Catalytic Enantioselective [3 + 2] Cycloaddition of α–Keto Ester Enolates and Nitrile Oxides

Abstract

An enantioselective [3 + 2] cycloaddition reaction between nitrile oxides and transiently generated enolates of α-keto esters has been developed. The catalyst system was found to be compatible with in situ nitrile oxidegeneration conditions. A versatile array of nitrile oxides and α-keto esters could participate in the cycloaddition, providing novel 5-hydroxy-2-isoxazolines in high chemical yield with high levels of diastereo- and enantioselectivity. Notably, the optimal reaction conditions circumvented concurrent reactions via Oimidoylation and hetero-[3 + 2] pathways.

Graphical Abstract

INTRODUCTION

Heterocycles are routinely encountered as structural units of biologically active unnatural¹ and natural² organic compounds. The contemporary chemist has at his or her disposal myriad modern methods to synthesize heterocyclic structures, yet the time-honored cycloaddition reaction between an olefin and reactive 1,3-dipoles is central because of its chemical efficiency and the breadth of structures accessible.^3,4 The development of new catalytic asymmetric 1,3-dipolar cycloaddition reactions holds significant potential for the preparation of enantioenriched five-membered heterocyclic scaffolds.⁵ In particular, new activation modes allow the union of 1,3-dipoles with unprecedented reaction partners. Herein we describe a catalytic asymmetric [3 + 2] cycloaddition between in situ-generated nitrile oxides and catalytically generated transition metal enolates of α-keto esters.

Our groups have reported that late transition-metal complexes act as catalysts for reactions that proceed via the soft enolization of carbonyl compounds.⁶ Transformations developed under this paradigm include Michael,^6a halogenation, ^6b,c Mannich,^6d and aldol^6e reactions of β-keto esters as well as Michael additions^6h,f and halogenations^6g–i using α-keto esters as pronucleophiles.⁷ The integration of unconventional electrophiles under this mechanistic construct carries with it the possibility to create heretofore unknown compounds: 1,3- dipolar synthons emerged as attractive targets. Sustmann classified 1,3-dipolar cycloadditions based on the relevant FMO interactions:⁸ type III cycloadditions are controlled by the HOMO of the dipolarophile and the LUMO of the dipole (Scheme 1, top). Drawing a parallel with electron-rich dipolarophiles,^5a–d we postulated that transition-metal enolates could engage dipoles in [3 + 2] cycloadditions. Precedent for achieving catalytic asymmetric dipolar cycloadditions by way of HOMO-raising activation exists. Yanagisawa and co-workers reported the cycloaddition of nitrones and alkenyl acetates and proposed the intermediacy of catalytically formed tin(IV) enolates.⁹ Additionally, enamine catalysis has been used to facilitate asymmetric [3 + 2] cycloadditions of aldehydes and azomethine imines.^10,11 Building on this concept, we recently developed a catalytic asymmetric [3 + 2] cycloaddition of α-keto ester enolates and nitrones¹² inspired by the structural characterization of a Ni(II)–diamine catalyst that merges Ni(II)–enolate and hydrogen-bonding activation modes.¹³ However, the scope of the process was limited to isolable (E)-nitrones, specifically dihydroquinoline derivatives. The generation of heretofore unknown isoxazolines via the reaction of an ephemeral metalloenolate with a transient nitrile oxide emerged as an attractive yet challenging next step (Scheme 1, bottom).¹⁴

Scheme 1.

Dipolar Cycloaddition by Dual Reagent Activation

RESULTS AND DISCUSSION

At the outset, we perceived a number of unique challenges to our reaction plan. Foremost among these is the inherent instability of nitrile oxides toward dimerization (Scheme 1, path A)^15,3b and the need to generate them in situ via base-mediated dehydrohalogenation of the corresponding hydroximoyl chloride or by the dehydration of nitroalkanes.¹⁶ Additionally, our laboratories have observed that α-keto esters 1 react via a homoaldol reaction (path B) in the absence of other viable pathways. The known hetero [3 + 2]-cycloaddition of nitrile oxides with α-keto esters defines a third relative-rate concern (path C),¹⁷ especially since the latter will be present in great excess relative to the derived metalloenolate. Finally, the inherent competition of O- versus C-trapping of metalloenolates (path D vs E) must be effectively managed for a successful outcome.

