Titanocene(III)-Catalyzed Formation of Indolines and Azaindolines

Abstract

Reductive cyclization of epoxides tethered to substituted anilines and aminopyridines in the presence of 3 mol % of titanocene(III) chloride and stoichiometric manganese metal promotes a radical annulation to form 3,3-disubstituted indolines and azaindolines.

Graphical abstract

The continued development of new methodology for the construction of indolines^1,2 and azaindolines³ is well justified by the tremendous therapeutic potential associated with these heterocyclic building blocks.⁴ Natural products containing the 3,3-disubstituted indoline motif include the clinically used antitumor agent vinblastine (1),⁵ the novel marine metabolite diazonamide A (2),⁶ and the cage-like structures of kopsidarine (3),⁷ and scholarisine A (4)⁸ (Figure 1).

Figure 1.

Representative natural products with 3,3-disubstituted indoline scaffolds (outlined in bold).

Alkyl radical cyclization onto pyridine rings to prepare azaindolines is common;⁹ in contrast, methods to construct indolines by alkyl radical cyclization onto the aromatic nucleus without using xanthate transfer are scarce.^2b,c While the vast majority of known protocols for indoline synthesis utilize nucleophilic, Lewis acid and transition metal cross-coupling techniques, we envisioned a radical process that installs a quaternary carbon simultaneous with pyrrolidine formation from readily available epoxidized allylic amines.¹⁰ The reductive opening of epoxides using either stoichiometric¹¹ or catalytic¹² titanocene(III) chloride was particularly attractive toward this goal. Nearly two decades after the first reports of using in situ prepared titanocene(III) chloride to generate reactive radical intermediates from epoxides, novel applications of this reagent continue to emerge.¹³

To test the potential of this methodology in the synthesis of 3,3-disubstituted indoline and azaindoline heterocycles, epoxide 4a¹⁴ was subjected to 10 mol % of Cp₂TiCl₂ at room temperature in degassed THF (0.1 M) in the presence of 2 equiv of zinc dust. However, these conditions gave only traces of the desired indoline and mainly undesired side products. Fortunately, upon lowering the concentration of the substrate to 0.03 M and using 10 mol % of precatalyst in the presence of 0.80 equiv of 20–50 mesh manganese powder under sonication¹⁵ at room temperature, a 3:1 ratio of indoline 5a and tetrahydroquinoline 6a was detected by GC analysis of the crude reaction mixture (Table 1, entry 1). Sub-stoichiometric amounts of manganese metal increased the ratio in favor of indoline 5a; a possible indication of reversible radical pathways.^12c Addition of a methyl group to the epoxide in substrate 4b (R²=CH₃) improved the efficiency of the titanocene(III) chloride catalyzed process and afforded 5b as a single product in 82% yield (entry 2). With 1,1-disubstituted epoxides, sonication was not essential and conventional magnetic stirring also afforded the product in good yield (entry 3).

Table 1.

Optimization of the titanocene(III) chloride catalyzed reductive cyclization of anilines.^a,^b

entry	epoxide	Cp2TiCl2 (mol %)	Mn (equiv)	concentration [M]	R1	R2	product(s), yield
1	4a	10	0.80	0.03	Ph	H	5a:6a (3:1)c,d
2	4b	10	0.65	0.03	Ph	CH3	5b, 82%d
3	4b	10	0.65	0.03	Ph	CH3	5b, 84%e
4	4c	10	1.5	0.1	Cbz	CH3	5c:7c (2:1)f
5	4c	3	1.5	0.1	Cbz	CH3	8c, 63%g
6	4b	3	1.5	0.1	Ph	CH3	5b, 89%
7	4a	3	1.5	0.1	Ph	H	5a:6a (3.8:1), 87%
8	4c	3	1.5	0.1	Cbz	CH3	5c, 14%h
9	4c	3	0	0.1	Cbz	CH3	5c, 0%
10	4c	0	1.5	0.1	Cbz	CH3	5c, 0%

^aFor epoxide preparation, see Supporting Information.

^bAll reactions were performed in degassed THF heated at reflux unless otherwise noted.

^cYield not determined; product ratio was determined by GC analysis of crude reaction mixtures.

^dReaction was performed in degassed THF at room temperature using sonication.

^eReaction was performed in degassed THF at room temperature using magnetic stirring.

^fYield not determined; product ratios were determined by ¹H NMR analysis of crude reaction mixtures.

^gYield was determined over 2 steps.

^hStarting material 4c was recovered in 43% yield.

