Given the prominence of fluoroalkyl groups in drug and agrochemical development42, 43, we first directed our attention to the insertion of fluoroalkylated carbon atoms into indoles, which would offer facile access to 3-fluoroalkylated quinolines. After optimization of various reaction parameters (See Tables S1 and S2), we found that the reaction of N-TBS-indole 1 with trifluoroacetaldehyde N-triftosylhydrazone (TFHZ-Tfs, 2)44 in the presence of NaH and the copper(I) catalyst TpBr3Cu(NCMe) (10 mol%), followed by in situ treatment of the intermediate cyclopropane 3' with TBAF and DDQ, directly afforded 3-(trifluoromethyl)quinoline 3 in 86% yield.
We next explored the scope of this one-pot, two-step skeletal ring expansion reaction of indoles with fluoroalkyl N-triftosylhydrazones. As shown in Fig. 2a, a range of N-TBS-indoles bearing electron-donating or electron-withdrawing substituents at the 4-, 5-, 6-, or 7-positions afforded the corresponding 3-trifluoromethyl quinoline products in good to excellent yields (5–22). Many functional groups, such as methyl (5–8), ester (9), acetyl (10), halogens (11, 18, 19, 21, and 22), ethers (12, 16, and 20), amine (13), phenyl (14), pyridine (15), and phenylethynyl (17) were well tolerated, although electron-donating substituents afforded slightly reduced yields of the 3-trifluoromethylated quinolines (e.g., 5–8, 12, 13, and 20). In contrast to this moderate electronic influence, steric factors can play a crucial role in the reaction; for example, 3-methyl-TBS-indole produced ring expansion product 4 in ~ 20% yield. Nonetheless, the extensive functional group compatibility, combined with the mild reaction conditions, offers much potential for the late-stage modification of complex molecules of biological relevance: for instance, a bioactive natural product, raputimonoindole B, as well as indoles derived from naturally occurring terpenes geraniol and perillyl alcohol, underwent smooth insertion to afford their corresponding quinoline homologs (23–25). In addition to trifluoroacetaldehyde N-triftosylhydrazone (TFHZ-Tfs), difluoroacetaldehyde N-triftosylhydrazone (DFHZ-Tfs)45 could also be used in our ring expansion protocol, affording 3-difluoromethylated quinolines (26–33) in high yields and with excellent chemoselectivity. Aside from indoles, the reaction of 7-azaindole with TFHZ-Tfs or DFHZ-Tfs also performed well to afford the corresponding ring expansion products 34 and 35.
We then investigated the reactivity of longer chain fluoroalkyl N-triftosylhydrazones. Under the standard conditions, the reaction of N-TBS-indole with pentafluoroethyl N-triftosylhydrazone produced the expected ring expansion product 36 in 60% yield along with small amounts (20%) of a hydrodefluorinative ring expansion product 37. Intrigued by the formation of 37, we re-optimized the reaction conditions to favor the formation of this hydrodefluorination product, which no longer required the addition of the oxidant DDQ (see Table S4); gratifyingly, we obtained 37 in 98% yield in the presence of caesium fluoride (CsF) in H2O/DMSO under air at 25 ℃. As shown in Fig. 2b, a wide variety of structurally diverse indoles with functionalities such as halogen, ester, ether, amine, methyl, alkynyl, and allyloxy substituents readily underwent this modified ring-expansion reaction, affording 3-fluoroalkylated quinoline products (38–48) in moderate to good yields, and could be extended to an azaindole (49). We found that longer-chain perfluoroalkyl N-triftosylhydrazones were also compatible with the transformation, producing the corresponding functionalized carbon-atom insertion products (50–55) in high yields, in which the fluorine atom at the α-carbon of the hydrazone chain has undergone selective defluorination.
