Allenoates in Enantioselective [2+2] Cycloadditions: From a Mechanistic Curiosity to a Stereospecific Transformation

Abstract

Identification of a novel catalyst–allenoate pair allows enantioselective [2+2] cycloaddition of α-methylstyrene. To understand the origin of selectivity, a detailed mechanistic investigation was conducted. Herein, two competing reaction pathways are proposed, which operate simultaneously and funnel the alkenes to the same axially chiral cyclobutanes. In agreement with the Woodward–Hoffmann rules, this mechanistic curiosity can be rationalized through a unique symmetry operation that was elucidated by deuteration experiments. In the case of 1,1-diarylalkenes, distal communication between the catalyst and alkene is achieved through subtle alteration of electronic properties and conformation. In this context, a Hammett study lends further credibility to a concerted mechanism. Thus, extended scope exploration, including β-substitution on the alkene to generate two adjacent stereocenters within the cyclobutane ring, is achieved in a highly stereospecific and enantioselective fashion (33 examples, up to >99:1 er).

Graphical Abstract

INTRODUCTION

Construction of multiple stereocenters in a single event has become an increasingly important strategy to build molecular complexity in an efficient and economical way. Arguably, one of the most powerful methods to achieve such a transformation is the Diels–Alder reaction.¹ It is generally postulated that the aforementioned reaction involves a symmetry-allowed cyclic transition state as predicted by the Woodward–Hoffmann rules.² Consequently, the reaction can be performed in a stereospecific fashion, allowing the generation of all possible isomers from the respective E- or Z-alkenes. Despite significant advances in the realm of enantioselective Diels–Alder reactions,³ the analogous, concerted [2+2] cycloaddition of alkenes has remained a challenge. Particularly, methods that utilize activated olefins are especially difficult. In contrast to the Diels–Alder reaction, recent methods to construct enantioenriched arylcyclobutanes by [2+2] cycloaddition often proceed through stepwise processes, resulting in decreased reaction selectivity (Scheme 1a).^4,5 With respect to Lewis acid catalyzed examples, gold complexes have been utilized to activate alkynes⁶ or allenes⁷ to achieve enantioselective cycloaddition with highly activated styrene derivatives. In addition, copper and zinc have been used with electron rich arylalkenes to give cyclobutanes with good control of enantioselectivity.⁸ Alternatively, chiral amines can be used to assemble cyclobutanes via ionic intermediates.⁹ Photochemical methods¹⁰ often exhibit increased tolerance for electron-poor styrenes, but generally involve excited state biradical intermediates, which lead to either stereoconvergence^10f or erosion of diastereoselectivity.^10g Overall, only a few examples,^9a report good stereospecificity through trapping of reactive intermediates at low temperatures.

Scheme 1.

Enantioselective Arylcyclobutane Synthesis

To address this problem, we recently focused our efforts on rendering alkene-allenoate cycloadditions enantioselective.¹¹ Based on several reports in the literature,^11b,12 alkene-allenoate cycloadditions appear to be concerted and therefore stereo-specific in nature, which provides a unique opportunity for an in-depth study (Scheme 1b). Herein, we report the enantioselective [2+2] cycloaddition of α-methylstyrene through identification of a novel alkene-allenoate pair. This method does not only enable catalytic enantioselective formation of quaternary carbon centers¹³ but also exemplifies how elucidation of reaction mechanism can go hand in hand with expansion of scope. Ultimately, we propose models that reliably explain the observed selectivity for a range of activated alkenes in this unique [2+2] cycloaddition.

RESULTS AND DISCUSSION

Optimization.

We initially envisioned accessing cyclobutanes bearing a quaternary center by using activated 1,1-disubstituted alkenes (e.g., α-methylstyrene). Preliminary data suggested that the identity of the allenoate ester has a significant influence on modulating reactivity as well as selectivity.^12d As such, investigations of various allenoates 1 were undertaken (Table 1).

Table 1.

Reaction Optimization

Entry	XR	Ar	Yieldb	erc
1	OBn (1a)	Ph (2a)	95%	78:22
2	OCH2CF3 (1b)	Ph (2a)	46%	76:24
3	SBn (1c)	Ph (2a)	72%	81:19
4	SBn (1c)	3,5-(CH3)2-C6H3 (2b)	71%	90:10
5	SBn (1c)	3,5-(tBu)2-C6H3 (2c)	90%	80:20
6	SBn (1c)	3,5-(OMe)2-C6H3 (2d)	56%	87:13
7	SBn (1c)	3,5-(CF3)2-C6H3 (2e)	96%	96:4
8	OBn (1a)	3.5-(CF3)2-C6H3 (2e)	87%	86:14

^aSee the Supporting Information for experimental details.

