Direct Stereospecific Synthesis of Unprotected N-H/N-Me Aziridines from Olefins

Abstract

Despite the prevalence of the N-H aziridine motif in bioactive natural products and the clear advantages of this unprotected parent structure over N-protected derivatives as a synthetic building block, no practical methods have emerged for direct synthesis of this compound class from unfunctionalized olefins. Here, we present a mild, versatile method for the direct stereospecific conversion of structurally diverse mono-, di-, tri- and tetra-substituted olefins to N-H aziridines using O-(2,4-dinitrophenyl)hydroxylamine (DPH) via homogeneous rhodium catalysis with no external oxidants. This method is operationally simple (i.e., one-pot), scalable and fast at ambient temperature, furnishing N-H aziridines in good-to-excellent yields. Likewise, N-alkyl aziridines are prepared from N-alkylated DPH derivatives. Quantum-mechanical calculations suggest a plausible Rh-nitrene pathway.

Aziridines, the triangular, comparably highly-strained nitrogen analogues of epoxides, are important synthetic intermediates (i.e., building blocks) en route to structurally complex molecules due to their versatility in myriad regio- and stereoselective transformations (ring openings and expansions as well as rearrangements).(1–6) The aziridine structural motif, predominantly N-H and to a lesser extent N-alkyl, also appears in biologically active natural products (e.g., azinomycins and mitomycins).(7–9) As a result, the synthesis and chemistry of aziridines have been the subject of intense research during the past 25 years, resulting in multiple aziridination methods.(10–23) The majority of these methods rely either on the transfer of substituted nitrenes, which are generated using strong external oxidants, to the C=C bond of olefins or the transfer of substituted carbenes to the C=N bond of imines. Normally, the result is an aziridine bearing a strongly electron-withdrawing N-protecting group (e.g., Ts = para-toluenesulfonyl, Ns = para-nitrophenylsulfonyl); removal of these N-sulfonyl protecting groups is problematic as it often results in the undesired opening of the aziridine ring. In addition, the high reactivity of N-protected nitrenes might give rise to non-productive allylic C-H amination products as well as the loss of stereospecificity. Clearly, the direct synthesis of N-H (i.e., N-unprotected) and N-alkyl aziridines would alleviate the above problems. However, a practical, functional group-tolerant and environmentally benign direct preparation of N-H aziridines from structurally diverse olefins has so far eluded synthetic chemists.(24–31) Herein, we report an operationally simple, inherently safe, chemoselective and stereospecific conversion of a wide range of olefins to the corresponding N-H/N-Me aziridines via a rhodium-catalyzed pathway free of external oxidants.

Recently, we developed a metal-free protocol for primary amination of arylboronic acids using only O-(2,4-dinitrophenyl)hydroxylamine (DPH, 1a, Fig. 1) as the stoichiometric aminating agent.(32) The transformation proceeds under neutral or basic conditions and can be conducted on a multi-gram scale to provide structurally diverse primary arylamines. The versatility and robustness of 1a prompted us to explore other uses of this aminating agent, specifically for the direct functionalization of readily available and inexpensive olefins. Our investigations began by subjecting 1:1.5 mixtures of cis-methyl oleate (7)/1a as well as styrenes (3a & 3b)/1a to a vigorous screening with a variety of transition metal complexes (see Tables S1 & S2, Supplementary Materials). This initial screen identified Rh₂(OAc)₄ as a promising catalyst for vic-amino-oxyarylation of olefins. Further evaluation of dimeric rhodium dicarboxylate complexes (see Table S3, Supplementary Materials), revealed that just 1 mol% loading of Du Bois’ catalyst(33–36) (2, Fig. 1) in acetonitrile (MeCN) leads to amino-oxyarylated styrenes 4a and 4b at room temperature in 56% and 75% isolated yields, respectively. These promising results prompted us to conduct a thorough solvent screen.

Fig. 1.

