B(C₆F₅)₃-Catalyzed C−H Alkylation of N-Alkylamines Using Silicon Enolates without External Oxidant

Abstract

An efficient method for the coupling of N-alkylamines with silicon enolates to generate β-amino carbonyl compounds is disclosed. These reactions proceed by activation of α-amino C–H bonds by B(C₆F₅)₃, which likely generates a “frustrated” acid/base complex in the presence of large N-alkylamines. The transformation requires no external oxidant and releases hydrosilane as a by-product. The utility of this method is demonstrated in the late-stage functionalization of bioactive molecules such as citalopram, atomoxetine, and fluoxetine.

Graphical Abstract

Enolate nucleophiles are commonly employed in addition reactions to C=N bonds (i.e., Mannich-type processes), since this process reliably generates several classes of important β-amino carbonyl molecules.^1–5 In these reactions, imine or iminium ion intermediates are either prepared in situ or in a separate operation through condensation of an amine and a carbonyl compound, α-fragmentation of iminium ion precursors, or oxidation of tertiary amines.^1–5 The latter method (i.e., oxidative Mannich-type reactions: Figure 1A) employs organometallic catalysts (e.g., Ru, Rh, Cu, V complexes) and stoichiometric oxidants (e.g., t-BuOOH, 2,3-dichloro-5,6-dicyano-1,4-benzoquinone, O₂) to convert N-alkylamines (I) to their corresponding iminium ions, which then react with various enolate equivalents (II).^2,3 Such a process, while notable, requires external oxidants, and the scope of amines is largely confined to tetrahydroisoquinoline and N,N-dimethylaniline derivatives.^2,3

Figure 1.

Coupling of N-Alkylamines and Silicon Enolates

We surmised that an attractive alternative to the oxidative Mannich-type reaction would entail rupture of an α-amino C–H bond of an N-alkylamine (1) by a boron-based Lewis acid (Figure 1B). Hydride abstraction from an N-alkylamine by organoborane compounds to furnish an iminium ion (IV) has been previously investigated.^6–12 An important advantage of this approach is that an array of N-alkylamines (including those that lack the fused N-aryl groups) can be converted to IV without the use of external oxidant.^6–10 Reaction of IV with an enol equivalent (2) would forge a C–C bond (V), subsequently releasing the β-amino carbonyl product 3, the Lewis acid catalyst, and a hydrosilane as an environmentally benign byproduct (VI). Here, we disclose the results of our studies regarding the realization of the above catalytic cycle.

To initiate our investigations, we needed an appropriate combination of acid catalyst and amine substrate that is capable of undergoing hydride transfer as opposed to forming a stable acid/base adduct. Accordingly, we considered pairing the strongly Lewis acidic B(C₆F₅)₃ with a hindered amine (Scheme 1). We first probed the ability of B(C₆F₅)₃ to convert a N,N-dimethylaniline to its derived iminium ion, which could then be trapped by a silyl ketene acetal (Scheme 1). Treatment of N,N-dimethylaniline (1a) and 1-methoxy-2-methyl-1-(trimethylsiloxy)propene (2a) with 20 mol % B(C₆F₅)₃ afforded 3a in 31% yield (DCE, 22 °C). We reasoned that electron-donating N-aryl substituents might improve the efficiency of B(C₆F₅)₃-catalyzed hydride abstraction by enhancing the hydride donor ability of the α-amino C–H bond and by stabilizing the resultant iminium ion intermediate. In the event, the reaction with electron-donating para-methoxy-substituted 1b afforded 3b in 25% yield. None of the desired product was observed with electron-deficient para-trifluoromethylphenyl-substituted 1c. With 3,5-di-tert-butyl-substituted 1d, the C–C bond forming product 3d was obtained in 38% yield.

Scheme 1.

Evaluation of N-Aryl Substituents ^a,b

^a Conditions: N,N-dimethylaniline (0.1 mmol), 1-methoxy-2-methyl-1-(trimethylsiloxy)propene (0.2 mmol), B(C₆F₅)₃ (20 mol %), dichloroethane (0.25 mL), under N₂, 22 °C, 12 h. ^b Yields were determined by ¹H NMR analysis of unpurified product mixtures with mesitylene as the internal standard. See the Supporting Information for details.