^bDetermi-nation by ¹H NMR of the crude reaction mixture utilizing a calibrated standard.

^cDetermined by HPLC analysis using a chiral column.

Changing from benzyl (1a) to the more reactive 2,2,2-trifluoroethyl ester (1b) under our previously optimized reaction conditions resulted in decreased yield, due to competitive polymerization of the starting materials under the reaction conditions (Table 1, compare entries 1 and 2). We next examined the use of thiobenzyl allenic ester 1c in the reaction and were pleased to find a slight increase in reaction selectivity, albeit with decreased overall yield (compare entries 1 and 3). Confident that reaction yield could be improved through catalyst control, we studied modifications of the diarylprolinol scaffold. Whereas increasing the steric size of the aryl groups from phenyl (2a) to xylyl (2b) provided a significant increase in reaction selectivity, further increase to sterically bulkier 3,5-(tBu)₂-C₆H₃ catalyst 2c resulted in a substantially less selective reaction (compare entries 3–5). Investigation of more electron rich aryl groups (i.e., entry 6, 2d) resulted in a marked decrease in overall reaction yield while providing higher reaction selectivity. Significant improvements in both reaction yield and enantioselectivity was observed when more electron deficient 3,5-(CF₃)₂-C₆H₃ catalyst 2e was examined in the reaction (entry 7). Interestingly, utilizing the same catalyst with benzyl allenoate resulted in a smaller increase in enantioselectivity (compare entries 1 and 8).

Mechanism.

To account for the observed enantiomer obtained in the reaction we propose that upon binding of the Lewis basic carbonyl oxygen to the Lewis acidic boron atom,¹⁴ the orientation of the allenoate may be fixed by a putative C— H⋯O hydrogen bonding interaction (Scheme 2a).¹⁵ As the bottom face of the allenoate is effectively blocked by the large aryl groups of the catalyst, approach of the alkene may only occur from the top face. Additionally, the sterically large phenyl group of the alkene is oriented distal to the large catalyst–substrate complex. Conveniently, the planar character of the phenyl group thereby also minimizes steric interaction with the protruding C–H bond of the allene, resulting in two plausible transition states (Scheme 2a, TS-A1 and TS-B1) for alkene approach.

Scheme 2.

Model for Enantioselectivity

We propose a concerted, asynchronous [_π2_s+(_π2_s+_π2_s)] cycloaddition in which the direction of rotation of the electron deficient allenic π-bond is dictated by the Woodward–Hoffmann rules (indicated by the blue arrows).^2,16 Because of the unusual symmetry of the system, both transition states lead to the same enantiomer. The absolute configuration of the cycloadduct 3c was proven through hydrolysis of the thioester and subsequent analysis of the corresponding carboxylic acid 4 via X-ray diffraction.¹⁷ To distinguish TS-A1 and TS-B1, cis-β-deutero-α-methylstyrene 5 was subjected to the reaction conditions (Scheme 2b). To our surprise, the cycloadduct 6 was obtained as an 83:17 Z:E mixture suggesting that both pathways are operating. Accordingly, trans-β-deutero-α-methylstyrene 7 furnished cyclobutane 8 as a 86:14 Z:E mixture. Assignment of the respective E- and Z-isomers was achieved through derivatization and subsequent NOE analysis of the respective tertiary alcohol 9 (see Supporting Information for details). Thus, we were able to deduce TS-A1 to be energetically favored, which can be explained by the alkene being distal to the bulky boroaryl group of the catalyst. In addition, the cycloaddition was highly stereospecific, as indicated by the two different pairs of products generated from the respective cis- and trans-deuteroalkene. This suggests a concerted mechanism, which was not necessarily to be expected considering the stabilization of a potential benzylic carbocation in a stepwise process.

