Application of Asymmetric Catalysis in the E/Z-Stereodivergent Synthesis of Alkenes

Abstract

Catalytic asymmetric reactions are primarily used to install new stereogenic centers highly enriched in either the R or S configuration. Here, we demonstrate that chiral catalysts can also be used to control the E/Z geometry of an alkene product devoid of chirality. The process is akin to a parallel kinetic resolution in the sense that a chiral catalyst converts two enantiomers of a starting material to two different products, in this case, only differing in the E/Z geometry of an alkene. The reaction is a reductive addition of a 1,1-dichloroalkene to a chiral secondary allylic alcohol. The catalyst controls the facial selectivity of addition to the allylic alcohol, and a subsequent anti-selective β-hydroxide elimination dictates the E/Z geometry of the alkene product. Synthetic applications of the skipped 1,4-diene products and studies probing the origin of stereoselectivity are described.

Graphical Abstract

INTRODUCTION

Control over E/Z geometry is one of the preeminent challenges in olefin synthesis (Figure 1A). Classically, the selective synthesis of an E or a Z alkene would rely on using two mechanistically distinct transformations (Figure 1B): for example, the semi-reduction of an alkyne under dissolving metal conditions (E-selective)¹ or using a transition metal-catalyzed hydrogenation reaction (Z-selective).² More recently, efforts have been directed at discovering catalytic reactions that can access both E and Z alkenes in a divergent fashion. In many cases, these processes involve a Z-selective reaction that can be coupled with an isomerization to the more thermodynamically stable E alkene.³ Less common are cases where both E and Z alkenes can be generated under kinetic control.⁴ If a general strategy for the catalyst-controlled, stereodivergent synthesis of E and Z alkenes could be identified, it would obviate the need to carry out challenging separations of olefin diastereomers or to plan entirely different routes to synthesize E and Z alkenes.

Figure 1.

Research background and reaction design. (A) E/Z diastereomers in natural products can have distinct biological functions and natural origins. (B) Classical approaches to the E and Z-selective alkene synthesis often require mechanistically distinct reactions. (C) In a typical catalytic asymmetric alkene addition reaction, the two enantiomers of the catalyst yield opposite enantiomers or diastereomers of the product when using achiral or chiral starting materials, respectively. (D) Here, we demonstrate that asymmetric catalysis can also be used to control alkene E/Z selectivity even in products that lack stereogenic centers.

Here, we demonstrate that the principle of asymmetric catalysis can be used to achieve E/Z diastereocontrol in the formation of an achiral olefin product. The reaction is a cobalt-catalyzed reductive coupling of a 1,1-dichloroalkene and an allylic alcohol to generate a skipped 1,4-diene product. In a typical catalytic asymmetric alkene addition, a chiral catalyst is used to add a reagent to one of the two prochiral faces of the alkene, yielding an enantioenriched product (Figure 1C).⁵ If the substrate contains a pre-existing stereocenter of defined configuration, the reaction becomes diastereoselective rather than enantioselective. We reasoned that if such a diastereoselective alkene addition could be coupled with a stereospecific β-elimination of an allylic leaving group, the E/Z geometry of the resulting alkene would be dictated by the enantiomer of catalyst used (Figure 1D).

RESULTS AND DISCUSSIONS

Our reaction design relies on a mechanistic observation that we made while studying catalytic enantioselective vinylidene additions to alkenes. Using a (PyBox)Co catalyst, additions of 1,1-dichloroalkenes to Z-alkenes generate highly enantioenriched axially chiral methylenecyclopropanes.⁶ However, when the substrate contained an allylic alkoxide, the methylenecyclopropane was not observed, and the product of β-alkoxide elimination was formed in high yield (Figure 2A).⁷ While there was no direct experimental evidence for the precise mechanism of elimination, DFT models suggested that a Lewis acid-assisted anti-elimination may be favored.

Figure 2.

(A) Observation of a β-alkoxide elimination in the cobalt-catalyzed reductive addition of 1,1-dichloroalkenes to 2,5-dihydrofuran. (B) Reductive additions of 1,1-dichloroalkenes to allylic alcohols yield achiral, skipped 1,4-diene products. The E/Z selectivity is determined by the chirality of both the allylic alcohol and the catalyst. (C) Proposed reaction mechanism and the origin of E/Z stereodivergence.

