Report

Complex N-Heterocycle Synthesis via Iron-Catalyzed, Direct C–H Bond Amination

See allHide authors and affiliations

Science  03 May 2013:
Vol. 340, Issue 6132, pp. 591-595
DOI: 10.1126/science.1233701

Closing the Cycle

Cyclic hydrocarbons that incorporate nitrogen in the ring are among the most heavily investigated compounds in medicinal chemistry. Hennessy and Betley (p. 591) demonstrate an iron catalyst that forms a range of such cyclic compounds by inducing linear alkyl azides to curl back on themselves, inserting the nitrogen at one end into a carbon-hydrogen bond further down the chain. The reaction furthers a trend of C-H bond activation chemistry that forms elaborate products from relatively simple precursors, without the need to install activating groups at unreactive sites.

Abstract

The manipulation of traditionally unreactive functional groups is of paramount importance in modern chemical synthesis. We have developed an iron-dipyrrinato catalyst that leverages the reactivity of iron-borne metal-ligand multiple bonds to promote the direct amination of aliphatic C–H bonds. Exposure of organic azides to the iron dipyrrinato catalyst furnishes saturated, cyclic amine products (N-heterocycles) bearing complex core-substitution patterns. This study highlights the development of C–H bond functionalization chemistry for the formation of saturated, cyclic amine products and should find broad application in the context of both pharmaceuticals and natural product synthesis.

Saturated, cyclic amines (N-heterocycles) are important building blocks for the synthesis of biologically active natural products, pharmaceutical agents, and materials. Current strategies for constructing saturated N-heterocycles are heavily dependent on functional group exchange, leading to inefficient synthetic protocols with poor atom economy and waste generation. A streamlined synthetic approach to this product class would rely on a catalyst capable of the direct amination of aliphatic C–H bonds. An advantage of this method is its potential to harness saturated hydrocarbon feedstocks. Unfortunately, current C–H bond functionalization protocols often require substrate preoxidation, directing groups, or strong chemical oxidants, which contribute to a lack of generality for this bond construction (14). Herein, we report an iron catalyst capable of functionalizing a broad range of aliphatic C–H bonds to form saturated, cyclic amine products.

A challenge to the development of a general and mild aliphatic C–H bond functionalization strategy is the unreactive nature of the substrates themselves. Saturated hydrocarbons are chemically inert due to the large C–H bond dissociation energy (93 to 105 kcal/mol) coupled with the energetic and spatial inaccessibility of the C–H bonding and antibonding orbitals. Nature provides a blueprint to overcome these obstacles. The reaction of dioxygen with heme iron in cytochrome P450 produces a strong oxidant consisting of an iron-oxygen multiple bond (iron-oxo) (5). The iron-oxo bond contains two electrons residing in Fe–O π* orbitals [Fe(dxz,dyz) – O(px,py)], which result in a weakened Fe–O bond vector possessing radical character, thus rendering the entire unit a reactive functionality. As a consequence of this electronic configuration, the iron-oxo bond can activate substrate aliphatic C–H bonds via an H–atom abstraction mechanism and thereby circumvent the orbital spatial restrictions that hinder oxidative addition pathways. Subsequent substrate functionalization results from recombination of the organic radical generated in the activation step with the open-shell iron-hydroxyl to produce an alcohol product with concomitant reduction of the iron. Despite this 30-year-old mechanistic precedent (6), viable catalysts fashioned with these design principles are only now being discovered.