Surprisingly, we found that the outcome of this reaction could be further tuned by treatment of the cyclopropane intermediate (3′) under identical conditions (CsF/H2O/DMSO under air), but at 40 ℃ instead of 25°C, which led to the formation of quinoline-3-carboxaldehyde (56) in 78% yield (Table S6). This unexpected defluorinative formylation affords a formal carbonylative insertion product, and turned out to be quite general, with a wide array of electronically differentiated indoles proving compatible with these reaction conditions. In all cases the corresponding quinoline-3-carboxaldehyde products (57–67) were obtained in high yields. Remarkably, indoles derived from naturally occurring terpenes such as geraniol and perillyl alcohol were also transformed smoothly into their corresponding quinoline-3-carboxaldehyde analogs 68 and 69 respectively, in spite of their potentially vulnerable ester and alkene functionalities, demonstrating the potential of our protocol for the late-stage modification of chemically-complex indole-containing bioactive molecules.
We recognized that the value of this chemistry would be significantly enhanced if other hydrazones could be applied, and turned our attention to the synthesis of 3-arylquinolines using aryl and heteroaryl aldehyde-derived N-triftosylhydrazones as carbene precursors. While the reaction of N-TBS-indole with 4-tolyl N-triftosylhydrazone under the previously optimized conditions delivered the desired ring expansion product 70 in only 20% yield, after significant re-optimization we were delighted to discover that TpBr3Ag(THF) improved this yield to 97% (Fig. 3a, see Table S6 for details). Under these new conditions, a wide variety of electronically differentiated aryl N-triftosylhydrazones were successfully inserted, with tolerance of functionalities such as phenyl, naphthyl, trifluoromethyl, trifluoromethoxy, halogens, and ethers (71–80). The practicality of our protocol was demonstrated by the gram-scale synthesis of 70. The extension of this method to N-triftosylhydrazones bearing a fused or hetero aromatic ring would be of great significance for drug discovery. We found that this indeed proved possible, with furan, benzofuran, benzothiophene, thiophene, and pyridine substituents all successfully introduced in high yields in the ring expansion to products (81–86). We could further expand the scope of N-triftosylhydrazone substitutents to alkyl, alkene, and even alkyne groups; these insertions also proceeded smoothly, producing the corresponding 3-alkyl-, alkenyl-, and alkynyl-quinolines (87–92) in good to excellent yields. The incorporation of these unsaturated moieties provides valuable functional handles for additional downstream synthetic transformations.
Having established the broad scope of the N-triftosylhydrazone partner that can be employed in this chemistry, we turned our attention to the indole substrate. To our delight, using 4-tolyl N-triftosylhydrazone as the carbene precursor, indoles adorned with bromo, nitro, ester, pyridine, allyloxy, hydroxy, amine, phenyl, halogen, methoxy, cyano, and acetal functionalities at any of the positions on the indole scaffold proved successful, affording the corresponding 3-arylquinolines (93–106) in moderate to excellent yields. Particularly notable is the successful insertion of sterically challenged 2,3-dimethyl- and 4-fluoro-5-OTBS-2-methyl-substituted indoles, which gave respectable yields of quinolines 107 and 108. Interestingly, OTBS (tert-butyldimethylsilyloxy) group can be tolerated in the first step, but is hydrolysed to hydroxyl in the second step, which provides a practical route to access hydroxyl-containing bioactive compounds (vide infra). In addition, 7-azaindole and 5-chloro-7-azaindole also delivered the corresponding ring expansion products (109 and 110). We also demonstrated the applicability of this silver-catalyzed protocol to the late-stage modification of bioactive indoles through the successful ring-expansion of tryptophol, melatonin, raputimonoindole B, and verticillatine B, which provided the respective ring-expansion products 111–114 in moderate yields. Key drug intermediates (e.g., pindolol), or indoles derived from steroids (e.g., pregnenolone) or terpenes (e.g., perillyl alcohol and geraniol) also proved good substrates, affording their 3-aryl quinoline analogs 115–118.
