Catalyst-Controlled Annulations of Strained Cyclic Allenes with π-Allylpalladium Complexes

Abstract

Strained cyclic allenes are a class of in situ-generated fleeting intermediates that, despite being discovered more than 50 years ago, has received significantly less attention from the synthetic community compared to related strained intermediates. Examples of trapping strained cyclic allenes that involve transition metal catalysis are especially rare. We report the first annulations of highly reactive cyclic allenes with in situ-generated π-allylpalladium species. By varying the ligand employed, either of two isomeric polycyclic scaffolds can be obtained with high selectivity. The products are heterocyclic and sp³-rich and bear two or three new stereocenters. This study should encourage the further development of fragment couplings that rely on transition metal catalysis and strained cyclic allenes for the rapid assembly of complex scaffolds.

In situ-generated strained intermediates that bear a functional group with a preferred linear geometry within a small ring have emerged as valuable synthetic building blocks. Arynes^1–6 and cyclic alkynes^7,8 (e.g., 1–3, Figure 1A) have been most well-studied with applications spanning the synthesis of heterocycles,^9,10 ligands (such as XPhos),¹¹ natural products,^12–14 agrochemicals,¹⁵ and organic materials.^16,17 A related class of strained intermediates discovered around the same time as arynes and cyclic alkynes is strained cyclic allenes, such as 1,2-cyclohexadiene (4).^18,19 The cumulated diene confined to a small ring leads to ~30 kcal/mol of strain energy, making cyclic allenes well-suited for strain-promoted reactions.^20–22 However, these intermediates have received significantly less attention compared to arynes and cyclic alkynes despite possessing many attractive attributes²³ that substantiate their value as building blocks for the synthesis of complex sp³-rich scaffolds. Recent studies of strained cyclic allenes have led to advances in cyclic allene generation protocols,^24–26 cycloaddition reactions,^27–36 metal-catalyzed processes,^24,37,38 trapping in a single-electron process,³⁹ DNA-encoded library synthesis,⁴⁰ and total synthesis.⁴¹

Figure 1.

(A) Strained cyclic intermediates 1–4, (B) approaches for metal-catalyzed reactions of strained cyclic allenes, and (C) overview of current study.

An attractive yet underdeveloped approach to leveraging strained cyclic allenes in synthesis utilizes transition metal catalysis.^42,43 Only three reports of such transformations have been demonstrated in the literature.⁴⁴ In these reactions, an organometallic intermediate 5 is generated catalytically and intercepts a transient cyclic allene (e.g., 6) (Figure 1B). This gives rise to annulated product 7. In the known literature reports, only σ-bound metal species have been used, namely, palladacycle 8,^24,45 nickelacycle 9,³⁷ and arylpalladium species 10.³⁸ With the aim of expanding the types of organometallic intermediates that can be used in reactions of strained cyclic allenes, we became interested in the use of π-allylpalladium species 11. The ability of 11 to undergo C–C bond formation at either of its termini^46–48 is distinct from the reactivity of organometallic intermediates used previously in cyclic allene annulations. As will be described herein, the use of π-allylpalladium complexes^49,50 to trap cyclic allenes provides opportunities to control regiochemistry and stereoselectivity, while building diverse and sp³-rich scaffolds.

In this manuscript, we demonstrate the first use of π-allylmetal complexes to trap strained cyclic allenes (12 and 6, respectively, Figure 1C). By varying the ligand employed, either of two different polycyclic scaffolds, 13 or 14, can be formed preferentially, via the formation of multiple bonds and stereocenters. Studies toward an enantioselective variant are also reported. These studies demonstrate the merger of transition metal catalysis and strained cyclic allenes for the rapid assembly of complex, sp³-rich scaffolds.

To initiate our studies, we investigated the reaction of vinyl benzoxazinone 15 with silyl triflate 16 using palladium catalysis (Table 1). Of note, benzoxazinone 15, a known π-allylmetal precursor,^51–55 was selected for these studies as it had been demonstrated to be compatible with fluoride-mediated generation of benzyne.⁵⁶ Based on known reactivity of 15, two products could plausibly arise, tricycle 17 or tetracycle 18. To generate soluble fluoride ions, CsF and Bu₄NOTf⁵⁷ were used in combination. Furthermore, elevated temperatures were employed to facilitate efficient generation of reactive intermediates. Select key results from reaction optimization are depicted, and additional data are available in the Supporting Information (SI).

