ReportsChemistry

Enantioselective cyanation of benzylic C–H bonds via copper-catalyzed radical relay

See allHide authors and affiliations

Science  02 Sep 2016:
Vol. 353, Issue 6303, pp. 1014-1018
DOI: 10.1126/science.aaf7783

Abstract

Direct methods for stereoselective functionalization of sp3-hybridized carbon–hydrogen [C(sp3)–H] bonds in complex organic molecules could facilitate much more efficient preparation of therapeutics and agrochemicals. Here, we report a copper-catalyzed radical relay pathway for enantioselective conversion of benzylic C–H bonds into benzylic nitriles. Hydrogen-atom abstraction affords an achiral benzylic radical that undergoes asymmetric C(sp3)–CN bond formation upon reaction with a chiral copper catalyst. The reactions proceed efficiently at room temperature with the benzylic substrate as limiting reagent, exhibit broad substrate scope with high enantioselectivity (typically 90 to 99% enantiomeric excess), and afford products that are key precursors to important bioactive molecules. Mechanistic studies provide evidence for diffusible organic radicals and highlight the difference between these reactions and C–H oxidations mediated by enzymes and other catalysts that operate via radical rebound pathways.

The clinical success of potential drug candidates has been shown to correlate with the fraction of sp3-hybridized carbon atoms and the number of stereogenic centers in the molecule (1). Consequently, synthetic methods that enable stereoselective functionalization of C(sp3)–H bonds could have a major effect on the discovery and development of new pharmaceuticals. Enzymatic oxygenases and halogenases, such as cytochrome P450 and nonheme iron enzymes (24), have inspired the discovery of synthetic methods for selective oxidation of C–H bonds [see, for example, (58)]. These enzymes and related small-molecule catalysts function via generation of a reactive metal-oxo species that abstracts an H atom from a C(sp3)–H bond, followed by radical rebound (9) to the metal-bound hydroxide or halide group (Fig. 1A). Here, we report a complementary radical relay strategy for C(sp3)–H oxidation whereby a reactive radical () abstracts a hydrogen atom from an sp3-hybridized carbon atom to produce a diffusible organic radical. This radical is then trapped by a second species that generates the functionalized sp3 center (Fig. 1B). This concept is demonstrated in copper-catalyzed enantioselective cyanation of benzylic C–H bonds (Fig. 1C). Optically pure alkyl nitriles are featured in many bioactive natural products, pharmaceuticals, and agricultural chemicals (10). Moreover, benzylic nitriles are direct precursors to phenethylamine derivatives, which represent a broad class of hormones, neurotransmitters, and psychoactive drugs (11), as well as arylacetic acid derivatives, which are among the most widely used nonsteroidal anti-inflammatory drugs (NSAIDs) (12). The stereoselectivity is proposed to arise from reaction of an achiral carbon radical with a chiral copper(II) species (L*CuII–CN) (13), a pathway that contrasts with most enantioselective C–H functionalization reactions that involve selective, stereospecific activation of a prochiral C–H bond within an organic molecule (14, 15).

Fig. 1 Strategies for hydrogen-atom-transfer–mediated C–H oxidation.

(A) Radical rebound mechanism for enzymatic and biomimetic C–H hydroxylation and halogenation (X = halide). (B) Radical relay pathway for C–H oxidation involving formation of a free radical that is trapped by a reactive metal center. (C) Proposed mechanism for Cu-catalyzed enantioselective cyanation of benzylic C–H bonds via radical relay.