The direct functionalization of C–H bonds based on a strategy exemplified by cytochrome P450 would be transformative in converting ubiquitous C–H bonds into functional group handles and would circumvent the traditional synthetic requirement for functional group exchange (7). The electronic structure of the cytochrome P450 reactive iron-oxo intermediate can, in principle, be replicated with any metal-ligand multiple bond (8) and would constitute a general strategy for the conversion of unactivated C–H bonds into a variety of C–heteroatom bond products. Indeed, metal stabilized carbene and nitrene transfer has garnered considerable interest through the use of noble metal catalysts (1, 914). Specifically, Fiori et al. (12) and Liang et al. (13) have developed a class of C–H amination Rh2-dicarboxylate catalysts capable of generating cyclic carbamate, guanidine, and sulfamide products. Recently, this methodology has been extended to include aryl azides to produce indolines via an intramolecular sp3 C–H amination, as reported by Nguyen et al. (14). In contrast, late, first-row transition metal complexes are potentially ideal catalyst candidates but have been less explored. Their high d-electron count and compressed ligand fields (compared with their second- and third-row analogs) favor population of metal-ligand antibonding orbitals leading to destabilization and reactivity akin to the cytochrome P450 iron-oxo intermediate (1524). With these design principles in mind, we describe an iron-dipyrrinato catalyst that can selectively aminate sp3 C–H bonds. Herein, we present the application of this catalyst toward the production of complex cyclic amine structures from simple linear aliphatic substrates requiring a single functionality, an azide (Fig. 1).

Fig. 1 Azide cyclization and iron-bound pyrrolidine products.

(A) Reaction of complexes 1 and 2 with linear azides to generate Fe-bound pyrrolidine products 3 to 7. (B) Solid-state core structures of (AdL)FeCl(2-Ph-NHC4H7) (3) from reaction with 1 and 1-azido-4-phenylbutane, (AdL)FeCl(2-Et-NHC4H7) (6) from reaction with 1 and 1-azidohexane, and (AdL)FeCl(2,2-Me2-NHC4H6) (7) from reaction with 2 and 2-azido-2-methylpentane, with the thermal ellipsoids set at the 50% probability level (Fe, orange; C, gray; H, white; N, blue; Cl, green).

We previously reported intermolecular amination of benzylic C−H bonds with aryl and alkyl azides using the ferrous dipyrrinato complex (RL)FeCl(solv) [R = 2,4,6-Ph3C6H2, Ad with meso-Ar = mesityl (1), 2,6-Cl2C6H3 (2); solv = Et2O, tetrahydrofuran; L, ligand; Ph, phenyl; Ad, adamantyl; Ar, aryl; Et, ethyl] (19). Isolation and characterization of the reactive intermediate elucidated the electronic structure of the high-spin, iron-bound imido radical, wherein a high-spin Fe(III) (S = 5/2) is antiferromagnetically coupled to the imido radical (S = –1/2) to give a high-spin ground state. This electronic structure places substantial radical character on both the Fe−N σ and π bond vectors, facilitating both radical H-atom abstraction and radical recombination pathways to proceed. Furthermore, the amination catalytic cycle remains in the quintet spin state (S = 2), making each step of the catalytic cycle spin-allowed. We envisioned that the intramolecular extension of this reactivity profile to linear, aliphatic azide substrates would generate cyclic amine products in a single operation.

To test the viability of catalyst 1 for intramolecular C–H amination, we subjected a variety of substituted aliphatic azides to complex 1 (Fig. 1A). Exposure of 1-azido-4-phenylbutane to 1 at room temperature in benzene resulted in consumption of the azide, as ascertained by the disappearance of the azide stretch in the infrared spectrum, and afforded a new paramagnetically shifted 1H nuclear magnetic resonance (NMR) spectrum. Crystallization occurred from a concentrated hexanes solution of the product at 23°C to yield crystals in which 2-phenylpyrrolidine was bound to the (AdL)FeCl complex (3, Fig. 1B) (25). Similarly, treatment of 1-azido-5-hexene with 1 afforded the cyclized product 2-vinylpyrrolidine as an iron-bound adduct (4, fig. S5). In addition to allylic and benzylic C–H bonds, less reactive tertiary C–H bonds could be similarly functionalized. The reaction of 1-azido-5-methylpentane and 1 under standard conditions gave the 2,2-dimethylpyrrolidine iron-bound product (5). Gratifyingly, even secondary aliphatic C–H bonds could be functionalized through this method. Addition of 1-azidohexane to 1 resulted in the rapid consumption of the azide to afford the 2-ethylpyrrolidine complex (6, Fig. 1B). In an attempt to activate the primary C–H bond of an aliphatic azide substrate, 1-azidobutane was exposed to 1. However, the only products observed in this transformation were linear n-butylamine and n-butylimine. To eliminate the potential for imine formation, through a process involving intermolecular C–H bond activation or β-hydride elimination, the gem-dimethyl substrate 2-azido-2-methylpentane was prepared and subjected to 1 at room temperature for 6 hours to afford the cyclized 2,2-dimethylpyrrolidine complex (7, Fig. 1B) in quantitative yield. The presence of the two α-Me (Me, methyl) substituents may facilitate the C–H bond functionalization and cyclization process through the Thorpe-Ingold effect (26).

