to the desired products 137 in 84% yield.
Finally, we pursued the development of a dearomative cyclopropanation reaction (44–46). After the systematic screening of reaction conditions, the subjection of N-methylindole (2.0 equiv) with 2 (1.0 equiv) in trifluorotoluene at 60°C in the presence of NaH and Rh2(OAc)4 afforded the corresponding cyclopropane (139) in 97% yield (Fig. 4C, See Table S2 for details). Applying the optimized conditions, we investigated the substrate scope of this dearomative cyclopropanation. As shown in Fig. 4C, a series of trifluoromethyl aryl N-triftosylhydrazones successfully participated in this reaction with N-methyl indole, affording N-methyl-tetrahydrocyclopropa[b]indoles (140–143) in high yields. Similarly, the N-triftosylhydrazones derived from disubstituted phenyl-, naphthyl-, piperonyl-, fluorenyl-, and furyl trifluoromethyl ketone were cyclopropanated (144–149). Both electron-donating (e.g., methyl) and electron-withdrawing (e.g., halogens, cyano, ester) groups on N-methyl indoles were compatible in the reaction (150–154). The little difference in yield confirms the effectiveness of our method in overriding inherent electronic preferences. N-Hexyl and N-benzyl indoles were also converted smoothly into corresponding cycloadducts 155 and 156 in 91% and 60% yield, respectively. The transformation was not limited to N-alkyl indoles; a variety of N-aryl indoles with various functionalities, N-naphthyl-, and N-pyrimidinyl indoles also performed well in this reaction (157–164). The identity of 163 was further confirmed by X-ray analysis. Other functional groups such as acetyl-, pivaloyl-, carbamoyl-, tosyl-, and TBS-protected indoles were found effective in producing tetrahydrocyclopropa[b]indole products (165–169). N-Triftosylhydrazone derived from pentafluoroethyl phenyl ketone was also a competent substrate, furnishing the corresponding product (170) in 83% yields.
The gram-scale synthesis of 108 (1.3 g, 96% yield) via C3-H gem-difluoroolefination also proceeded smoothly, demonstrating the scalability of the process (Fig. 5A). To demonstrate the synthetic utility of gem-difluoroalkenyl group, we performed a cyclization reaction between 108 and benzoyl hydrazine in the presence of Cs2CO3, which afforded an unsymmetrical 2,5-disubstituted 1,3,4-oxadiazoles (171) in 61% yield. Given the importance of trifluoromethyl and gem-difluorides in medicinal chemistry, the late-stage skeletal and peripheral editing of indole alkaloids could benefit the discovery of new drug analogous (Fig. 5B). Subjecting excess of Verticillatine B and Raputimonoindole B, two bioactive compounds isolated from neotropical plants, to our optimized TpBr3Ag/NaH catalytic system generated the corresponding carbon insertion products 172 and 175, respectively, whereas the C3-gem-difluoroolefination products 173 and 176 were isolated using Rh2(OAc)4/DIPEA catalytic system. Treatment of Verticillatine B and Raputimonoindole B with excess N-triftosylhydrazones 2 provided the tandem cyclopropanation and N1-gem-difluoroolefination products 174 and 177 under the TpBr3Ag/NaH catalytic system.
Mechanistic Investigations
To understand the origin of chemoselectivity, a series of control experiments were conducted using phenyl trifluoromethyldiazomethane as a carbene precursor (Fig. 6B). These results suggest that the chemoselectivity of the reaction is dependent on the base (NaH or DIPEA) used and the ratio of silver carbene and indole (Fig. 6B). No N1 − H functionalization was observed in all cases, which differs fundamentally from previously reported reactions of N-unprotected indoles with metal carbenes (44, 45). To gain deeper insight into the reaction mechanism and the origin of this unusual chemoselectivity, we performed density functional theory (DFT) calculations at the SMD(DCM)//B3LYP-GD3(BJ)/6–31 + G(d,p)-SDD(Ag, Br or Rh) level of theory (For details see Figs. S1 to S6). As shown in Fig. 6A, electrophilic metal carbene preferentially attacks the more nucleophilic C3 position of indole 178 to produce IntII-Ag and IntII-Rh. The calculated energy for TSI-Ag to IntII-Ag was 4.0 kcal mol− 1 lower than that for TSI-Ag', resulting in unfavourable N1 − H functionalization (Fig. 6D). Fukui function analysis (53, 54) was also performed on TSI-Ag and TSI-Ag′, which support the hypothesis that the indole C3 position is more nucleophilic than the N1 position (Fig. 6E).
When using DIPEA as the base, DIPEAH+-assisted β-F elimination of IntII-Rh preferentially occurred, where the hydrogen bonding (C − H···F and N − H···F in TSIII-Rh) between the DIPEA and CF3 group playing a dual role in lowering the energy barrier of β-F elimination process and assisting deprotonation to form C3-gem-difluoroolefination product 179. The lack of such hydrogen-bonding effect with the TpBr3Ag/NaH catalytic system favours 2,3-cyclopropanation of IntII-Ag over β-F elimination (Fig. 6C). In the presence of excessive indoles, the cyclopropane intermediate IntIII undergoes a tandem rate-determining hydrogen atom abstraction by NaH (ΔG≠ = 12.3 kcal mol− 1), reversible ring opening (ΔG≠ = 12.1 kcal mol− 1), and water-assisted protonation (ΔG≠ = 3.0 kcal mol− 1) to give the carbon insertion product 181 (Fig. 6A). When silver carbene IntI-Ag is excessive, the nucleophilic attack of IntIII on silver carbene IntI-Ag via TSIII-Ag' occurs to form kinetically favourable ylide IntIV', which then undergo β-F elimination to give N1-gem-difluoroolefination product 180 (for details see Fig. S4). The calculated energy for TSIII-Ag' is 0.6 kcal mol− 1 higher than that of TSI-Ag (Fig. 6D) and 10.4 kcal mol− 1 lower than that of TSIII-Ag. These observations are consistent with the experimental results that using an excess of indoles affords carbon-atom insertion product 181, and vice versa.