With this precedent in mind, we began our investigations by examining the reductive coupling of 1,1-dichloroalkene 1 and allylic alcohol 2 (Figure 2B). Using racemic alcohol 2 in combination with the cobalt catalyst bearing (S,S)-L1, the skipped 1,4-diene product 3 was obtained in 78% yield as a 1.3:1 mixture of E and Z diastereomers (entry 1). In a complementary experiment, the achiral ligand L2 in combination with (S)-2 (>99% ee) similarly yielded 3 as a 1:1 E/Z mixture (entry 2). Hypothesizing that the E and Z alkenes may be formed in two diastereomeric pathways, we next tested the enantiopure ligand (S,S)-L1 with enantiopure (S)-2. Accordingly, product 3 was formed in 65% yield as a 13:1 E/Z ratio. With the inverted ligand (R,R)-L1, the reaction favored Z-3 in a 1:13 E/Z ratio. Thus, the reductive vinylidene addition is not intrinsically E or Z selective but can be rendered highly diastereoselective using a chiral catalyst.

Based on the observed stereoselectivity of the reaction, a proposal for the origin of diastereoinduction is illustrated in Figure 2C. Sequential reduction of the Co(II) catalyst and oxidative addition of the 1,1-dichloroalkene generates the cobalt vinylidene intermediate II. Subsequent [2 + 2]-cycloaddition with the allylic alcohol is regioselective, placing the alcohol in a β-position relative to the Co–C(sp³) bond. The most sterically accessible trajectory would be on the side opposing the vinylidene aryl substituent and in a quadrant not occupied by the ligand i-Pr groups. Cobaltacyclobutane III then undergoes an anti-selective β-hydroxide elimination to generate the vinylcobalt intermediate IV. In the case of the (S,S)-L1/(S)-2 combination, the stereoselectivity of the [2 + 2]-cycloaddition and β-elimination steps yields an E-alkene, whereas the (R,R)-L1/(S)-2 combination yields a Z-alkene. Finally, vinylcobalt IV undergoes transmetallation with ZnX₂, and the resulting vinylzinc species is quenched by either the alcohol starting material or H₂O to yield 3.

The substrate scope of the stereodivergent vinylidene addition to allylic alcohols is summarized in Figures 3 and 4. A broad range of aryl- and heteroaryl-substituted 1,1-dichloroalkenes react effectively. Functional groups commonly used in transition metal-catalyzed cross-coupling reactions, such as boronate esters (6) and aryl chlorides (7), are well-tolerated.

Figure 3.

Substrate scope of 1,1-dichloroalkenes. Standard reaction condition: 1,1-dichloroalkene (1.0 equiv), 2 (2.0–3.0 equiv), Co(dme)Br₂ (15 mol%), (S,S)- or (R,R)-L1 (18 mol%), Zn (3.0 equiv), ZnI₂ (3.0 equiv), DMA (0.13 M), 40 °C, 24 h. Isolated yields and E/Z ratios were determined after purification by column chromatography. ^a4 (2.0 equiv) used instead of 2 and ZnBr₂ used instead of ZnI₂. ^bReactions run for 3 d. ^cMn used instead of Zn. See Supporting Information for additional details.

Figure 4.

Substrate scope of allylic alcohols. Standard reaction condition: 17 (1.0 equiv), allylic alcohol (2.0–3.0 equiv), Co(dme)Br₂ (15 mol%), (S,S)- or (R,R)-L1 (18 mol%), Zn (3.0 equiv), ZnI₂ (3.0 equiv), DMA (0.13 M), 40 °C, 24 h. Isolated yields and E/Z ratios were determined after purification by column chromatography. See Supporting Information for additional details.

Tetrasubstituted 1,1-dichloroalkenes did not yield any coupled products using allylic alcohol 2. Instead, the 1,1-dichloroalkene was converted to the corresponding monochloroalkene. We hypothesized that undesired protonation of the cobalt(chloroalkenyl) intermediate may be predominating over further reduction to the vinylidene for these more hindered substrates. Accordingly, the allylic benzyl ether (S)-4 lacking an acidic proton was employed to accomplish these couplings (products 13–16).