To gain further insight into the reaction mechanism, we became interested in differentiating 1,1-biaryl alkenes based on their steric and/or electronic properties (Scheme 3a). Herein, only one aryl ring is in conjugation with the π-system, resulting in substantial steric differentiation of the two aryl groups (Scheme 3a, TS-A2). We propose preferential reaction occurs with the more reactive conformer of alkene (TS-A2, X = more electron donating than Y), providing an excellent setting to undertake a more detailed Hammett study. Electronically differentiated biaryl alkenes 10 were evaluated in the reaction. Moderate to good yields were obtained depending on the electronic properties of the biarylalkenes. In agreement with our model, increased disparity between the two aryl groups resulted in improved reaction enantioselectivity (products 11a–11f). We found a good correlation between log(er) and σ⁺ with a ρ value <1, suggesting the build-up of positive charge in the transition state in a less sensitive fashion than the parent S_N1 reaction.¹⁸ This correlates with a concerted, highly asynchronous cycloaddition. According to the regression equation obtained from the small training set used for the Hammett study, an enantiomeric ratio of 92:8 was predicted for alkene 12 bearing two electronically altered rings¹⁹ (herein Δσ⁺ was obtained from the parent σ⁺ values for para-CF₃ and para-OMe).¹⁸ When 12 was subjected to the reaction conditions, cycloadduct 13 was obtained in 58% yield and 92:8 er highlighting the potential of this type of enantiodiscrimination (Scheme 3b). Proof of absolute stereochemistry was achieved by hydrolysis of product 13 and X-ray analysis of the respective biarylcyclobutanecarboxylic acid 14.

Scheme 3.

Distal Differentiation of 1,1-Biarylalkenes

To further probe our hypothesis, differentially substituted 1,1-biaryl olefin 15 was synthesized and examined in the reaction (Scheme 3c). As the aryl groups of this alkene possess more similar electronic properties, the rotation of one aryl group out of conjugation is primarily driven by adverse intramolecular steric interactions. We hypothesized the ortho-tolyl group would preferentially rotate out of plane to minimize 1,3-allylic strain (TS-A3, Scheme 3c), thus providing a similar steric environment as proposed in TS-A2. Gratifyingly, 16 was obtained in 48% yield and 89:11 er, lending support to our mechanistic hypothesis.

Scope.

With a catalyst system that allowed for the cycloaddition of activated alkenes in hand, we examined the substrate scope of the reaction (Scheme 4). α-Methylstyrene underwent [2+2] cycloaddition in 93% yield and 96:4 er (product 3c) on gram scale (5.26 mmol) with no loss in reaction selectivity. Increasing the steric size of the α-substituent was investigated and proceeds with high enantioselectivity (products 3d and 3e). High chemoselectivity for the activated alkene in the presence of an unactivated alkene was observed to provide 3e in 72% yield and 96:4 er, with no trace of cycloadducts derived from the reaction of the unactivated olefin. Several steric and electronic perturbations of the aromatic ring have been investigated (products 3f–3l). The cycloaddition proceeded in good yield with sterically encumbered (product 3f), halogenated (products 3h and 3j), and electron-poor (products 3i and 3j) vinyl arenes. Spirocyclic cyclobutane derivatives can also be accessed from the requisite 1,1-disubstituted olefin in good yield and high enantioselectivity (product 3k and 3l). Heterocycles bearing weakly basic heteroatoms, such as thiophene, were also tolerated (product 3l). In some cases, dropping the temperature improved the yield by decreasing the rate of alkene polymerization. Interestingly, enynes underwent cycloaddition without any interference of the alkyne moiety yielding 3m and 3n in moderate yield and excellent enantioselectivity.

Scheme 4.

1,1-Disubstituted Alkene Substrate Scope*

Replacement of the aryl group with a cyclohexyl group resulted in substantial decrease in enantioselectivity, demonstrating that the aryl group is necessary to obtain highly enantioenriched products (Scheme 4, compare products 17 and 3c). Substitution at the α-position was also essential for successful reaction, as styrene itself performed poorly in terms of reactivity and selectivity under several reaction conditions (product 3o, see Supporting Information for further details). Partial polymerization of styrene presumably accounts for the low yield, whereas low enantioselectivities for 17 and 3o can be rationalized by lack of steric differentiation with the protruding C–H bond of the allene.

To further expand the reaction scope, we investigated commodity dienes, such as isoprene (18), in the cycloaddition reaction (Scheme 5). Initially, when EtAlCl₂ was used as a Lewis acid, low periselectivity was observed, favoring Diels–Alder product 20. In stark contrast to EtAlCl₂, catalyst 2e allowed for >99:1 selectivity favoring [2+2] cycloadduct 19. The reaction also occurred with high chemoselectivity, as only the more substituted alkene of isoprene underwent cyclo-addition; however, the observed enantioselectivity was only moderate for this reaction. We assume that upon binding of allenoate 1c to catalyst 2e, the internal π-bond (Scheme 5a, marked in gray in TS-A4) is sufficiently blocked by the large boroaryl group of the catalyst, leaving the distal π-bond (marked in red in TS-A5) more readily accessible for [2+2] cycloaddition.²⁰ It should be noted that our system complements previous reports on Diels–Alder reactions between allenoates and cyclic dienes.^14b,21,22 α-Substitution, as imposed by 2,3-dimethylbutadiene (21), significantly improved the enantioselectivity while preserving the high level of peri-selectivity (Scheme 5b, product 22).