With a reliable protocol for the stoichiometric C–H functionalization of aliphatic azides in hand, we attempted to render the reaction catalytic [5 to 10 equivalents (equiv.) of azide per 1 equiv. of 1 or 2]. Unfortunately, examination of the cyclization reaction under catalytic conditions did not markedly increase the yield of the resultant free heterocyclic product. We attribute the lack of catalyst turnover to product inhibition, in which a tight Lewis acid/base pair between the dipyrrinato iron and the heterocyclic nitrogen atom is formed (Fig. 1B). To overcome product inhibition, we performed the cyclization reaction in the presence of an in situ protection reagent, which would reduce the nucleophilicity of the product N-heterocycle while avoiding the generation of by-products that might retard or prevent catalysis. Accordingly, treatment of a solution of 1-azido-4-phenylbutane and 9-fluorenylmethyl N-succinimidyl carbonate (Fmoc-OSuc; a base-cleavable protecting group) in benzene at room temperature with a stoichiometric amount of 2 for 12 hours afforded the Fmoc-protected 2-phenylpyrrolidine in 98% yield (Table 1, entry 1). Similarly, addition of an equivalent of 2 to a solution of 1-azido-4-phenylbutane and di-tert-butyl dicarbonate (Boc2O; an acid-cleavable protecting group) under similar reaction conditions afforded the tert-butyloxycarbonyl (Boc)–protected product 1-Boc-2-phenylpyrrolidine in 93% yield. As catalyst loading was decreased, the N-hydroxysuccinimide by-product of Fmoc-protection led to catalyst decomposition through ligand protonation and limited the reaction to a single turnover. Fortunately, the by-products of protection with Boc2O (tBuOH, CO2; tBu, tert-butyl) did not inhibit catalyst turnover, permitting the heterocycle to be synthesized with catalytic amounts of 2 (chosen to eliminate benzylic C–H bonds from catalyst meso-aryl substituent).

Table 1 Catalytic synthesis of pyrrolidine products.

R1: Ph, CHCH2, Me, Et, H, CO2Et, (CH2)5; R2: H, Me, (CH2)5; R3: H, Et, Me; R4: H, OTMS (trimethylsiloxy), Ph, Me; R5: H, Me, Ph; R6: H, Me. dr, diastereomeric ratio.

View this table:

We investigated application of catalytic quantities of 2 to the established in situ protection protocol for C–H functionalization and cyclization (Table 1). Exposure of substrates containing allylic, benzylic, or tertiary C–H bonds to 2 (10 mol %) provided the corresponding Boc-protected pyrrolidine in good isolated yield (49 to 72%, entries 1 to 3). Catalytic functionalization of a secondary C–H bond was also possible, affording 1-Boc-2-ethylpyrrolidine in a lower isolated yield (19%, entry 4). The formation of undesired linear side products is competitive with cyclization at the strong secondary C–H bond and contributes to the low yield. Even a primary C–H bond in 2-azido-2-methylpentane could be functionalized to give 1-Boc-2,2-dimethylpyrrolidine in 17% isolated yield (entry 5), with a mass balance of unreacted azide. Unlike the previous example, we hypothesize that the diminished yield reflects the ability of the tertiary azide to access the iron catalyst. In situ 1H NMR monitoring of the stoichiometric reaction between 1 and 2-azido-2-methylpentane to generate the 2,2-dimethylpyrrolidine iron adduct requires 55 min at room temperature (with no detectable buildup of an intermediate), whereas conversion of 1-azido-5-methylpentane to the same product is complete in 5 min (figs. S1 and S2). The long reaction time to aminate the primary C–H bond likely has an adverse effect on the overall catalysis, as competitive catalyst inactivation occurs on a similar time scale. We then expanded the substrate scope to include hetero-atom–containing functional groups. Exposure of ethyl-5-azidopentanoate to 2 under standard catalytic conditions resulted in only linear primary amine and imine products. Again, blocking the α position of the azide with gem-dimethyl substituents led to productive cyclization (entry 6), albeit in the low yield characteristic of tertiary azide substrates. Introduction of heteroatoms between the reactive functionalities allowed for the formation of 1-Boc-2-phenyloxazolidine in 47% yield (entry 7). The reaction is also tolerant of siloxy groups, as shown in the formation of 1-Boc-2-vinyl-4-trimethylsiloxypyrrolidine in 68% isolated yield (entry 8).