Aryl, heteroaryl, and alkyl substituents can be incorporated into the allylic alcohol partner. For substrates containing multiple alkenes (product 25), terminal alkenes react in preference to more hindered disubstituted and trisubstituted alkenes. By taking advantage of the numerous advances in asymmetric synthesis, chiral secondary alcohols of the type used in this reaction are readily available in highly enantioenriched form using methods such as kinetic resolutions of secondary alcohols⁸ and asymmetric vinyl nucleophile additions to aldehydes.⁹

Natural products and bioactive molecules containing chiral tertiary allylic alcohols, such as linalool, sclareol and nabumetone, proved to be suitable substrates for the reaction, yielding products 26–29 with high E/Z selectivity. Compound E-30 is a natural product isolated from the Ottelia genus and functions as an inhibitor of α-glucosidase.¹⁰ By using the requisite starting materials, E-30 and its unnatural diastereomer Z-30 are accessible with high diastereoselectivity. The skipped 1,4-diene E-31 shows inhibitory activity towards the metamorphosis of silkworm larvae.¹¹ Previously, E-31 was synthesized using a Wittig olefination as a mixture of stereoisomers. By contrast, E-31 and Z-31 could be independently synthesized with a >20:1 dr using the vinylidene and allylic alcohol coupling reaction.

The 1,4-diene substructure can also be found in fragrance compounds and precursors, which are commonly derived from terpene natural products. The fragrance precursor Z-32 can be synthesized efficiently, and the novel analog E-32 is also accessible.¹² Finally, 1,4-dienes are also featured as intermediates in total synthesis. For example, E-33 was used in Nicolaou’s synthesis of the endiandric acids, and it was prepared in 9 steps (35% overall yield) as a 1:1 E/Z mixture at the C3–C4 alkene.¹³ Only the E-isomer possesses the requisite stereochemistry for the synthesis of endiandric acids A–Gm, necessitating separation from the undesired Z-isomer by column chromatography. The vinylidene coupling reaction provides E-33 in 51% yield with >20:1 E/Z selectivity at the C3–C4 alkene and a more modest 4:1 E/Z selectivity at the C6–C7 alkene.

To take advantage of the putative organozinc intermediate in this reaction, we next sought to prevent its protonation by replacing the free alcohol in the substrate with a methyl ether (Figure 5B). The reductive coupling of 1,1-dichloroalkene 17 and allylic methyl ether 34 yielded 36 in comparable yield and stereoselectivity to that obtained with the corresponding allylic alcohol. When the crude reaction mixture was quenched with CD₃OD, 72% deuterium incorporation was observed at C2, consistent with the proposed identity of the alkenylzinc species. Furthermore, quenching with NIS or catalytic cross-coupling yielded the products of iodination (37), arylation (39),¹⁴ and allylation (41).¹⁵ The corresponding Z-selective reaction can also be carried out to yield Z-41. Thus, the reductive coupling provides streamlined access to stereodefined trisubstituted skipped dienes and trienes that would be challenging to prepare using other approaches to olefin synthesis.

Figure 5.

(A) Diastereoselective synthesis of natural products, bioactive compounds, and synthetic intermediates containing 1,4-dienes. (B) Functionalizations of the alkenylzinc intermediate. (C) Divergent reactivity of (S)-cyclopent-2-en-1-ol using the (S,S)- and (R,R)-enantiomers of the catalyst.

The E/Z stereoselectivity of the vinylidene coupling reaction is ultimately determined by both the facial selectivity of the alkene addition step and the syn/anti selectivity of the β-elimination step.¹⁶ We therefore sought to design a reaction that would allow us to interrogate these two steps independently. Accordingly, we examined a reaction between 1,1-dichloroalkene 42 and (S)-cyclopent-2-en-1-ol (43). Using (S,S)-L1, there are two [2 + 2]-cycloaddition transition states that would allow the alkene to approach in the open quadrant of the Co=C=CR intermediate (44a and 45a). Intermediate 45a suffers from prohibitive steric interactions, and the corresponding product 47b is not observed. Instead, the chiral skipped diene (S)-46 is formed in 54% yield and >99% ee, indicating very high facial selectivity in the [2 + 2]-cycloaddition step. Using (R,R)-L1, the steric interaction between the hydroxyl group and the aryl substituent of the cobalt vinylidene is less pronounced, and both [2 + 2]-cycloaddition transition states are energetically accessible. The [2 + 2]-cycloaddition transition state leading to 45b places the hydroxyl group in a γ-position relative to the Co–C(sp³) bond. Thus, β-elimination cannot occur, and the methylenecyclopropane 47 is formed in 48% yield and >99% ee. In the alternative metallacycle 44b, the hydroxyl group is in a suitable position for β-elimination but is syn to the Co–C(sp³) bond. Both the syn β-elimination and C–C reductive elimination pathways are competitive, and products (R)-46 and 47a are formed in 17% and 6% yield, respectively, both with >99% ee.