In recent years, we have been investigating Cu-catalyzed methods for oxidation of styrenes and other alkenes [for example, (16, 17)], as well as C–H bonds (18). The former reactions are proposed to proceed via reaction of a benzylic radical intermediate with copper(II), generating an organocopper(III) species that undergoes reductive elimination to afford a new C–C or C–heteroatom bond. This pathway, which resembles a number of recently reported nickel- and copper-catalyzed cross-coupling methods that involve the reaction of a carbon-centered radical with a transition-metal center [e.g., (13, 1923)], prompted us to consider how analogous concepts could be applied to C–H oxidation. Kharasch-Sosnovsky allylic oxidations and related reactions provide relevant precedent for this idea (24, 25); however, these reactions face constraints that have greatly limited their synthetic utility over the nearly 60-year history of the reaction. For example, the alkene substrate is almost always used in large excess relative to the peroxide-based oxidant, a feature that severely restricts the net product yield and, in turn, the scope of suitable substrates. The majority of examples feature only simple cyclic alkene substrates, such as cyclopentene and cyclohexene. Reports of enantioselective reactivity retain this limited substrate scope and are characterized by poor catalytic efficiency that can result in multiday reaction times (26, 27). A synthetically useful radical-relay C–H oxidation method would need to overcome these limitations.

The pharmaceutical interest in benzylic nitriles and products derived therefrom, together with the formation of benzylic nitriles in certain alkene difunctionalization reactions (17, 28) [for broader context, see (29)], prompted us to target a method for enanatioselective cyanation of benzylic C–H bonds. Initial efforts focused on generation of benzylic radicals via hydrogen-atom abstraction in the presence of copper(II) and a cyanide source. We posited that reductive activation of peroxides and other diheteroatom-bonded oxidants (see X–Z, Fig. 1C) with a copper(I) source should afford a reactive radical species and copper(II). A series of oxidants, including tBuO2H, (tBuO)2, PhCO3tBu, (PhCO2)2, oxone, N-fluorobenzenesulfonylimide (NFSI), and Selectfluor-BF4, were tested in the cyanation of 1-ethylnaphthylene, 1a, with trimethylsilyl cyanide (TMSCN) as cyanide source and [Cu(MeCN)4]PF6 as the copper(I) catalyst (see the supplementary materials for screening data). An achiral bis(oxazoline) ligand (L1) was included in the reactions, but virtually all of the reactions failed to give the desired cyanation product. NFSI was the only oxidant that afforded any amount of the benzylic nitrile (2a). The yield increased upon changing the solvent from acetonitrile to chlorobenzene and using CuOAc as the copper source (L1, 25% yield) (Fig. 2). Replacing the ligand L1 with an enantiomerically pure, benzyl-substituted bisoxazoline (BOX) derivative L2 and performing the reaction in benzene led to a significant improvement in the yield (up to 91%) and also afforded product 2a in excellent (96%) enantiomeric excess (ee) (Fig. 2). In each of the reactions in Fig. 2, small quantities (trace to 12%) of benzylimide product derived from coupling with the bis(sulfonyl)imide of NFSI (30) were detected by proton nuclear magnetic resonance (1H NMR) spectroscopy.

Fig. 2 Reaction optimization and representative ligand screening data.

General reaction conditions: 1a (0.2 mmol), trimethylsilyl cyanide (TMSCN) (0.6 mmol), NFSI (0.3 mmol), CuOAc (0.02 mmol), and ligand (0.024 mmol) in 1.0 mL of solvent at r.t. for 10 hours. Yield determined by 1H NMR, CF3CONMe2 as internal standard. The ee value was determined by high-performance liquid chromatography (HPLC) with a chiral stationary phase. *Reaction conducted in chlorobenzene.

Assessment of various ligands showed that BOX ligands were the most effective ligand class, and testing of other BOX-ligand variants led to further insights and improvements (Fig. 2). Replacement of the gem-dimethyl substituents with a spirocyclic cyclopentyl group (L3) improved the enantioselectivity to 95% ee, whereas use of tert-butyl substituents on the oxazolines (L4), rather than benzyl groups, led to a significant dropoff in yield and enantioselectivity. Only trace product was observed with a ligand having an unsubstituted methylene linker between the two oxazolines (L5). Some of the best enantioselectivities were obtained with indane-derived BOX ligands, in particular, L6 and L8.