Next, we explored the synthesis of highly substituted pyrrolidine products using substrates accessible via cuprate-assisted epoxide opening (27) followed by azide formation (Table 1, entries 9 to 18). The use of commercially available epoxide and alkyl halide building blocks permits virtually any substitution pattern to be programmed into the ensuing heterocyclic product. After Li2[CuCl4] promoted epoxide opening, the resultant primary or secondary alcohols were tosylated and displaced with sodium azide to provide the desired cyclization precursor. Alternatively, tertiary and benzylic alcohols were directly converted to the corresponding azide by exposure to trimethylsilylazide and boron trifluoride diethyl etherate (28). Allylic, tertiary, and secondary C–H bond substrates available through either reaction sequence underwent facile C–H functionalization and cyclization (entries 9 to 17, 58 to 98% yield). Notably, this method provides access to an all-carbon spiro-center (entry 15, 67%). Functionalization of primary C–H bonds required stoichiometric catalyst loading (entry 18, 78% yield), for reasons stated above. Use of (R)-2-phenyl-5-azidopentane [95% enantiomeric excess (ee)] in the catalytic transformation resulted in (S)-2-methyl-2-phenylpyrrolidine with retention of configuration (entry 13, 75%, 93% ee). Application of stoichiometric quantities of catalyst 1 to the reaction gave the corresponding (S)-2-methyl-2-phenylpyrrolidine iron-bound adduct whose absolute stereochemistry was verified by x-ray diffraction (fig. S8).

Last, we investigated the potential of this method to generate N-heterocycles of various ring sizes. We anticipated that a vinyl directing group could be employed to encourage the site-selective functionalization of the allylic C–H bond within the acyclic precursor. Gratifyingly, treatment of 1-azido-6-heptene and Boc2O (1 equiv.) with 2 (1 equiv.) at room temperature generated the six-membered 1-Boc-2-vinylpiperidine (Table 2, entry 1) as the exclusive reaction product. Use of the vinyl activating group to target seven-membered azepane products, however, led to exclusive formation of the corresponding pyrrolidine (entry 2). In contrast, a phenyl activating group was not effective in favoring the formation of a six-membered-ring product. Addition of 2 to 1-azido-5-phenyl-pentane under standard conditions resulted in a 1:0.85 mixture of both 1-Boc-2-phenylpiperidine and 1-Boc-2-benzylpyrrolidine (entry 3). Similarly, the use of a tertiary C–H bond to favor six-membered-ring formation in the case of 1-azido-5-methylhexane resulted in a mixture of piperidine and pyrrolidine products (entry 4). An attempt was made to block the potential for pyrrolidine formation; exposure of 2-azido-2,5,5-trimethylhexane and Boc2O to 2 resulted in both the anticipated 1-Boc-2,2,5,5-tetramethylpiperidine and the unexpected 1-Boc-2,2-dimethyl-4-tert-butylazetidine (entry 5). Alternatively, use of catalyst 1 and omission of Boc2O allowed for the characterization of the corresponding iron-bound piperidine and azetidine adducts by x-ray analysis (figs. S10 and S11).

Table 2 Product distribution for azetidine, pyrrolidine, and piperidine formation.

R: H, Me; R1: CHCH2, (CH2)2CHCH2, Ph, Me, CMe3; R2: H, Me; GC/MS, gas chromatography–mass spectrometry.