With these results in hand, a DFT model for the origin of diastereoselectivity was developed (Figure 6A and Figure 6B). Using (S,S)-L1, the most favorable [2 + 2]-cycloaddition transition state of the Co=C=CR intermediate involves the approach in the least hindered quadrant away from both the vinylidene substituent and the ligand i-Pr groups. Additionally, the hydroxy group is in the β-position relative to the Co–C(sp³) bond, making it suitable for the subsequent elimination step. The regioisomeric transition state in which the hydroxyl group is γ to the Co–C(sp³) bond is higher in energy by 9.1 kcal/mol due to unfavorable steric interactions between the ligand i-Pr groups and the ZnCl₂-coordinated alcohol (see Supporting Information for additional details).

Figure 6.

(A) DFT model for the major reaction pathway using the (S,S)-catalyst; (B) DFT model for the major reaction pathway using the (R,R)-catalyst. DFT method: UM06L/6–311G(d,p); (C) Experiments probing the relative rates of the pro-E and pro-Z [2 + 2]-cycloaddition and β-elimination steps.

Metallacycle S-B-48a undergoes a nearly barrierless (0.7 kcal/mol) ZnCl₂-assisted anti β-elimination to form the E-alkene product. The alternative syn β-elimination to form a Z-alkene product has a higher barrier of 10.9 kcal/mol. For (R,R)-L1, the [2 + 2]-cycloaddition step generates the diastereomeric metallacycle R-B-48a. Anti β-elimination is favored over syn β-elimination by a smaller energy difference of 2.6 kcal/mol, and formation of the Z-alkene product is favored. The higher barrier for β-elimination using the (R,R)-enantiomer of the catalyst is likely due to the need to access a higher-energy rotamer in order to place the OH group in an anti orientation.

Additional experiments were carried out to probe key aspects of the proposed mechanism. In the coupling of 1,1-dichloroalkene 1 and the racemic allylic alcohol 2 (2.0 equiv), recovered 2 at the end of the reaction (52% conversion of 2) has an ee of only 9%, indicating that there is minimal kinetic resolution. This result suggests that there is no significant kinetic preference for addition to the two enantiomers of the allylic alcohol. In a complementary experiment, the symmetrical bis(allylic) alcohol 49 was reacted with 1,1-dichloroalkene 17 and produces the triene product 50 in 45% yield as a 1:1.5 E/Z mixture. Thus, the catalyst does not distinguish between the two alkenes leading to the pro-E and pro-Z metallacycles.

DFT models suggest that both the pro-E and the pro-Z β-elimination have low activation barriers. To experimentally interrogate this step, a reaction was carried out using the vinyl-substituted cyclic acetal 51. Using (S,S)-L1, the vinyl ether product 52 was obtained in 66% yield as a 1.3:1 E/Z mixture. Therefore, following the [2 + 2]-cycloaddition step and formation of the cobaltacycle intermediate, there is no intrinsic preference for the pro-E or pro-Z β-elimination transition states.

CONCLUSIONS

In summary, the asymmetric reductive vinylidene and allylic alcohol coupling reaction demonstrates that chiral catalysts can be used to control diastereoselectivity even in the formation of achiral products. Whereas most asymmetric addition reactions to chiral substrates result in the formation of a new stereogenic center in a diastereoselective process, a unique feature of this reaction is that both stereogenic centers are eliminated to generate an alkene. By using two different enantiomers of the catalyst, it is possible to access either the E or Z alkene in a stereodivergent fashion with high selectivity. DFT models indicate that the E/Z selectivity is dictated both by the alkene facial selectivity in the [2 + 2]-cycloaddition step and by the preference for anti over syn β-elimination.

The products of these reactions are skipped polyenes where control over E/Z geometry is challenging using other coupling or olefination strategies. Furthermore, the product of the coupling is an alkenylzinc species generated from a transmetallation from cobalt to zinc. Thus, allylic ether substrates, which lack acidic protons, can be used to retain the alkenylzinc functionality and enable a tandem functionalization to form trisubstituted alkenes of defined stereochemistry. Applications of this anti-selective β-elimination process in other stereoselective transformations are the subject of ongoing investigation.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information. Experimental procedures, spectroscopic data, X-ray crystallographic data, and computational details (PDF). This material is available free of charge via the Internet at http://pubs.acs.org.