Of high importance in molecular editing is the ability to apply a single reaction to multiple heterocyclic cores. As such, we next addressed the application of our protocol to the transmutation of pyrrole into pyridine. Although N-TBS-protected pyrroles failed to provide the targeted cyclopropanation product using phenyl N-triftosylhydrazone, reaction of 1H-pyrrole, using Rh2(esp)2 as catalyst, afforded the ring expanded pyridine 119 in 48% yield (see Table S7). With suitable modified editing conditions established, we first investigated the range of N-triftosylhydrazone coupling partners that could be used. As shown in Fig. 3b, we found that a variety of mono- and di-substituted aryl N-triftosylhydrazones bearing trifluoromethyl, fluoro, ester, methyl, and chloro substituents directly afforded the corresponding 3-aryl pyridines (120–125). A brief survey of pyrrole scope revealed that 2-(trichloroacetyl)-1H-pyrrole provided the desired pyridine 126 in 59% yield, while 3-methyl-1H-pyrrole resulted in a 4:3 mixture of regioisomeric ring expansion products 127 and 128, in 56% yield.
To further demonstrate the utility of the ring-editing protocol, we targeted the streamlined syntheses of a number of bioactive quinolines of medicinal interest (Fig. 4). 3-Aryl quinolines are common motifs in such compounds, however their construction typically requires multiple steps for the synthesis and functionalization of the quinoline ring, which can limit development. Our methodology, on the other hand, is simple, easy to implement, and generally results in high yields of 3-aryl quinolines by direct transmutation of more readily-available indoles. For example, we successfully synthesized potential anticancer agents compounds 130 and 131 from indole 129 in a single step, in 77% and 53% yield respectively. Previous approaches required three step protocols starting from 3-methoxyaniline, which proceeded with overall yields of 3.3% (130) and 9.2% (131)46, clearly demonstrating the efficiency of our ring expansion protocol. Similarly, we were able to prepare quinoline 133, which is used for the treatment of hypopharyngeal cancer, in two steps from indole 132 with 70% overall yield via C-D insertion followed by Buchwald-Hartwig amination with morpholine. Previous routes employed 1,2-difluoro-4-nitrobenzene as starting material, and gave 133 in 47% overall yield via a five step protocol47. The efficiency of preparation of the antimycobacterial treatment adjuvant 134, previously prepared via a five-step synthetic protocol with an overall yield of 20%48, could be nearly doubled by employing the C-D insertion (overall yield 37%) while reducing the number of synthetic steps. Moreover, 3-arylquinoline-2-carboxaldehyde 137, a key intermediate for the synthesis of pharmaceuticals with antiproliferative activity (138) and anti-dengue activity (139), was previously obtained from isatin in three steps with 53% total yield49. However, using the herein developed methodology, compound 137 can be accessed in two steps from 2-methylindole 134 in 62% overall yield. Subsequently, compound 137 could be readily converted into 138 by a Claisen-Schmidt condensation or to compound 139 via a Cannizzaro reaction. Finally, this silver-catalyzed one-carbon insertion could also be used for de novo syntheses of anti-inflammatory compound 14050 in four steps with 38% overall yield from indole 134, and β-glucuronidase inhibitor 14251 in three steps with 40% overall yield from indole 141.
Mechanistic investigations
To gain insight into the mechanism of the halogen-free C-D reaction, we performed a series of experiments to study the reaction pathway for the formation of the various products (Fig. 5). First, trifluoromethyl cyclopropane 144 was isolated as a single diastereomer on reaction of indole 143 with TFHZ-Tfs (2) (Fig. 5a). Treatement of 144 with CsF in dry DMSO under a nitrogen atmosphere at room temperature for 10 minutes gave a mixture of 145 and 146 in a ~ 1:1 ratio (82%). Subjection of these products to additional CsF and water, under air at 40°C, resulted in the formation of formal formylation product 62 in 80% yield. This suggests that 1,4-dihydroquinoline 145 or 3,4-dihydroquinoline 146 are both key intermediates in the formation of the aldehyde product. Further, resubjection 3-(difluoromethyl)quinoline 29 to the same conditions did not afford any aldehyde 62, ruling out the possibility of difluoromethyl hydrolysis to the aldehyde.