View this table:

As illustrated in Fig. 2A, we hypothesize that the C–H bond functionalization reaction occurs via a three-step process involving: (i) oxidation of the FeII catalyst to an FeIII imido radical by the alkyl azide substrate, (ii) intramolecular H-atom abstraction to generate an alkyl radical and an FeIII amide (path Ia), and (iii) radical recombination to form the observed N-heterocyclic product (path Ib). Alternatively, a direct C–H bond insertion by the FeIII imido radical intermediate cannot be excluded (path II). Both mechanisms require that the substrate C–H bond be brought into close proximity to the reactive Fe-imido radical. Based on our previous findings, the imido radical likely resides in the plane defined by the iron and the dipyrrin ligand, flanked by large pyrrolide adamantyl substituents. Such a conformation requires that the C–H bond substrate approach the imido radical opposite the chloride ligand. We expect that once this orientation is obtained, C–H bond functionalization is rapid. This hypothesis is supported by the retention of stereochemical information during the cyclization of (R)-2-phenyl-5-azidopentane (Fig. 2B). This stereoretention probably reflects the spatial constraints imposed by the ligand adamantyl units to inhibit racemization of the carboradical intermediate. Additionally, cyclization of 1-azido-4-deutero-4-phenylbutane provides an intramolecular kinetic isotope effect of 5.3 at 25°C and 5.1(2) at 65°C (Fig. 2C). This value is similar to the kinetic isotope effect (KIE) observed in the hydroxylation of 1,3-dideuteroadamantane catalyzed by tetramesitylporphyrin iron with oxone [KIE = 4.1(2)] (29). Finally, addition of the radical clock substrate (2-(4-azidobutyl)cyclopropyl)benzene to 2 exclusively furnishes the pyrrolidine product 1-Boc-2-(2-phenylcyclopropyl)pyrrolidine with the cyclopropyl unit intact (Fig. 2D). The nonunity intramolecular kinetic isotope effect suggests a stepwise mechanism for benzylic substrates (Fig. 2A, path I), which is consistent with our previously reported intermolecular amination reaction (19). The stereospecificity of the cyclization and the preservation of the cyclopropyl unit in the radical clock experiment suggest that if a stepwise mechanism is operative, the radical intermediate following H-atom abstraction is short-lived [recombination rate > 1011 s–1 (30)]. Alternatively, the reaction mechanism may change to a direct-insertion pathway (Fig. 2A, path II) when stronger substrate C–H bonds are functionalized.

Fig. 2 Mechanistic studies.

(A) Proposed mechanistic pathways for intramolecular C–H amination of linear alkyl azides with Fe catalyst 1 or 2 to form N-heterocycles. Pyrrolidine formation is depicted, although azetidine and piperidine products are also accessible. (B to D) Substrates designed to probe the mechanism of C–H functionalization and distinguish between paths I and II.

The foregoing results have demonstrated the oxidative potency of the transiently formed, high-spin iron imido radical for the functionalization of both activated and unactivated aliphatic C–H bond substrates. This iron-mediated cyclization of linear azides provides facile entry into complex N-heterocyclic products from readily available substrates that cannot be achieved by azide photolysis (31) or via classic Hoffmann-Löffler-Freytag methodologies (32). We anticipate the methodology described herein can be extended to produce a wide variety of saturated, cyclic structures.

Supplementary Materials

www.sciencemag.org/cgi/content/full/340/6132/591/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S11

Table S1

References (3352)

References and Notes

  1. Materials, methods, and x-ray diffraction details are available as supplementary materials on Science Online.
  2. Acknolwedgments: This work was supported by a grant from the NSF (CHE-0955885) and Harvard University. E.T.H. is grateful for a predoctoral fellowship from the U.S. Department of Energy Office of Science Graduate Fellowship Program (DE-AC05-060R23100) administered by Oak Ridge Institute for Science Education–Oak Ridge Associated Universities. T.A.B. is grateful for a George W. Merck Fellowship. We thank S.-L. Zheng for assistance with crystallography; J. Hong, A. Saghatelian, and B. Milgram for helpful discussions; and E. N. Jacobsen and T. Ritter for the generous use of their high-performance liquid chromatography and GC/MS instruments, respectively. The crystallographic data CCDC-905454 – 905461 can be obtained free of charge from the Cambridge Crystallographic Data Centre (www.ccdc.cam.ac.uk/data_request/cif).
View Abstract

Navigate This Article