Exploration of DPH (1a) as a versatile aminating agent. (A) Rh₂(esp)₂ is an effective catalyst for olefin difunctionalization. (B) In 2,2,2-trifluoroethanol (TFE or CF₃CH₂OH), 7 undergoes direct aziridination to N-H aziridine 8 in excellent isolated yield.

In methanol, we observed the incorporation of the MeO group at the benzylic position (5) in addition to the amino-oxyarylated product 4b; these compounds were isolated in a combined yield of 78%. At this juncture, we reasoned a highly polar, hydroxylic and non-nucleophilic solvent such as 2,2,2-trifluoroethanol (CF₃CH₂OH, TFE) would completely avoid the incorporation of solvent into the products. Indeed, 3b was cleanly amino-oxyarylated in TFE and 4b was isolated in 66% yield. It was unclear if the transformation 3b→4b involved the opening of a highly reactive aziridine (6) or an alternative process. Surprisingly, when 7 was reacted in trifluoroethanol as solvent, cis-N-H aziridine 8 was isolated in excellent yield (83%) instead of the expected amino-oxyarylated product. The transformation proceeded with complete stereospecificity as no traces of the trans-N-H aziridine were detected by ¹H- and ¹³C-NMR analysis (⁤2% sensitivity).

Encouraged by this unexpected, yet most welcome result, a systematic study was initiated using representative aliphatic olefins with a wide range of substitution patterns and functionalities (Fig. 2). Terminal aliphatic olefin substrates (entries 1–3, Fig. 2) either did not react or reacted sluggishly (i.e., days) when 1 mol% of catalyst 2 was used; however, increasing the catalyst loading to 5 mol% led to rapid conversion at room temperature to the corresponding N-H aziridines (10a–c). We empirically found that in some of the reactions (i.e., entries 4, 5, 7, 9, 11, 14 & 20) addition of the catalyst in several 1 mol% portions minimized decomposition of both the catalyst and aminating agent and invariably led to higher isolated yield of product. Remarkably, the N-H aziridination took place efficiently in the presence of a labile terminal epoxide (10c) as well as an unprotected primary alcohol (10a); these functionalities typically interfere with currently used aziridination protocols. In case of the transformation 9c→10c, only the product was detected in the crude reaction mixture by NMR analysis. In the presence of 1 mol% of catalyst 2, both cis- and trans-1,2-disubstituted aliphatic olefins (entries 4–10, Fig. 2) underwent smooth and stereospecific N-H aziridination at room temperature as established by ¹³C-NMR analysis (≤2% sensitivity). The presence of an unprotected secondary alcohol in substrate 9i (entry 9) did not influence the stereochemical outcome of the N-H aziridination and 10i was isolated as a 1:1 mixture of diastereomers.

Fig. 2.

Direct and stereospecific N-H aziridination of olefins. Reactions were conducted at 0.1M using 2,2,2-trifluoroethanol as solvent and at 0.5 mmol scale unless otherwise indicated. To obtain crystalline material, 10s was O-acetylated and N-tosylated (Ts = para-toluenesulfonyl) to afford derivative 10ss.

Benzoyloxy and acetyloxy cis-olefins 9k and 9m (entries 11 & 14), when exposed to 1 mol% of catalyst 2 and 1.2 equiv of aminating agent 1a at 50 °C, were smoothly aziridinated followed by an in situ aziridine ring-opening (via transacylation) to yield the corresponding trans-2,3-disubstituted furans 10kk and 10mm in 84% and 61% yields, respectively. On the other hand, when olefin 9k was exposed to 5 mol% loading of catalyst 2 and 1.2 equivalents of 1a at 25 °C, the expected N-H aziridine 10k (entry 12) was formed in just 2 hours and isolated in 69% yield. As anticipated, when the rate of N-H aziridination is slow and elevated temperatures are used, secondary processes (i.e., intramolecular annulation) that consume the initially formed N-H aziridines can dominate. Apparently, a five-fold increase in catalyst loading increased the rate of N-H aziridination sufficiently that it could take place rapidly at ambient temperature.