Scheme 5.

Catalyst Induced Periselectivity

Application to β-Substitution.

The potential to generate two adjacent stereocenters piqued our curiosity to further study β-substitution of the alkene starting materials and test our proposed models. We initiated this survey with cyclic alkene 23. Gratifyingly, the reaction proceeded in good yield, regioselectivity, and with excellent enantioselectivity, but resulted in the formation of both alkene isomers (Scheme 6, Z-24 and ent-E-24) in a 69:31 ratio. Separation by column chromatography revealed Z-isomer Z-24 as the major product. In accordance with the deuteration experiment and as a consequence of β-substitution on the alkene, the two TS do not lead to the same absolute configuration within the cyclobutane ring (see Scheme 6a, TS-A6 and TS-B2). To verify this hypothesis, Z-24 and ent-E-24 were individually transformed to ketone 25 by oxidative cleavage using a modified Lemieux–Johnson oxidation.²³ As expected, 25 revealed opposite absolute configuration indicated by opposite optical rotation.

Scheme 6.

Initial Study on Trisubstituted Alkenes

Acyclic trisubstituted alkenes should, based on our mechanistic study, undergo cycloaddition in a concerted, stereospecific fashion. To demonstrate the utility of such an attribute, different pairs of acyclic E- and Z-alkenes were subjected to the optimized conditions (Table 2). As seen for alkene 23, modest E/Z selectivity was observed for Z-alkenes; however, the respective cyclobutanes 27a and 27b were formed with very high enantioselectivity (entries 1 and 2, the respective TS-A6 and TS-B2 from Scheme 6 are likely operating and explain the observed selectivity for alkenes 26a and 26b). Gratifyingly, no detectable amounts of the diastereomeric product originating from a nonstereospecific process was observed (compare entry 1 and 3). Interestingly, E-alkenes (entries 3–7) performed significantly better in terms of E/Z selectivity of products. Considering the two transition states TS-A7 and TS-B3, this is not surprising because TS-A7 encounters a substantial steric interaction between the E-β-substituent and the boroaryl group, whereas TS-B3 is less affected by this substituent pointing away from the large boroaryl group (see Scheme 7a for details). Thus, E-isomeric cyclobutanes were formed almost exclusively from E-alkenes; however, product 27c (entry 3) was obtained with low enantioselectivity even when the temperature was decreased to −20 °C. A modest improvement was accomplished by exchanging catalyst 2c with 2a to provide the desired product in 90:10 er. Interestingly, when benzyl allenoate 1a was used instead of its thio-analog 1c, good enantioselectivity was achieved while maintaining the high level of E/Z selectivity (entry 4). Steric bulk, as imposed by ethyl groups on their respective positions (alkene 26d and 26e), was well tolerated giving products 27e and 27f in good yield and enantioselectivity. Finally, cyclic alkene 26f proceeded in 88% yield and 94:6 er with thiobenzyl allenic ester 1c. Its absolute stereochemistry was unambiguously determined by X-ray diffraction of the respective pentabromophenyl ester 28 and was found to be in agreement with the proposed models (Scheme 7b).

Table 2.

Stereospecific [2+2] Cycloadditions

^aDetermination by ¹H NMR of the crude reaction mixture utilizing a calibrated standard. Reactions run under optimized conditions using catalyst 2e.

^bYields reported of pure major isomer as average of two experiments.

^cEnantiomeric ratio of the major isomer (see Supporting Information for er of minor isomers).

^dReactions run at −20 °C.

^eCombined yield of both isomers in parentheses.

Scheme 7.

TS for E-Alkenes

CONCLUSION

In summary, a method for enantioselective [2+2] cyclo-additions of activated alkenes with allenoates has been developed. Supported by mechanistic evidence, this reaction resembles a rare example of a concerted, enantioselective [2+2] cycloaddition with activated alkenes. As such, its potential to generate molecular complexity, with precise control of stereochemistry, makes the reaction especially attractive toward the synthesis of cyclobutane containing targets.