We hypothesized that the Ni(II) catalyst developed for the cycloaddition of nitrones with α-keto ester 1¹² would display a similar efficiency with nitrile oxides and that running the reaction at −40 °C would preclude the undesired dioxazole coproduct 4 (Table 1). Catalytic amounts of tertiary amine bases have been shown to accelerate reactions involving Ni(II)–enolates of α-keto esters,^6h thus nitrile oxides were generated in situ from hydroximoyl chlorides using triethylamine. Hydroximoyl chloride 2a was chosen as a model substrate. As shown in entries 1 and 2, when utilizing Ni(OAc)₂ complexes of ligands A and B in the reactions, the conditions did not favor the formation of the desired 5-hydroxy-2- isoxazoline 3a. In addition to the anticipated dioxazole coproduct 4a, the intervention of the O-imidoylation pathway was concurrently observed (5a).¹⁸ However, we were encouraged by the promising levels of stereoselectivity observed for 3a (Table 1, entry 1: −82% ee, 11:1 dr and entry 2: −53% ee, 11:1 dr) without interference by the forecasted homodimerization (path A) and homoaldol (path B) pathways. A solvent screen (entries 3–5) revealed that the chemoselectivity could be dramatically improved (Table 1, entry 5: 10:1 3a/5a) if N,N-dimethylformamide was employed as the solvent. In contrast, the use of ⁱPrOH as the solvent induced a preference for 5a (Table 1, entry 6: 1:2 3a/5a). Further screening of conditions with Ni(II) failed to yield satisfactory results (See Supporting Information for details).

Table 1.

Reaction Optimization

entry	ligand	M(OAc)2	solvent	yield of 3a [%]c	ratio 3:4:5d	dr (3a)d	ee major [%]e
1f	A	Ni(OAc)2	THF	34	1:0.3:1	11:1g	−82
2f	B	Ni(OAc)2	THF	20	1:1:1	11:1g	−53
3f	A	Ni(OAc)2	DCM	22	1:0.3:1	18:1	n.d.
4f	A	Ni(OAc)2	PhMe	25	1:0.6:1	13:1	n.d.
5f	A	Ni(OAc)2	DMF	65	10:1:1	11:1	−58
6f	A	Ni(OAc)2	iPrOH	24	1:_:2	13:1	−78
7h	B	Cu(OAc)2	THF	<17	1:5:_	11:1g	54
8h	C	Cu(OAc)2	THF	15	1:4:_	12:1	74
9h	C	Cu(OAc)2	DCM	16	1:5:_	12:1	74
10h	C	Cu(OAc)2	PhMe	38	1.6:1:_	10:1	70
11h	C	Cu(OAc)2	iPrOH	77	>20:1:_	11:1	70
12h	D	Cu(OAc)2	iPrOH	86	8:1:_	9:1	66
13h	E	Cu(OAc)2	iPrOH	72	>20:1:_	10:1	64
14h	F	Cu(OAc)2	iPrOH	79	>20:1:_	10:1	83
15h	G	Cu(OAc)2	iPrOH	82	>20:1:_	10:1	91
16I	G	Cu(OAc)2	iPrOH	75	>20:1:_	13:1	90
17I,j	G	Cu(OAc)2	iPrOH	83	>20:1:_	13:1	91
18I,j	H	Cu(OAc)2	iPrOH	86	>20:1:_	11:1	92
19j,k	I	Cu(OAc)2	iPrOH	9	1:3:0.2	12:1	8

^aReactions were run on a 0.1 mmol scale; M(OAc)₂ refers to hydrate.

^bUsed 20 mol % ligand and metal acetate unless otherwise noted.

^cIsolated yield.

^dDetermined by ¹H NMR analysis of the crude reaction mixture, unless noted otherwise; _ = not observed

^eHPLC analysis using a chiral stationary phase, n.d. = not determined.

^fIsolated complex used.

^gIsolated dr.

^hCatalyst complexed in situ.

^IUsed 5 mol % of isolated complex.

^jConsisted of 1.5 equiv 2, 2.5 equiv Et₃N.

^kScale of 0.2 mmol. Catalyst complexed in situ; 5 mol % loading.

Switching to Cu(OAc)₂·H₂O as the metal source in THF completely eliminated the O-addition byproduct while favoring the opposite enantiomer of 3a (Table 1, entry 7: 54% ee, 11:1 dr),¹⁹ although the concurrence of the undesired hetero [3 + 2]-pathway was still observed. Diamine complexes of Ni(II) and Cu(II) exhibit distinct coordination geometries,²⁰ thereby leading to different coordination modes for the derived enolates. We postulate that these differences are manifested in opposite enantiofacial selectivity and preferential chiral recognition of geometrically different nitrones and nitrile oxides by the Ni(II) and Cu(II) complexes, respectively, via hydrogen bonding with the H-bond-donating diamine ligand.¹² A detailed discussion of this effect can be found in the SI. Along these lines, the inclination of Ni(II) enolates to undergo O-trapping may, in part, result from a kinetic preference for this pathway induced by the distinct geometry of Ni(II)-enolates.