Cyclohexene 9n (entry 15) was aziridinated at room temperature to afford cyclic N-H aziridine 10n; no traces of allylic C-H amination (i.e., 1-amino-2-cyclohexene) could be detected by ¹H-NMR analysis (⁤2% sensitivity), in sharp contrast with other metal nitrene-based aziridination methods.(37) Geraniol (9o, entry 16) and geranyl acetate (9q, entry 18), which incorporate two trisubstituted C=C double bonds, were N-H aziridinated regioselectively, favoring the double bond at the Δ^6,7-position over the Δ^2,3-position in both cases.

The shift of the regioisomeric ratio from 1:5 in 10o to 1:14 in 10q suggests a subtle directing effect of the free allylic alcohol and/or an inductive deactivation by the acetate; perhaps the extent of H-bonding in the solvent also plays a role. Entry 17 stands as a testament to the extraordinarily mild reaction conditions as trisubstituted olefin 9p, which possesses a highly sensitive epoxy alcohol, was aziridinated rapidly and efficiently to epoxy N-H aziridine 10p in excellent yield. The transformation 9q→10q (entry 18) could be readily scaled up (6 mmol) with minimal erosion of the isolated yield to provide gram quantities of 10q. N-H aziridination of limonene 9r (entry 19) favored the trisubstituted ring double bond with 9:1 regioselectivity; however, the chiral center had no evident influence on the diastereoselectivity (1:1 dr). In contrast with the lack of stereoselectivity in 9i, cholesterol 9s (entry 20) exclusively yielded the β-N-H aziridine 10s in 71% yield; this unexpected stereochemical outcome, confirmed by single crystal X-ray analysis of 10ss (a crystalline derivative of 10s), suggests a directing effect by the adjacent C(3)-β-alcohol not observed in conformationally more mobile acyclic molecules such as 9i. The success with cholesterol and other natural products (7, 9h, 9i, 9o and 9r, Fig. 1 & 2) highlights the prospective utility of this method in the straightforward elaboration of molecules of biomedical interest (e.g., for ¹⁵N-labeling studies).

Next, we turned our attention to the direct N-H aziridination of di-, tri- and tetra-substituted styrenes and stilbene (entries 21–28, Fig. 3A). In general, styrenes were more reactive than aliphatic olefins, and often lower temperatures (−10 to 25°C) were adequate. Conspicuously, cis- β-methyl styrene 11d furnished the corresponding cis-2-Ph-3-Me N-H aziridine (12d, entry 24) without isomerization. Similarly, trans-β-methyl styrene 11c readily furnished trans-2-Ph-3-Me N-H aziridine (12c, entry 23) even on a 1 to 8 mmol scale. The N-H aziridine derived from 2-Me indene (12h, entry 28) was not isolated due to its high reactivity, but instead reduced in situ to amine 12hh. Evaluation of the effect of catalyst loading on the reaction 11f→12f (entry 26) revealed the lowest practical loading of catalyst 2, without decreasing the isolated yield or drastically increasing the reaction time, was 0.5 mol%. This low catalyst loading renders the process economical and environmentally friendly. A further five-fold reduction in catalyst loading (from 0.5 mol% to 0.1 mol%) resulted in a 25-fold increase in reaction time and a 30% drop in the isolated yield of 12f. To our delight, tetrasubstituted olefin 11g (entry 27) was easily N-H aziridinated at room temperature; 12g was isolated in 70% yield. The attempted direct N-H aziridination of 1-Ph-1-cyclopropylethene (11b) yielded only amino-oxyarylated product 12b; the complete lack of cyclopropane ring-opening products corroborate an aziridination pathway that does not involve long-lived radical or carbocation intermediates (see more detailed discussion of the mechanism in the computation section and also in Fig. 4).