In an effort to explore the scope of the benzylic cyanation reaction, three conditions identified in Fig. 2 were tested with a wide range of alkylarene substrates (Fig. 3). Simple 1-alkyl–substituted naphthylenes proved to be good substrates, providing the desired products 2a to 2c with excellent enantioselectivities (96 to 97% ee) in yields of 72 to 91%. Product 2d, with a free primary alcohol, was obtained in excellent enantioselectivity (95% ee), but the reaction proceeded with only moderate conversion, resulting in a 42% product yield with 52% unreacted starting material. The analogous homobenzylic acetate derivative, however, exhibited very good reactivity (87% yield of 2e), while retaining excellent enantioselectivity (99% ee). Versatile functional groups appended to the aliphatic chain, such as chloride, azide, phthalimide (Nphth), and carboxylic ester, were very well tolerated (2f to 2j). A trifluoromethane sulfonate (OTf) group on the aromatic ring did not interfere with the reaction (products 2k and 2m) and provides a basis for subsequent derivatization of the product into more-complex molecules. The compatibility of these potentially sensitive functional groups included a primary alkyl bromide precursor to a CCR1 receptor antagonist of interest for treatment of chronic inflammation (31), highlighting the mildness of reaction conditions (Fig. 3B, i). 2-Alkyl naphthylenes also proved to be good substrates for the enantioselective cyanation, affording products 2l and 2m in good yield and enantioselectivity. Desymmetrization of acenaphthene (2n) was achieved with good enantioselectivity (79% ee).

Fig. 3 Substrate scope and synthetic applications of Cu-catalyzed cyanation of benzylic C–H bonds.

(A) Substrate scope. (B) Synthetic applications. *Reaction conditions: alkylarene as the limiting reagent (0.2 to 1 mmol scale), 10 mole % (mol %) CuI(OAc), 12 mol % ligand; 1.5 equivalent (equiv) NFSI, 3 equiv TMSCN, C6H6 (0.2 M substrate); under N2; isolated yields reported. Absolute stereochemical assignments are based on x-ray crystallographic characterization of (R)- and (S)-2h; see section 5 in the supplementary materials for details. Negative ee values reflect the opposite enantiomeric configuration relative to that depicted graphically. Chlorobenzene used as the solvent. ee value of product after recrystallization. §See (34).

Ethyl- and related alkylbenzene derivatives were similarly effective for this transformation (products 2o to 2z and 3a and 3b). The cases with reduced yields (e.g., 40 to 60%) typically reflect poor conversion, which would allow for recovery of valuable unreacted starting material, if desired. The enantioselectivity typically remained excellent (>90%). With substrates featuring more than one benzylic C–H site (e.g., 2n, 2r to 2t, 2y, and 3a), selectivity for the monocyanation product was observed, with consistently high ee. Good results were observed in the reaction of 4-isobutyl biphenyl (2z), demonstrating tolerance of branching in the homobenzylic position (see below for further discussion). Tertiary benzylic C–H bonds and substrates with larger substituents in the homobenzylic position were found to be unreactive with these catalyst systems, presumably because they present too much steric hindrance for effective hydrogen-atom abstraction. Numerous substrates containing pharmaceutically relevant heterocycles were tested, including those with a pyrazole, carbazole, benzothiazole, benzoxazole, thiophene, and pyridine (3c to 3i). Yields ranged from moderate to excellent (up to 89%), whereas enantioselectivities were consistently good to excellent (up to 98% ee). Alkyl phenol and anilide substrates were found to be unreactive. In these cases, the presence of weak O–H and N–H bonds could serve as radical traps. Benzene was typically used as the solvent in these reactions, on the basis of the initial optimization results, but a number of substrates were also tested in chlorobenzene as an alternate less toxic solvent, and excellent results were obtained (see 2a, 2c, 2e, 2f, 2h, 2p, 2r, 2v, 3c, and 3h). In addition, four substrates were demonstrated on a 1- to 5-g scale to establish practical laboratory-scale utility (2c, 2h, 3d, and 3h).

The results in Fig. 3 represent the most streamlined method to date for the preparation of enantiomerically enriched benzylic nitriles from readily available starting materials. Asymmetric hydrocyanation of vinylarenes is perhaps the most compelling alternative. Such methods have been known for some time (32) and are the subject of ongoing development (33), but the present methods exhibit advantages with respect to substrate accessibility (i.e., access to alkyl- versus vinyl-substituted arenes), functional-group compatibility and substrate scope, and enantioselectivity relative to asymmetric hydrocyanation of alkenes.

