Enantioconvergent synthesis of functionalized γ-butyrolactones via (3 + 2)-annulation.

A dynamic kinetic resolution of β-halo α-keto esters in an asymmetric homoenolate reaction is described. A chiral N-hetereocyclic carbene catalyzes the a(3) → d(3)-umpolung addition of α,β-enals to racemic α-keto esters, forming γ-butyrolactones with three contiguous stereocenters. The addition occurs with high regio-, diastereo-, and enantiocontrol. This methodology constitutes an intermolecular DKR process to set three stereocenters during the key bond forming event.

Dynamic kinetic asymmetric transformations are potent methods to generate functionalized and stereochemically defined products from racemic starting materials, and a number of enzymatic and chemocatalytic reactions have been put forward utilizing this paradigm.[1] Concomitant with the burgeoning number of dynamic methodologies has been an increase in the stereochemical complexity generated in these systems. At the bottom of this gradient resides methods that generate a single stereocenter; a prototypical example is the conversion of racemic alcohols into optically pure acetates enabled by redox processing (Scheme 1).[2] The second echelon in complexity-generation includes dynamic pathways that generate two stereocenters. The asymmetric hydrogenation of configurationally labile α-substituted β-keto esters is the archetypical example.[3]

Scheme 1

Degrees of Stereocomplexity in Dynamic Kinetic Transformations

Stereodynamic methods that generate three stereocenters are limited.[4,5] The preponderance of these methods use catalyst or substrate control to independently establish one stereocenter and dynamic bond formation to furnish the other two. The intramolecular DKR transformation of β-keto esters to β-lactones reported by Scheidt and co-workers serves as a counter-example, whereby simultaneous generation of all three stereocenters can occur during the same step.[6] To the best of our knowledge, an intermolecular DKR or dynamic kinetic asymmetric transformation (DyKAT) that establishes three stereocenters during the key bond-forming event is heretofore unknown. In this communication we describe a DKR utilizing carbene-generated homoenolate equivalents for the chemoselective formation of γ-butyrolactones from α,β-unsaturated aldehydes and racemic α-keto esters with excellent levels of diastereo- and enantiocontrol.

The exploitation of homoenolate (d3) nucleophiles generated by the union of N-heterocyclic carbenes (NHC) and α,β-unsaturated aldehydes has seen widespread use.[7] This method of catalytic umpolung (polarity inversion) has grown to include the use of imines,[8] carbonyls,[9] and Michael acceptors[10] as electrophilic components; however, to this point the reaction of enals with linear α-keto esters been reported in only low diasterocontrol (1.5:1) and moderate enantioselectivity (78% ee).[11] In connection with our interest in enantioconvergent carbon–carbon bond constructions involving racemic electrophiles,[12] we sought to develop NHC-catalyzed homoenolate addition of α,β-unsaturated aldehydes to configurationally labile α-keto esters. This endeavor presents a significant challenge in rate constant management: eight stereoisomers are possible and byproducts arising from mechanistically validated cross-benzoin[12c,13] and enal dimerization pathways were a legitimate concerns (Scheme 2).[14]

Scheme 2

Chemoselectivity Challenges for the Carbene Catalyzed Coupling of α,β-Enals (Blue) and Enolizable α-Keto Esters (Red)

Our studies began by examining the reaction of cinnamaldehyde and α-keto ester 1a-Me. A preliminary screen showed that NHC catalyst A(12c) delivers γ-butyrolactone 2 in low regio- and diastereoselectivity (Table 1, entry 1).[15] Using 1a, which has a more sterically demanding tert-butyl ester, in combination with A yielded 2a as a single product in a 6:1 diastereomeric ratio (dr) (Table 1, entry 2). Taking this result as an indication that a sterically hindered ester was likely necessary for the efficacy of this transformation, we began a systematic screening of carbene catalysts with 1a. Catalyst B and C revealed no marked increase in stereoselectivity (Table 1, entry 3–4). In tandem, these results indicated that increasing the steric bias of phenylalanine derived NHC catalysts was ineffectual to increasing reaction selectivity. Aminoindanol-derived catalyst D(16) showed poor differentiation between the acyl anion and homoenolate pathway yielding a 2:1 ratio of products 2a/3a (Table 1, entry 5). Catalyst E provided 2a as the sole product with a 3:1 dr and an enatiomeric ratio (er) of 78:22 (Table 1, entry 6).

Table 1

Catalyst and Substrate Optimization

entrya	R	cat	conv (%)	2a/3ab	drb,c	erd
1	Me	A	100	5:1	3:1	90:10
2	tBu	A	100	>20:1	6:1	84:16
3	tBu	B	88	>20:1	3:1	73:27
4	tBu	C	64	3.5:1	2:1	85:15
5	tBu	D	100	2:1	6:1	−
6	tBu	E	100	>20:1	3:1	78:22
7	tBu	F	100	>20:1	9:1	93:7
8	tBu	G	100	>20:1	33:1	99:1
9e	tBu	G	40	>20:1	>20:1	−
10f	tBu	G	100	>20:1	33:1	99:1

All reactions were run on a 0.10 mmol scale.