To evaluate the effect of using more hindered amines, we tested a range of ortho-disubstituted anilines. Whereas 2,6-difluoro-N,N-dimethylaniline (1e) gave none of the Mannich-type product, reaction of larger and more electron-rich N,N,2,6-tetramethylaniline (1f) resulted in the formation of 3f in 78% yield, which marks a considerable improvement in efficiency. Encouraged by this finding, we studied the reaction with 4-methoxy-N,N,2,6-tetramethylaniline (1g), which gave 3g in 56% yield; the N-aryl substituent in 3g was removed under oxidative conditions.¹³ By using the less hindered and more electron-deficient 2,6-difluoro-4-methoxyphenyl-substituted 1h, we were able to obtain 3h in 51% yield.

With the aim to further increase efficiency, we set out to identify the optimal conditions with the transformation that affords 3g, serving as the representative process (Table 1). There were no products formed without any B(C₆F₅)₃ present (entry 2).

Table 1.

Evaluation of Various Reaction Parameters ^a,b

entry	Lewis acid	catalyst loading (%)	solvent	yield (%)
				3g	4g
1	B(C6F5)3	20	DCE	56	35
2	none	0	DCE	0	0
3	B(C6F5)3	10	DCE	71	22
4	B(C6F5)3	10	Et2O	83	17
5	B(C6F5)3	5.0	Et2O	22	<5
6	B(C6F5)3	10	THF	38	<5
7	B(C6F5)3	10	toluene	75	21
8	B(C6F5)3	10	benzene	81	16
9	B(C6F5)3	5.0	benzene	>95	<5
10	BF3•OEt2	10	benzene	0	0
11	BPh3	10	benzene	0	0

^aConditions: 4-methoxy-N,N,2,6-tetramethylaniline (0.1 mmol), 1-methoxy-2-methyl-1-(trimethylsiloxy)propene (0.2 mmol), Lewis acid, solvent (0.25 mL), under N₂, 22 °C, 12 h.

^bYields were determined by ¹H NMR analysis of unpurified product mixtures with mesitylene as the internal standard. See the Supporting Information for details.

In some instances (entries 1 and 3–9), secondary amine 4g was also obtained, probably through the cleavage of the N–Me bond due to the reaction of in situ generated iminium ion with water. Among the N,N-dimethylanilines evaluated, only in the cases of ortho-dimethyl-substituted 1f and 1g were the latter type of byproducts formed. At lower catalyst loading (10 mol %) loss of the methyl unit was minimized and 3g was formed more selectively in 71% yield (entry 3). Next, we examined the effect of using ethereal solvents to investigate if these more polar solvents could facilitate the Mannich reaction which involves ionic intermediates. In diethyl ether, and with 10 mol % catalyst loading, 3g was formed in 83% yield, but with 5.0 mol % catalyst there was a considerable diminution in efficiency (22% yield, entries 4–5). With THF as the solvent, 3g was obtained in 38% yield (entry 6). Use of less polar aromatic hydrocarbons such as benzene and toluene as the solvent and with 10 mol % B(C₆F₅)₃ led to the formation of 3g in 81% and 75% yield, respectively (entries 7–8). With benzene and 5.0 mol % B(C₆F₅)₃ loading, 3g was obtained with high selectivity (<5% of 4g) and in >95% yield (entry 9). Use of less hindered BF₃•OEt₂ or less acidic BPh₃ proved to be ineffective (entries 10–11), providing support for the hypothesis that acidic B(C₆F₅)₃, along with sterically demanding and electron-rich N-alkylamines, represent the most effective catalyst/substrate combination.

A significant assortment of N,N-dialkylanilines may be used (Scheme 2). Reaction of N,N-dimethyl-substituted 1g afforded 3g in 80% yield after purification (5.0 mol % B(C₆F₅)₃). With N-benzyl and N-cyclopropylmethyl-substituted substrates, 3i and 3j were isolated in 64% and 56% yield, respectively; thus, in these instances, α-amino methylene C–H bonds remained intact. α-Amino C–H bond of N-arylpyrrolidine 1k could be converted to a C–C bond by the use of 10 mol % B(C₆F₅)_3, affording 3k in 90% yield. 1-(4-Methoxy-2,6-dimethylphenyl)pyrrolidin-3-ol (1l), possessing an unprotected hydroxyl group, furnished the desired products in their O-silylated forms 3l (31% yield, 4.0:1 trans:cis) and 3m (23% yield, >20:1 trans:cis).¹⁴ Reaction with α-methyl-substituted pyrrolidine delivered 3n in 70% yield and trans:cis of 7.0:1. When N-arylpiperidine 1o was used, 3o was isolated in 38% yield with the reaction being performed at 50 °C. A series of trialkyl-substituted amines that lack the fused N-aryl group were coupled efficiently with ((1-methoxyprop-1-en-1-yl)oxy)trimethylsilane (2b), leading to the formation of 5a−5d (88%–95% yield). Reaction of methyl (S)-3-(bis((S)-1-phenylethyl)amino)-2-methylpropanoate with 2b delivered 5d as a 1.6:1 mixture of diastereomers, which were separable through silica gel chromatography, allowing us to produce the β-amino esters in enantiomerically pure form (see the Supporting Information for details).