We next performed a series of deuterium labeling studies (Fig. 5b). Treatment of deuterated cyclopropanated indole 147-d with TBAF in D2O resulted in the formation of the bis-deuterated 1,4-dihydroquinoline 148-d2 (80% D incorporation). Reaction of 148-d2 under the standard conditions (CsF / H2O, DMSO, air, 40°C) afforded 56-d with 74% D incorporation at the carbonyl carbon and 85% D retention at the C4-position. In contrast, reaction of the mono-deuterated 1,4-dihydroquinoline 148-d1 (99% D) afforded 56-d with 23% D incorporation at the carbonyl carbon and 75% D retention at the C4-position, indicating a kinetic isotope effect (kH/kD) of 3:1. These experiments demonstrate that the aldehyde hydrogen of 56 is derived from the C4-position of the 1,4-dihydroquinoline intermediate, and appear to suggest an internal 1,3-hydride shift, which may be the rate-determining step of the reaction. Interestingly, treatment of deuterated cyclopropane 149-d with TBAF in THF at 25°C under a nitrogen atmosphere afforded the deuterated quinoline 150 in 98% yield, with 40% D incorporation at the N1-position and 30% D incorporation at the C4-position, which indicated an imine-enamine tautomerization process. Susequently, 150 could be oxidized (by air), giving rise to 3-arylquinoline 70-d in 97% yield and with 10% D retention at the C4-position, suggesting that the direct carbon-insetion may proceed through an oxidative dehydrogenation of 1,4-dihydroquinoline. Finally, subjection of trifluoromethyl cyclopropane 147 to the standard reaction conditions, but with H218O instead of H2O, produced the 18O-incorporated quinoline-3-carboxaldehyde 56-18O, showing that the oxygen atom in the aldehyde is derived from water (Fig. 5c).
Based on these experiments, possible reaction pathways for the C-D editing of indoles with fluoroalkyl carbenes is shown in Fig. 5d. First, [2 + 1] cycloaddition of the indole with the in situ generated fluoroalkyl carbene generates the N-TBS protected cyclopropane intermediate I, which undergoes deprotection under the influence of TBAF or CsF. Concurrent or subsequent ring-opening yields the 3,4-dihydroquinoline anion II, which is protonated (by water) to furnish the 1,4-dihydroquinoline intermediate IV after imine-enamine tautomerization. From this key intermediate, two pathways are possible, depending on the reaction conditions. Firstly, intermediate IV (R = F) can undergo oxidative aromatization mediated by the strong oxidant DDQ to produce the 3-(trifluoromethyl)quinoline (3) (pathway I). Alternatively, the gem-difluoromethylene intermediate V can be generated via a base-mediated elimination of fluoride (pathway II). As evidenced by the deuterium labeling studies (vide supra), intermediate V (R = perfluoroalkyl group) undergoes a 1,3-hydride shift to deliver the defluorinated product 37. However, for intermediate V (R = F), oxa-Michael addition of water occurs to form intermediate VI, followed by HF elimination to generate intermediate VII. From VII, a 1,3-hydride shift occurs to produce an intermediate VIII, which eliminates another molecule of HF to furnish the quinoline-3-carboxaldehyde product 56.