The enantiomerically enriched nitriles derived from these reactions are excellent precursors to phenethylamine and arylacetic acid derivatives. For example, benzylic nitriles 2d and 2h were subjected to hydrogenation over a Raney nickel catalyst, and the resulting primary amines were converted to a Boc (tert-butyloxycarbonyl) derivative before isolation. This sequence afforded the desired products in good yield with no loss in ee relative to the starting material (Fig. 3B, ii). Traditional methods for nitrile hydrolysis (e.g., in the presence of aqueous acid) erode the enantiomeric excess, but a number of biocatalytic processes are emerging for nitrile hydrolysis that proceed under very mild conditions (34). Although these enzymatic methods were not explored in the present study, they provide an excellent strategy for conversion of nitriles to arylacetic acids and could even provide a means to upgrade the enantioselectivity (Fig. 3B, iii).

Several observations suggest that site selectivity in these reactions is influenced by both steric and electronic effects (Fig. 4A). Subjection of 2-ethylfluorene 4b to the Cu-catalyzed cyanation conditions results in selective cyanation of the benzylic site of the 2-ethyl group; no cyanation of the highly activated, doubly benzylic methylene group of the fluorene is observed. We attribute this outcome to steric effects. The contribution of electronic effects is evident in the reactivity of the substituted phenethylnapthalene derivatives 4c to 4f. The site selectivity with these substrates is quite high, always favoring the benzylic position adjacent to the naphthalene, but the preference increases from 12:1 to 29:1 as the substituent at the para position changes from an electron-donating tert-butyl group to an electron-deficient –CF3 group. Excellent enantioselectivity (97 to 98% ee) is observed in each case. The site selectivity is further evident from the reaction of the homobenzylic methyl ether substrate 4g, which affords only the benzylic C–H cyanation product in good yield and excellent ee. Similarly, benzylic over tertiary C–H activation is observed in the formation of 2z and suggests that the reactive imidyl radical shows exquisite selectivity in the hydrogen-atom-transfer (HAT) reaction. This observation contrasts with more potent HAT reagents, such as photoexcited benzophenone, which exhibit only modest selectivity toward different C–H bonds (35).

Fig. 4 Experimental and computational studies providing insight into the mechanism of the copper-catalyzed cyanation reaction.

These include (A) reaction selectivity studies, (B) kinetic isotope effects, (C) radical trapping experiments, and (D) DFT computational studies.

Stereoselectivity issues were probed in the reaction of enantiomerically pure homobenzylic acetate (R)-4h. Catalysts with each of the enantiomers of ligand L3 were tested, and both led to products with excellent diastereoselectivity, reflecting high levels of catalyst- rather than substrate-controlled stereoselectivity. The differences in diastereomeric ratio and yield of the two products probably reflect a matched-mismatched effect between the ligand and substrate chirality.

A competition deuterium kinetic isotope effect (KIE) study, employing a mixture of 1a and 1a-d2, revealed a KIE of 3.5 (Fig. 4B). On the other hand, independent measurement of the reaction rate of these two substrates revealed only a modest difference in rate (KIE = 1.6 ± 0.2). These values suggest that C–H cleavage is only partially turnover-limiting and that another mechanistic step also contributes to the catalytic turnover rate (e.g., generation of the reactive radical via activation of NFSI by CuI). Additional mechanistic studies are ongoing to secure further insights, but several observations provide clear evidence for formation of a diffusible carbon-centered radical in the reaction (Fig. 4C). Subjection of the cyclopropane-containing substrate 4i to the reaction conditions results in a mixture of the benzylic cyanation product 5i and a ring-opened product (7) in 7% and 32% yield, respectively, together with a mixture of other unidentified by-products. Additional evidence for a radical intermediate was obtained by performing the catalytic reactions of 1-ethylnaphthalene (1a) in the presence of O2 or BrCCl3. Under ambient air as a source of O2, the reaction generated the racemic benzylic alcohol and ketone in 24% and 55% yields, respectively. This product mixture is consistent with trapping of the organic radical by O2 and subsequent conversion of the organic peroxyl radical into the alcohol/ketone mixture. No cyanation product (2a) was observed under the aerobic conditions. When the reaction was performed in the presence of BrCCl3 (2 equivalents), a mixture of the benzylic nitrile and bromide products was obtained (44 and 24% yields, respectively). This product ratio, together with precedent for efficient trapping of organic radicals by BrCCl3 (36), implicates very rapid trapping of organic radicals by a chiral L*CuII(CN) species under the reaction conditions.