Even though exchanging the cyclohexyl-substituted ligand B with the benzyl-type ligand C did not lead to an improvement in chemoselectivity, a noticeable enhancement in enantioselectivity was observed (Table 1, entry 8: 74% ee). Further solvent screening was performed. Whereas no improvement in chemoselectivity occurred in DCM (Table 1, entry 9: 1:5 3a/4a), 3a was favored slightly in PhMe (Table 1, entry 10: 1.6:1 3a/4a). Along with others, we have noted that reactions involving transition-metal enolates display enhanced rates in alcoholic solvent.^6h,21 In fact, the use of ⁱPrOH gave a >20:1 preference for the desired product 3a (Table 1, entry 11: 77% yield, 70% ee, 11:1 dr). The protic solvent may promote fragmentation of less-active catalyst oligomers,²² or catalyst turnover from the Cu(II)-alkoxide intermediate proceeding [3 + 2]-cycloaddition via protonolysis.

Having overcome the issue of low chemical yield, we shifted our focus toward the influence of ligand structure on enantioselectivity. The electronic features of the benzyl-type ligands D and E have little influence on enantioselection (Table 1, entry 12: 66% ee and Table 1, entry 13: 64% ee). However, the results with ligand F suggest that nonbonding interactions are important (Table 1, entry 14: 83% ee). Ligand G provided satisfactory levels of enantioselectivity (entry 15: 91% ee), and we were pleased to find that the catalyst loading could be reduced to 5 mol % without adversely influencing reactivity or selectivity (entry 16). Additionally, the amounts of hydroximoyl chloride 2a and Et₃N could be reduced to 1.5 equiv and 2.5 equiv, respectively (entry 17). The 2-CF₃-substituted benzyltype ligand H also provided high selectivity for this substrate (entry 18: 92% ee) and was later found to be uniquely suited for achieving high enantioselectivity with 2-substituted benzenenitrile oxides. Finally, the reduced yield and stereocontrol obtained with N-Me ligand I suggest the secondary diamine as a key design feature of the present catalyst system (entry 19: 9% yield, 8% ee).

With suitable reaction conditions in hand, we sought to learn how the steric and electronic parameters of the reaction partners might affect the process (Table 2). Nitrile oxides bearing electron-withdrawing groups in the 4-position were tolerated under the optimized conditions. For instance, 4- CF₃C₆H₄-substituted 3b and 4-FC₆H₄-substituted 3c were obtained in good chemical yields with high levels of diastereoand enantioselectivity (Table 2, 3b: 91% yield, 90% ee, 10:1 dr and 3c: 88% yield, 91% ee, 10:1 dr). Electron-rich aromatic nitrile oxides also participated, and 4-PhC₆H₄-substituted 3d and 4-MeC₆H₄-substituted 3e were obtained with similarly high levels of diastereo- and enantioselectivity (Table 2, 3d: 58% yield, 88% ee, 11:1 dr and 3e: 75% yield, 90% ee, 13:1 dr). Mesityl ligand G displayed lower levels of enantioselection when 2-substituted aryl nitrile oxides were used; however, the 2-CF₃ analogue H promoted high selectivity with these substrates. This effect was not as marked with smaller substituents, as 2-FC₆H₄-subsituted 3f was obtained with high levels of enantioselectivity using both catalysts (Table 2, ligand G: 90% ee and ligand H: 92% ee). The difference in selectivity increased upon moving from 2-OMe to the more sterically demanding 2-Me substitution (Table 2, 3g: 80% ee vs 92% ee and 3h: 76% ee vs 90% ee). The enantioselectivity difference was accentuated in the case of 2-BrC₆H₄-substituted 3i, which was obtained in 62% ee using the mesityl-substituted ligand G, whereas the 2-CF₃C₆H₄-substituted ligand H provided 91% ee. Heteroaromatic nitrile oxides were compatible with the optimized reaction conditions; the 2-keto glutaric acid-derived 3-pyridyl-substituted 3j could be formed, albeit with lower enantioselectivity (Table 2, 3j: 90% yield, 74% ee). Cycloadduct 3k was obtained in good yield with excellent diastereoselectivity, and with high levels of enantioselectivity (Table 2, 3k: 85% yield, 93% ee). The absolute stereochemistry of 3k was determined via X-ray crystallography, and the configurations of the remaining products are assigned by analogy. Other electronically diverse nitrile oxides reacted well with this α-keto ester (3l–o). The homobenzyl-substituted α-keto ester furnished the corresponding cycloadduct 3p in an 88% ee with high diastereoselection when the 2-FC₆H₄- substituted hydroxyimidoyl chloride was employed. In contrast, the n-propyl-substituted product 3q was formed with lower levels of diastereoselectivity. A γ-branched α-keto ester could be employed when the catalyst loading was increased to 20 mol % (Table 2, 3r: 45% yield, 14:1 dr, 83% ee). A β-aryl-substituted α-keto ester (R² = 3-tolyl) provided low diastereo- and enantiocontrol during the formation of the cycloadduct (59% yield, 1.3:1 dr, 62% ee and 68% ee, see SI for details). Similarly, lower levels of enantiocontrol were observed for a product derived from a Me-substituted α-keto ester (R² = Me) and 2- fluorobenzonitrile oxide, although satisfactory levels of diastereocontrol were maintained (79% yield, 54% ee, 7:1 dr, see SI for details). The isoxazoline, derived from tert-butyl pyruvate (R² = H), was furnished in racemic form (58% yield, 2% ee, see SI for details). The identity of the ester influenced the enantiocontrol, as lower levels of enantioselectivity were obtained for product 3s using the ⁱpropyl-substituted analogue of our model α-keto ester 1a (Table 2, 88% ee). Finally, alkenyl nitrile oxides are viable partners under the reaction conditions, and the styryl isoxazoline 3t was obtained in good yield, as a single diastereomer, with high levels of enantioselectivity (Table 2, 65% yield, 20:1 dr, 90% ee). Unfortunately, aliphatic nitrile oxides are not compatible at this stage of optimization (i.e. 3u). Aliphatic nitrile oxides are less stable than their sp²- carbon-substituted counterparts,¹⁵ and we believe that competitive decomposition is to blame.