Scheme 2.

Evaluation of Different N-Alkylamines ^a,b

^a Conditions: N,N-dialkylaniline (0.2 mmol), 1-methoxy-2-methyl-1-(trimethylsiloxy)propene (0.4 mmol), B(C₆F₅)₃, solvent (0.5 mL), under N₂, 22 °C, 12 h. ^b Yield of purified products. ^c B(C₆F₅)₃ (5.0 mol %) was used. ^d Benzene (0.5 mL) was used. ^e THF (0.5 mL) was used. ^f B(C₆F₅)₃ (10 mol %) and benzene (0.5 mL) were used and reaction was performed at 50 °C. ^g Benzene (0.5 mL) was used and reaction was performed at 70 °C. See the Supporting Information for details.

Next, we explored the range of silicon enolates (Scheme 3). Cyclopentyl- and cyclohexyl-substituted variants reacted with 1g to give 3p and 3q in 73% and 79% yield, respectively. The less sterically demanding ketene acetal (2b) afforded 3r in 66% yield. A broader range of ketene acetals proved to be applicable to reactions with N-arylpyrrolidine. α-Cycloalkylesters and methyl propionate could thus be installed, furnishing the corresponding 2-substituted pyrrolidine products 3s−3u in 60% to >95% yield. Nucleophilic partners derived from isopropyl isobutyrate and 3-methyldihydrofuran-2(3H)-one were found to be compatible, as indicated by efficient synthesis of 3v (90% yield) and 3w (70% yield, 1.3:1 dr). In addition to silyl ketene acetals, trimethyl((1-(methylthio)vinyl)oxy)silane could be used, as the transformation affording β-amino thioester 3x illustrates (52% yield).

Scheme 3.

Evaluation of Various Silicon Enolates ^a,b

^a Conditions: N,N-dialkylaniline (0.2 mmol), silicon enolate (0.4 mmol), B(C₆F₅)₃ (10 mol %), solvent, under N₂, 22 °C, 12 h. ^b Yield of purified products. ^c Benzene (0.5 mL) was used. ^d THF (0.5 mL) was used. See the Supporting Information for details.

The catalytic protocol is scalable, as highlighted by the 1.0 mmol synthesis of 3k, which was obtained in 95% yield by the use of 5.0 mol % B(C₆F₅)₃ (Scheme 4A). The 4-methoxy-2,6-dimethylphenyl group of 3k could be readily removed under oxidative conditions to deliver 6k in 97% yield.¹³ This method is applicable in the late-stage functionalization of bioactive molecules containing an N-alkylamine moiety such as citalopram (antidepressant), atomoxetine (treatment for ADHD), and fluoxetine (antidepressant) (Scheme 4B). N-Methyl C–H bonds of these drug compounds were selectively converted to C–C bonds to afford a useful amount of β-amino carbonyl compounds 8a (23% yield, 0.30 g), 8b (25% yield, 92 mg), and 8c (27% yield, 113 mg).

Scheme 4.

Scale-Up Synthesis and Late-Stage Functionalization of Bioactive Molecules

^a See the Supporting Information for details.

In summary, we have developed a catalytic method for a B(C₆F₅)₃-catalyzed C–C bond forming process that provides access to an assortment of β-amino carbonyl compounds. Acyclic as well as cyclic amines may be used as pro-electrophiles. On the basis of our mechanistic hypotheses, it should be possible to develop an enantioselective version of this C–C bond forming reaction through design of a chiral Lewis acid catalyst,¹⁵ and also to expand the scope of nucleophiles. Investigations along these lines are currently underway.