The proposed reaction of organic radicals with a L*CuII(CN) species is the subject of ongoing investigation, but the high enantioselectivity observed in the reactions suggests that the benzylic position is in close proximity to the chiral Cu center in the enantioselectivity-determining step. A plausible pathway involves reaction of the intermediate organic radical with CuII, an open-shell, d9 metal center, to afford a benzyl–CuIII species. Preliminary density functional theory (DFT) calculations support the viability of this pathway (Fig. 4D). An ethylbenzene-derived radical reacts at CuII in an (L2)CuII(CN)2 species with a low barrier to afford a benzyl-CuIII species. Subsequent C(sp3)–CN reductive elimination generates the benzylic nitrile product. The calculations suggest that the transition state for the addition of the benzylic radical to CuII is lower in energy than reductive elimination, indicating that formation of the alkyl-CuIII species is reversible and that enantioselectivity is determined by the reductive elimination step. An analogous conclusion was reached in a recent DFT study of Ni-catalyzed cross-coupling reactions, wherein reversible addition of an alkyl radical to a NiII catalyst precedes enantiodetermining reductive elimination (37). For the present system, the difference in the computed activation free energy (ΔΔG) for reductive elimination of the two enantiomeric benzylic nitriles (1.6 kcal/mol) is very similar to the ΔΔG associated with the experimentally observed enantioselectivity (e.g., ΔΔG = 1.5 kcal/mol for 2q) (Fig. 3). Moreover, the reductive elimination transition states for this pathway are calculated to be slightly lower than the calculated barrier for bromine transfer from BrCCl3 to an ethylbenzene-derived radical (ΔG = 12.2 kcal/mol) (Fig. 4D), matching the slight preference for cyanation over bromination noted in Fig. 4C. Although the agreement between the computed pathways and several key experimental observations is quantitatively closer than can be justified by the uncertainties in the computational methods, the qualitative agreement is consistent with the proposed radical coupling pathway with L*CuII(CN)2.

The copper-catalyzed method for enantioselective cyanation of C(sp3)–H bonds described herein represents a valuable demonstration of radical relay catalysis. The ability to use the C(sp3)–H substrate as the limiting reagent, together with the mild reaction conditions, excellent enantioselectivity, and broad substrate scope, provide key foundations for the pursuit of other chemo-, regio-, and stereoselective C–H oxidation reactions.

Supplementary Materials

www.sciencemag.org/content/353/6303/1014/suppl/DC1

Materials and Methods

Figs. S1 to S3

Tables S1 to S3

References (3860)

References and Notes

Acknowledgments: Financial support for this project was provided by the National Basic Research Program of China (973-2015CB856600), the National Natural Science Foundation of China (nos. 21225210, 21421091, 21532009, and 21472217) (G.S.L.), a National Institutes of Health predoctoral fellowship (F31-GM116443) (S.D.M.), and the Department of Energy (DE-FG02-05ER15690) (S.S.S.). The NSF provided partial support for the computational resources (CHE-0840494). G.S.L. thanks X. L. Wan at Shanghai Institute of Organic Chemistry for his assistance with HPLC analysis. Metrical parameters for the structures of (R)- and (S)-2h are available free of charge from the Cambridge Crystallographic Data Centre under accession numbers CCDC-1496364 and -1495535, respectively.

Stay Connected to Science

Navigate This Article