Table 2.

Scope of Reaction^a

^aThe reactions were run on a 0.2 mmol scale, see SI for reaction times. Yields are for isolated products. Diastereomeric ratio (dr) values were determined by ¹H NMR analysis of the crude reaction mixture, isolated dr values were written in parentheses. The enantiomeric excess (ee) values were determined by HPLC using a chiral stationary phase.

^bLigand G employed.

^cLigand H employed.

^d20 mol % Cu(OAc)₂–diamine employed.

^e3.0 equiv of hydroximoyl chloride, 3.6 equiv of Et₃N.

Having established the scope of this transformation, we sought to study the reactivity of the C=N bond installed during the cycloaddition reaction (Scheme 2). We found that the 2-ketoglutaric acid-derived cycloadduct 3k could be prepared on a gram scale (2.5 mol % catalyst). Under the action of NaBH₄/NiCl₂,²³ 3k (90% ee) was converted to γ-lactam 6k containing three vicinal stereocenters (2.5 mmol scale). Lactam 6k was moderately enriched upon recrystallization, and the structure of this product was obtained via X-ray crystallography.²⁴

Scheme 2.

Gram Scale Reaction and Isoxazoline Reduction

The isoxazoline products delivered by the title reaction are hemiacetals that could in principle coexist with their ringopened acyclic γ-oximo-α-keto ester isomers (7, Scheme 3). Such structures would appear to be quite vulnerable toward racemization via facile keto–enol tautomerization because of the high C–H acidity of the methine proton (7 ⇆ 8). The obtention of isoxazoline adducts 3 and a downstream adduct such as 6k with high enantiomeric enrichment allows us to draw conclusions regarding the intervention of the equilibria depicted in Scheme 3. We hypothesize that the lack of racemization stems from the high stability of the isoxazoline hemiacetal that disfavors ring opening. Circumstantial evidence against the formation of the acyclic oxime 7 is the constant product diastereomeric ratio as a function of time: ring opening/ring closure would presumably change the diastereomeric composition. The kinetic reluctance to form the oxime 7 may stem from the system’s preference to avoid the creation of a highly electrophilic α-keto ester in the presence of multiple electronegative atoms/functional groups. The isoxazoline hemiacetal thus insulates a potentially labile stereocenter from undesired racemization.

Scheme 3.

Isoxazoline Hemiacetals Do Not Racemize via Tautomerization

SUMMARY

We have developed an enantioselective [3 + 2] cycloaddition reaction between nitrile oxides and transiently generated enolates of α-keto esters catalyzed by a chiral copper(II)– diamine complex. The catalyst system was found to be compatible with in situ nitrile oxide generation. This constitutes the first catalytic enantioselective preparation of this unique class of heterocycles, which can be transformed into interesting lactams. With this data in hand, we are currently assessing the reactivity of the 5-hydroxy-2-isoxazolines in other downstream transformations in addition to studying the mechanism of this transformation in detail. These results will be reported in due course.