To further support this proposed mechanism shown in Fig. 5d, we performed a computational study of the ring-opening of cyclopropane intermediate Int1. Nucleophilic attack of fluoride on the TBS group in Int1, via transition state TS1 (ΔG‡ = 15.2 kcal/mol), generates the zwitterionic intermediate Int3, which is exergonic by 28.3 kcal/mol (from Int-1) providing the thermodynamic driving force for the ring-opening (Fig. 6a). Protonation of Int3 with H2O via transition state TS2 (ΔG‡ = 7.8 kcal/mol) results in imine intermediate Int5, which can isomerize to the more stable enamine Int6 with a free energy release of ΔG0 = -3.6 kcal/mol. The formation of Int6 from Int4 is reversible. These calculated mechanism for ring-opening of Int1 are fully supported by our deuterium labelling studies on the ring-opening of deuterium-labeled cyclopropane 149 to afford the N1 and C3 deuterium labeled intermediate 150 (Fig. 5b). The 1,4-dihydroquinoline Int6 then undergoes a CsF-mediated 1,4-elimination of fluoride via TS3 to generate the gem-difluoromethylene intermediate Int8 with a modest barrier of 13.0 kcal/mol (Fig. 6a). The relatively low barrier is due to the F∙∙∙Cs interaction as well as C–H∙∙∙O and C–H∙∙∙F hydrogen bond interactions between the solvent and the indole that stabilizes transition state TS3, which results in facile elimination of HF. This facile formation of the gem-difluoromethylene is also supported by our experimental results (Fig. 5b). Subsequently, a 1,4-oxa-Michael addition of H2O to the gem-difluoromethylene Int-8 affords intermediate Int10 with an energy barrier of 6.2 kcal/mol (Fig. 6b). In turn, Int10 must overcome a relatively large energy barrier of 22.4 kcal/mol (via TS5) to eliminate CsF and subsequently form the enol intermediate Int11. A CsF/DMSO assisted 1,3-H shift via TS6 (ΔG‡ = 24.3 kcal/mol) produces Int12, which is the rate-determining step in the reaction. This calculated rate-determining step is in line with the experimental observation that C–H bond cleavage is involved in the rate limiting step (supported by its kinetic isotope effect, Fig. 5b). Finally, elimination of HF from Int12 furnishes the quinoline-3-carboxaldehyde product 56. The elimination of HF in the final step is assisted by CsF, lowering the barrier to ΔG‡ = 5.7 kcal/mol (Fig. 6b).
We also investigated computationally the origin of the difference in reaction outcomes between the CF3- and C2F5-substituted carbenes, via analysis of the key transition states that lead to defluorinative formylation insertion product 56 and hydrodefluorination insertion product 37 (Fig. 6c). Our calculations show that Int8 – generated from the CF3-substituted carbene – strongly prefers the Michael addition of H2O (via TS4, ΔG‡ = 6.2 kcal/mol) over CsF-assisted 1,3-H transfer that leads to defluorinative product 26 (via TS4', ΔG‡ = 18.1 kcal/mol) (Fig. 6c). The preference of the Michael addition over the 1,3-H shift is consistent with our experimental findings that 3-(difluoromethyl)quinoline 26 failed to give any of carbonylation product 56 under the standard conditions (Fig. S17). By contrast, for Int8-1 – generated from the C2F5-substituted carbene – opposite chemoselectivity is observed. For Int8-1, the 1,3-H transfer is significantly preferred over the Michael addition reaction (compare TS4-1': ΔG‡ = 1.2 kcal/mol vs. TS4-1: ΔG‡ = 7.2 kcal/mol). The origin of this reversed selectivity appears to derive from the difference in electronegativity between O and C atoms in TS4 (Δe = 2.005), which is larger than that in TS4-1 (Δe = 1.489). The large difference in electronegativity is due to the strong electron-withdrawing effect of the trifluoromethyl group compared to the fluorine atom, which also explains the relatively lower activation barrier of TS4 (Fig. S18). In addition, conformational analysis indicates that the energy barrier for the 1,3-H transfer may be influenced by the conformational change between the transition state and its precursor. More specifically, the electrostatic repulsion between the fluorine and oxygen atoms in transition state TS4-1' is stronger than those in TS4', making the change in conformation between TS4-1' and its precursor Int8-1 smaller than that between TS4' and Int8 (Fig. S18).
Taken together, our combined experimental and DFT calculation results indicate that the formal carbon-atom insertion proceeds through a cyclopropanation/fragmentation cascade to generate a key 1,4-dihydroquinoline intermediate, which can undergo an unprecedented defluorinative aromatization to deliver the hydrodefluorination insertion product 37 or the defluorinative carbonylation insertion product 56, depending on the carbene precursor and reaction conditions. In both processes, CsF and DMSO play critical roles in controlling the chemoselectivity and lowering the activation energy through transition state stabilization via hydrogen bonding interactions.