Fig. 3.

Direct and stereospecific N-H and N-Me aziridination of olefins. Reactions were conducted at 0.1M using 2,2,2-trifluoroethanol as solvent and at 0.5 mmol scale unless otherwise indicated. CSA = camphorsulfonic acid.

Fig. 4.

Selected DFT-examined pathways for N-H aziridination of styrene in 2,2,2-trifluoroethanol solvent. R = esp ligands. Energies in kcal/mol. MECP = minimum energy crossing point.

The practicality and broad scope of the preceding direct and stereospecific N-H aziridination of olefins (Fig. 2 & Fig. 3A) prompted an investigation of direct N-Me aziridination. Towards this end, several di- and trisubstituted aliphatic olefin and styrene substrates (entries 29–33, Fig. 3B) were examined in the presence of 1b as the stoichiometric aminating agent and 1 to 2 mol% of catalyst 2. The N-Me aziridination of olefins also proceeded stereospecifically (entries 29 & 30) and, in the case of geraniol acetate 9q, the regioselectivity increased from 1:14 (in 10q) to >1:30 (in 13c), favoring the Δ^6,7-olefin in both cases.

Two of the N-H aziridine products (12c and 12f) were subjected to ring-opening transformations (Fig. 3C). Upon catalytic hydrogenation, aziridine 12c afforded a 94% yield of amphetamine 15, the active pharmaceutical ingredient (API) in Adderall™, an approved medication for attention deficit hyperactivity disorder (ADHD) as well as narcolepsy that is marketed as a mixture of enantiomers. Under acidic conditions, at slightly elevated temperature (40 °C) in MeOH, 12c was converted to O-Me-norephedrine 14 with complete regioselectivity and in nearly quantitative yield. Likewise, the ring-opening of trisubstituted N-H aziridine 12f with sodium azide furnished azidoamine 16 in 79% yield. These transformations by example illustrate how readily a nitrogen atom can be introduced into molecules.

We also examined prospective reaction mechanisms using quantum mechanical density-functional theory calculations (Fig. 4). Our (U)M06 calculations were carried out in Gaussian 09 (38) using a polarizable conductor continuum solvent model for trifluoroethanol. Details of calculated transition states and intermediates are given in the Supplementary Materials.

We first examined plausible rhodium nitrene pathways. Generation of a rhodium nitrene intermediate is possible if the amino group of 1a coordinates to Rh₂(esp)₂ followed by loss of dinitrophenol (Pathway A, Fig. 4). Calculations suggest that the triplet-spin state of the nitrene (³17) is more than 8 kcal/mol lower in energy than the open-shell singlet and reaction pathways identified on the triplet-spin energy surface were found to be lower in energy than reaction pathways on the singlet-spin energy surface.(39, 40) Because the Rh₂(esp)₂ catalyst and aziridine product have singlet-spin ground states, the reaction pathway must involve spin interconversion. The mechanism outlined in Fig. 4 provides a route for stereospecific aziridination if ³17 reacts with alkenes by forming the first C–N bond via triplet transition state TS1 followed by spin interconversion along the pathway to diradical intermediate 19 or fast spin interconversion at the diradical intermediate.(41) After spin interconversion, the second C–N bond is formed by the coupling of singlet-paired electrons without a barrier and leads directly to aziridine 20.

As alternatives to nitrene pathways we also explored polar mechanisms involving Rh-amine and Rh-alkene coordination modes (see Supplementary Materials). One of several possible polar mechanisms is outlined as Pathway B in Fig. 4. This pathway is akin to the mechanism proposed for amination of aryl boronic acids with 1a.(32) While this mechanism may account for amino-oxyarylated products (e.g., 4a & 4b) observed under some experimental conditions, the calculated barrier for this mechanism, as well as alternative polar mechanisms, are higher in energy than the nitrene mechanism presented in Pathway A.