Determined by 1H NMR analysis of the crude reaction mixture.

dr is only reported for homoenolate product 2a.

Determined by chiral SFC analysis.

Conducted on the corresponding β-bromo analogue of 1a.

Five mol % of catalyst G.

Catalyst F,[17] which is a derivative of pyroglutamic acid, delivered exclusively 2a, in 9:1 dr and 93:7 er (Table 1, entry 7). Deploying catalyst G furnished 2a in 33:1 dr and 99:1 er (Table 1, entry 8). Using the β-bromo α-keto ester of 1a under identical conditions maintained high levels of isomer selectivity but suffered from poor reactivity (Table 1, entry 9). Lowering the catalyst loading to 5 mol % had no deleterious effects on reaction efficiency or selectivity (Table 1, entry 10).

With suitable conditions in hand we began to probe the allowable steric and electronic parameters of this annulation, initially by varying the identity of the α,β-unsaturated aldehyde (Table 2). Changing the electronic features of the aldehyde delivered 2b and 2c without loss of reaction fidelity. While both meta- and para-toluyl-derived cinnamaldehydes cleanly delivered 2d and 2e, the heightened steric encumbrance of ortho-methylcinnamaldehyde resulted in no reaction. Similarly, 2g and 2h, products that would arise from the addition of a trisubstituted alkene[12d] and (Z)-cinnamaldehyde, respectively, were inaccessible. Heteroaromatic 2i was isolated in 45:1 dr and 98.5:1.5 er, while 2j was obtained with 6:1 dr and 95:5 er. Products 2k and 2l demonstrated the viability of nonaromatic substitution, albeit with low dr for the addition of (E)-4-oxobut-2-enoic acid ethyl ester.

Table 2

Variation of α–β Unsaturated Aldehydes in the Homoenolate Addition to β-Chloro-α-Keto Estersa

All reactions were run on a 0.20 mmol scale at room temperature for 14 h. No acyl anion addition was observed for any example. Diastereomeric ratios were determined by 1H NMR; enantiomeric ratios by chiral SFC. Yields are of isolated products.

Using (Z)-cinnamaldehyde.

Yield shown is a 1H NMR yield of the major diastereomer utilizing mesitylene as an internal standard.

Variation of the α-keto ester also provided information regarding reaction scope (Table 3). Reducing the chain length of the starting α-keto ester delivered 2m with 1.5:1 dr and 94.5:5.5 er. Similarly, 2n was isolated with high enantioselectivity but as a 4:1 mix of isomers, while sterically encumbered 1o resulted in no reaction. Replacing chlorine with fluorine gave 2p in 4:1 dr and 97:3 er. Products 2q and 2r showcase the efficacy of substrates bearing β-propargyl and β-allyl substitution while heteroatom containing 2s was obtained in low yield but with 97.5:2.5 er.

Table 3

Variation of β-Halo α-Keto Esters in the Homoenolate Addition of Cinnamaldehydea

All reactions were run on a 0.20 mmol scale at room temperature for 14 h. Diastereomeric and product ratios were determined by 1H NMR; enantiomeric ratios by chiral SFC. Yields are of isolated products.

Using THF as the solvent.

Yield shown is a 1H NMR yield of the major diastereomer utilizing mesitylene as an internal standard.

Unambiguous differentiation between regioisomers and diastereomers was not feasible; an isomer ratio (ir) is reported.

Unambiguous identification of isomer ratios was not possible. Isolation of 2s was achieved via chromatography.

The reaction of 1a with (E)-3-(thiophen-2-yl)acrylaldehyde on a 1 g scale resulted in 84% yield of 2i as a single stereoisomer. An X-ray diffraction study was carried out to assign the relative and absolute stereochemistries as (3R,4R,5R) (Scheme 3).[18] The strong stereochemical influence of the β-chloro substitutent is manifested by the conserved anti-relationship between the nascent tertiary alcohol and the resident halogen.[12b,12c] This outcome is consistent with stereocontrol based on Felkin-Anh or Cornforth models.[19]

Scheme 3

Gram Scale Reaction of 1i and Cinnamaldehyde

In conclusion, we have developed the first stereoconvergent homoenolate reaction that utilizes racemic electrophiles. This NHC-catalyzed process between β-halo-α-keto esters and α,β-unsaturated aldehydes also constitutes the first intermolecular dynamic kinetic resolution in which three stereocenters are established during the enantiodetermining step. The resultant γ-butyrolactones bear a fully substituted glycolic acid moiety and are often obtained as single products in high diastereo- and enantioselectivity. Further manipulations of this product class and continued expansions of complexity-generating dynamic processes are of ongoing interest in our laboratory.