Enantioselective Potassium-Catalyzed Wittig Olefinations

Abstract

We report asymmetric potassium-isothiourea-boronate–catalyzed Wittig olefinations of 4-substituted cyclohexanones with non-stabilized phosphorus ylides to afford highly enantioenriched axially chiral alkenes. The optimal catalyst features an unusual macrocyclic amide-potassium-boronate chelate. Kinetic and spectroscopic analyses are consistent with a Lewis acid mechanism for the catalytic olefination that results in formation of the oxaphosphetane adduct under cryogenic conditions. Thermal fragmentation of the oxaphosphetane to the alkene product occurs after the reaction is complete. Computational studies indicate that cycloaddition proceeds via a stepwise mechanism involving enantiodetermining polar 1,2-addition to afford an intermediate potassium betaine complex.

Graphical Abstract

The Wittig reaction has stood as one of the essential methods for stereoselective alkene synthesis since its initial report,^1,2 finding continuous application in both industry and academia to this day.^3-7 Olefination of suitably substituted prochiral ketones permits the formation of axially chiral alkene products,^8-13 and numerous approaches to asymmetric Wittig and Horner–Wadsworth–Emmons reactions have been demonstrated using stoichiometric chiral controlling elements,¹⁴ with canonical examples relying on covalent chiral auxiliaries^15-17 or chiral ligands.^18,19 In contrast, examples of asymmetric catalysis of Wittig-type olefinations to access axially chiral products are exceedingly rare; to our knowledge there exist only three reported examples, using either Brønsted acid, H-bond-donor, or phase-transfer catalysts and attaining up to 75% enantiomeric excess (ee) (Scheme 1A).^20-23 Herein, we report highly enantioselective Wittig reactions of 4-substituted cyclohexanones catalyzed by a novel potassium-isothiourea-boronate complex.

Scheme 1.

Synthetic and Mechanistic Aspects of Stereoselective Wittig Olefinations

The generally accepted mechanism for the Wittig reaction involves an initial irreversible asynchronous concerted [2+2] cycloaddition between a carbonyl compound and a phosphorus ylide to form a mixture of cis- and trans-oxaphosphetane (OPA) species, followed by an irreversible stereospecific cycloreversion to afford the corresponding alkene and phosphine oxide (Scheme 1B).^24-27 Zwitterionic “salt-free” betaine species have not been observed during Wittig reactions, and have been conclusively excluded as possible intermediates in the reactions of certain classes of ylides.²⁸ However, Lewis acidic metal cations (typically Li⁺) have been shown to stabilize betaine species and thereby influence the outcomes of Wittig olefinations.^29-32 Reactions of aldehydes with non-stabilized ylides exhibit a substantial decrease in Z-selectivity upon addition of lithium salts, consistent with the establishment of a lithium-mediated cycloaddition pathway via a betaine intermediate with lower kinetic selectivity for the formation of the cis-OPA (Scheme 1C).^24,29,33,34 Additionally, reactions of ketones with stabilized ylides are accelerated dramatically in the presence of lithium salts.³⁵

Recently, our group discovered novel chiral lithium-isothiourea-boronate derivatives that catalyze highly enantioselective Matteson homologation reactions (Scheme 1D).^36,37 Given the observed effects of lithium salts on the rate and stereoselectivity of standard Wittig reactions, we considered whether this new family of chiral Lewis acids might exert enantiocontrol over Wittig reactions generating axially chiral alkenes. In particular, we envisioned that chiral alkali metal-based Lewis acids might catalyze the addition of non-stabilized phosphorus ylides to 4-substituted cyclohexanones to generate enantioenriched OPA intermediates (Scheme 1E). The resulting intermediate could then undergo stereospecific cycloreversion on warming to afford an enantioenriched axially chiral alkene product. Key to this strategy is the low-temperature stability of OPAs derived from non-stabilized ylides,^25,29 as the Lewis basic phosphine oxide byproduct of fragmentation to the alkene would likely be a potent catalyst poison.^38-40

A preliminary assay of alkali metal-isothiourea-boronate derivatives developed in our study of the Matteson reaction revealed that while lithium- and sodium-based catalysts Li-3a and Na-3a afforded only racemic product in the olefination of 4-phenylcyclohexanone with ylide 2a, the potassium-isothiourea-boronate derivative K-3a catalyzed the model reaction in 64% ee (Table 1). Reaction enantioselectivity proved highly responsive to the identity of the arylpyrrolidine moiety on the catalyst, with the unsubstituted derivative K-3h affording nearly racemic product and the 3-phenanthryl–substituted K-3f inducing the highest levels of selectivity (92% ee). The Na analog Na-3f was less enantioselective (30% ee), while the Li derivative afforded only racemic product. While removal of the α-boryl substituent had little effect on ee (K-3g, 90% ee), catalysts epimeric at the α-boryl stereocenter (K-epi-3a–f) promoted very poorly selective reactions in all cases.