Model substrates 4a and 4b were designed to facilitate the cyclization by the presence of two symmetrical N-phenyl substituents. However, in order to broaden the scope of this reaction, we intended to replace one of them with a suitable nitrogen protective group. For reasons that are not completely clear, the secondary amine (R¹=H) mostly decomposed in the reaction mixture, and only a minor amount of reduced amino alcohol was formed. Alternatively, among protecting groups at this position, including the p-toluenesulfonyl, benzyl, trifluoroacetyl, and t-butoxycarbonyl functions, the benzylcarbamate (Cbz) group proved to be the most versatile substituent after we re-optimized the reaction parameters. At room temperature, anilide 4c afforded a ~2:1 mixture of 5c:7c at 0.1 M concentration in the presence of 10 mol % of titanocene (Table 1, entry 4). The undesired reduced epoxide 7c could be suppressed by lowering the precatalyst loading to 3 mol % of Cp₂TiCl₂ while increasing the reaction temperature to THF at reflux.¹⁶ Under these conditions, indoline 8c was isolated in 63% yield over 2 steps (entry 5). When these optimized conditions were applied to the earlier model system 5a, indoline 5b was formed in 89% yield (entry 6). High yields were also obtained for epoxide 4a, which provided a 3.8:1 mixture of 5a and 6a in 87% yield. We also explored the sensitivity of the conversion to air. When the epoxide-opening rearrangement with substrate 4c was performed in a flask kept open to the atmosphere in non-degassed THF, the reactivity was greatly diminished (entry 8). Additional control experiments using either solely the precatalyst (entry 9) or the manganese metal (entry 10) under otherwise identical reaction conditions failed to afford indoline products.

The general scope of indoline formation was further illustrated by the conversions summarized in Table 2. Alkyl-substituted substrates, including tetrahydroquinoline 15 afforded the corresponding indolines in modest to good yields (entries 2–4). The electron-rich 5-methoxyindoline 18 was isolated in low yield (21% over 2 steps). Electron-deficient substrates underwent the epoxide-opening rearrangement to afford substituted indolines in good yields (entries 6 and 7). However, an attempt to prepare the tricyclic pyrroloindole 24 from epoxide 23 did not yield any of the cyclized product (entry 8). Meta-substituted anilines generally led to 1:1 mixtures of regioisomeric indolines (data not shown).

Table 2.

Indolines prepared using titanocene(III) catalysis

entry	epoxide	indoline	yield
1			65%
2			62%a
3			35%a
4			69%a
5			21%a
6			56%a
7			41%
8			–

^aYields determined over 2 steps.

Azaindoles are attractive bioisosteres of indoles in pharmaceutical research.¹⁷ We therefore investigated the possibility of applying this methodology toward the formation of azaindolines, using aminopyridines in place of anilines. Chemoselective epoxidation of alkenes in the presence of a pyridine ring is precedented, however, attempts to efficiently prepare the epoxide on an unsubstituted aminopyridine substrate were unsuccessful. This was primarily due to the reactivity of the nuclephilic pyridine nitrogen toward either m-CPBA or DMDO.¹⁸ In contrast, an ortho-chlorine substitution¹⁹ proved to be sufficient to attenuate the nucleophilic character of the pyridine nitrogen. Curtius rearrangement of the known carboxylic acid 25²⁰ and trapping of the intermediate isocyanate with benzyl alcohol afforded aminopyridine 26 (Scheme 1). Subsequent methallylation of the Cbz-protected amine and epoxidation with m-CPBA led to epoxide 27. Under the optimized titanocene(III) chloride conditions, cyclization provided an intermdiate 4,6-dichloro-5-azaindoline, which was concurrently deprotected and dechlorinated with Pd/C under an atmosphere of H₂ to give azaindoline 28 in 52% yield over 2 steps.

Scheme 1.

Preparation of 5-azaindoline 28

As a mechanism for the titanocene(III) chloride catalyzed epoxide-opening annulation, we propose that the formation of the β-titanoxy radical 30^11,12 is followed by a reversible cyclization onto the aromatic ring, forming the cyclohexadienyl radical intermediate 31 (Figure 2).²¹ Oxidation of the dienyl radical by trace amounts of dioxygen and proton loss affords the indoline 32.²² Furthermore, protodemetallation by collidinium hydrochloride leads to product 33, and presumably regenerates the precatalyst Cp₂TiCl₂.²³

Figure 2.

Proposed catalytic cycle for the titanocene(III) chloride catalyzed epoxide-opening annulation.

The titanocene(III) chloride catalyst or the manganese byproducts may also serve as a Lewis acid to promote an epoxide-opening Friedel-Crafts alkylation. However, 5-membered ring benzene annulations by epoxide openings under Friedel-Crafts conditions are quite inefficient,²⁴ and a stoichiometric amount of reducing agent is required to drive the titanocene(III) chloride catalyzed process to completion.²³ Furthermore, it has previously been shown that the MnCl₂ produced in the reaction is not Lewis-acidic enough to promote epoxide-opening reactions.^12a

In summary, we have developed a novel titanocene(III) chloride catalyzed epoxide-opening arene annulation that affords 3,3-disubstituted indolines and tolerates a range of substituents on the aromatic ring. This methodology can be extended toward other five-membered heterocycles, as demonstrated by the preparation of a 3,3-disubstituted 5-azaindoline.