Table 1.

Metal-Isothiourea-Boronate Catalyst Structure-Enantioselectivity Relationships

We sought to elucidate the basis for the ligand effects and the superior performance of potassium-based catalysts in this transformation. An X-ray crystal structure analysis of K-3f·THF revealed an unusual macrocyclic chelate in which the potassium cation is coordinated between the amide oxygen and the anterior boronate oxygen (Figure 1A). This ligation mode is dramatically different from the five-membered amide-Li-isothiourea chelate previously observed in the X-ray structure of a lithium-isothiourea-boronate derivative.^36,37

Figure 1.

Characterization of alkali metal-isothiourea-boronate catalyst structures by (A) single-crystal X-ray diffraction; (B) IR spectroscopy; and (C) NMR studies. Hydrogen atoms and close contacts in the crystal structure of K-3f·THF are omitted for clarity. (D) Electrostatic potential map of anion derived from deprotonation of 3f. Red = electron-rich, blue = electron-poor. (E) Proposed model of catalyst structure-enantioselectivity relationship.

To assess the relevance of these solid-state structural differences to the corresponding solution structures, solution IR spectra of M-3f (M = Li, Na, K) were acquired and compared to spectra of the two chelates predicted by DFT (Figure 1B). The unmetallated isothiourea-boronate 3f (M = H) exhibits absorbances corresponding to the amide C─O (1660 cm⁻¹) and isothiourea N─C─N (1590 cm⁻¹) moieties that are closely reproduced in the calculated spectrum. The spectra of K-3f and Na-3f display similar patterns in the carbonyl region, with a redshift of the C─O stretch by roughly 40 cm⁻¹ and a flattening of the N─C─N stretch relative to the spectrum of 3f. These spectral changes are in good agreement with predicted IR spectra of the amide-metal-boronate chelates. In contrast, the spectrum of Li-3f reveals a redshift of both peaks by roughly 40 cm⁻¹, in good agreement with the calculated amide-metal-isothiourea chelate (for in-depth discussion of IR spectra, see SI).^36,37

These fundamental differences in chelation mode were found to translate into differences in relative Lewis acidity, as demonstrated in a Gutmann–Beckett analysis^38,39 (Figure 1C). A slight tempering of the Lewis acidity of K-3f relative to KHMDS and of Na-3f relative to NaHMDS is observed, whereas a significant increase in Lewis acidity is measured for Li-3f relative to LiHMDS. In an electrostatic potential map of the 3f-derived isothiourea-boronate anion (Figure 1D), a high degree of electron density appears at the boronate oxygen atoms, consistent with the observed buffering of Lewis acidity in the two proposed amide-metal-boronate chelates. The low enantioselectivities obtained with catalysts K-epi-3a–f (Table 1) can be attributed to the steric effect of the syn α-boryl methyl group, which likely interferes with formation of the amide-K-boronate chelate (Figure 1E).

The scope of enantioselective olefinations catalyzed by K-3f was examined with a set of representative cyclohexanone and ylide substrate combinations (Figure 2). Reactions proceeded effectively with ylide derivatives bearing linear alkyl substituents, with the targeted products generated in high yields and generally high enantioselectivities (5a–d). A particular sensitivity to the steric properties of the ylide was noted, as evidenced by a dramatic decrease in ee observed with a β-phenyl derivative (compare 5a with 5e). Cyclohexanone derivatives bearing a variety of C4 aryl substituents (5f–h) underwent olefination with excellent levels of enantioselectivity, with only slightly lower selectivities obtained for 4,4-disubstituted derivatives (5i, 5j). Reactions with C4 alkyl-substituted derivatives were slightly less enantioselective, with a relatively minor effect associated with substituent steric variation (77-83% ee, 5k–m).

Figure 2.

Substrate scope. All reactions were performed on 0.1 mmol scale. Yields correspond to isolated yields of purified products. See Supporting Information for details on the assignment of absolute configuration. ^aReaction time 4 hours. ^bReaction time 48 hours. ^cNMR yield.

We sought to glean insight into the basis for the intriguing transannular electronic effects on ee revealed in the substrate scope study (compare 5c with 5f). Through-space electric-field effects have been documented previously in hydride additions to 4-substituted cyclohexanones.⁴¹ Computed free energy of activation barriers for the uncatalyzed olefination of a series of para-substituted 4-arylcyclohexanones were found to correlate well with the calculated electric field intensity along the forming C─C bond in an equatorial addition of the ylide (Figure 3A). An examination of the K-3f–catalyzed olefinations of the same set of substituted 4-arylcyclohexanones (Figure 3B) revealed that higher enantioselectivity was obtained with substrates bearing more electron-deficient aromatic rings. The enantioselectivity data, expressed as ΔΔG^‡ (= −RTln(enantiomer ratio)), was plotted against several different standard parameters and found to correlate best with σ_meta (Figure 3C, see SI for correlations with other parameters).^42,43 A significant effect on rate was also observed, with electron-withdrawing 4-substituents leading to faster reactions in both the catalyzed and uncatalyzed olefinations (Figure 3D).

Figure 3.

(A) (top) Proposed model of electric field–induced rate acceleration; (bottom) Computed electric field intensity along the reaction axis correlates with computed rates of olefination for para-substituted 4-phenylcyclohexanones. (B) Model reactions performed to study the effect of arylcyclohexanone substitution on ee. (C) Correlation between experimental enantioselectivities and the σ_meta parameter for R. (D) One-pot competition experiments reveal faster rates in both the catalyzed and uncatalyzed reaction for the more enantioselective substrate.

A full kinetic analysis of the olefination of ketone 1a was performed with data collected by UV-visible spectroscopy, monitoring the disappearance of the intensely orange-red ylide 2d at 520 nm (Figure 4). Plots of the concentration versus time data in a pair of “same excess” experiments with different initial concentrations (Figure 4A, black and red traces) revealed poor graphical overlay, consistent with a catalyst deactivation process.⁴⁴ This deactivation could be ascribed directly to product inhibition of K-3f by OPA 4d, as excellent overlay was obtained by addition of the appropriate amount of racemic OPA 4d to the reaction with lower initial concentrations (Figure 4A, blue trace). A set of “different excess” experiments (Figure 4B) revealed approximate doubling of reaction rate with a doubling of initial ylide concentration, but only a small increase in rate with a doubling of initial ketone concentration. These observations are most consistent with a first-order kinetic dependence on ylide and saturation kinetics in ketone. Variation of the total catalyst concentration [K-3f]_T revealed a first-order kinetic dependence (Figure 4C), which in combination with the absence of a nonlinear effect (Figure 4D) is consistent with a monomeric catalyst species throughout the catalytic cycle.

Figure 4.

Kinetic studies and computational analysis of K-3f–catalyzed olefination. (A) Same-excess experiments revealing evidence for inhibition by the OPA product. (B) Different-excess experiments consistent with first-order kinetic dependence on [ylide] and a shift from zero-order to positive-order dependence on [ketone] over the course of the reaction. (C) First-order kinetic dependence on [K-3f]_T. (D) Absence of nonlinear effect. (E) Proposed catalytic cycle. (F) Proposed rate law and fitted kinetic parameters. (G) Computational analysis of enantiodetermining step reproduces experimental selectivity.

These kinetic results lead us to postulate the catalytic cycle shown in Figure 4E, in which free catalyst K-3f reversibly binds ketone 1a to form complex A, followed by turnover-limiting and enantiodetermining irreversible addition of ylide 2d to form betaine complex B, which reversibly dissociates the OPA product 4d to return free K-3f. Consistent with the proposed inhibition of K-3f via reversible P─O bond cleavage of 4d to form B, cleavage of OPA intermediates by LiBr to form lithium betaine species has been demonstrated to be facile at −78 °C.²⁹ On the basis of the kinetic data, it can be estimated that K-3f exhibits a threefold preference for binding of OPA 4d over ketone 1a. This would induce a shift in catalyst resting state over the course of the reaction from the ketone complex A to the betaine complex B, manifesting as a roughly zero-order kinetic dependence on ketone in the early stages of the reaction that develops into a first-order dependence as the reaction progresses.

DFT calculations of the putative catalyzed enantiodetermining transition states leading to product 5b with the 1:1:1 catalyst:ketone:ylide stoichiometry supported by the kinetic analysis reproduced both the correct sense and magnitude of enantioinduction (Figure 4G). The reaction proceeds via a polar 1,2-addition to forge the C─C bond and afford an intermediate potassium betaine complex.³⁴ Overall transition state geometry appears to be dictated primarily by two factors: (1) facial selectivity for equatorial addition to the cyclohexanone, and (2) minimization of steric interactions between the triphenylphosphonium moiety and the catalyst framework, dictating the orientation of the ylide substituent. Bulky ylide substituents may hinder steric differentiation with the triphenylphosphonium moiety, leading to lower enantioselectivity (Figure 2, 5e). While this computational model recapitulates the observed enantioselectivity in the formation of 5b, the exact basis for enantioinduction is complex and not fully elucidated at this stage.

In summary, highly enantioselective Wittig olefinations of 4-substituted cyclohexanones are catalyzed by a chiral Lewis-acidic potassium-isothiourea-boronate complex possessing a novel macrocyclic chelate structure. The enantiodetermining cycloaddition is a stepwise process involving irreversible polar 1,2-addition to form a potassium betaine complex followed by a reversible cyclization. Future studies will focus on application of the mechanistic observations made here to enantioselective catalysis of other challenging alkali-metal–mediated transformations.