Table 1.

Select Results from Optimization Studies

^aReactions conditions: 15 (3.0 equiv), 16 (1.0 equiv), catalyst (5 mol %), ligand (as shown), CsF (10 equiv), Bu₄NOTf (5.0 equiv), additive (as shown), DMF (0.050 M), 70 °C, 2 h.

^bYield determined by an isolation experiment.

^cRatio of 17:18 and dr were determined by ¹H NMR analysis. In entries where 17 is the major product, dr = 11:1; in entries where 18 is the major product dr = >20:1.

^d[Pd] (10 mol %), 0.025 M.

The use of the dialkylbiaryl ligand DavePhos, which had previously been identified as a competent ligand in a cyclic allene annulation,³⁸ did not lead to a productive reaction (Table 1, entry 1). However, switching to the PPh₃ ligand gave tricyclic annulation product 17, albeit in a low yield of 20% (entry 2). Alternatively, the use of the bidentate ferrocene ligand dppf delivered constitutional isomer 18 as the major product (entry 3). With initial results for the selective formation of annulation product 17 or 18, we performed optimization to selectively access each of the constitutional isomers. For practical reasons, we elected to pursue the use of Pd(PPh₃)₄, which gave 17 in slightly improved yield and excellent selectivity (entry 4; compared with entry 2). Because a significant amount of decomposition of benzoxazinone 15 was observed under these conditions,⁵⁸ we explored additives to mitigate deleterious side reactions. Empirically, the addition of water led to significant improvement, generating tricycle 17 in 65% yield after 2 h (entry 5). Extensive efforts were made to selectively generate tetracycle 18. We ultimately identified that the use of Buchwald precatalyst dppf Pd G3 and water as an additive delivered 18 in 73% yield after 2 h (entry 6; compared with entry 3). Thus, either tricycle 17 or tetracycle 18 could be accessed by the judicious choice of ligand. It should be noted that the annulation of benzyne with benzoxazinone 15 only delivers the tetracyclic counterpart to 18, without formation of the tricyclic counterpart to 17.⁵⁶

Having identified suitable reaction conditions for the selective formation of either constitutional isomer 17 or 18, we examined the scope of the methodology beginning with the formation of tricyclic products 21 (Figure 2). The parent substrate, along with its N-mesyl derivative, each gave a 60% yield of tricyclic product (17 and 22, respectively). A variety of substituents on the arene (R² = Me, F, OMe, Ph, 2-thiophene) were also tolerated, as evidenced by the formation of tricycles 23–29.⁵⁹ The use of a naphthalene-derived substrate proceeded smoothly to furnish 30 in 64% yield. Additionally, we evaluated a benzoxazinone bearing an isopropenyl group, which afforded tricycle 31. Lastly, heterocyclic allene precursors (i.e., 20) could be employed under modified reaction conditions to enable further modulation of the product structure. More specifically, interception of oxa-²⁹ and aza-derivatives²⁸ of 20 delivered heterocycles 32 and 33, respectively. It should be noted that formation of the corresponding tetracyclic products was either negligible or not observed for the results shown in Figure 2. Additionally, in most cases, the observed diastereoselectivities are synthetically useful.

Figure 2.

Scope of the annulation reaction using PPh₃ as the ligand. Reactions performed on 0.05 mmol scale with 19 (3.0 equiv) and 20 (1.0 equiv) at 0.050 M. Yields reflect an average of two isolation experiments. ^a Reactions were performed with CsF (5.0 equiv) in MeCN (0.050 M) at 60 °C for 6 h, without Bu₄NOTf additive.

Subsequently, the scope of the methodology for preparing tetracycles utilizing the dppf Pd G3 catalyst was assessed (Figure 3). Although our conditions shown in Table 1, entry 6 were generally useful, some modifications were made to obtain optimal yields based on empirical observations. Beginning with the parent substrate combination, 18 was isolated in 70% yield and >20:1 dr. Variation of the N-substituent, as well as steric and electronic perturbations of the aryl ring, were tolerated as seen by the formation of 35–43 in good yields. Finally, trapping of an oxacyclic allene gave tetracycle 44. Notably, the annulations to afford tetracycles 34 proceed via the formation of three new bonds and three stereocenters with excellent diastereoselectivity (>20:1 dr in all cases).

Figure 3.

Scope of the annulation reaction using dppf as the ligand. Reactions performed on 0.05 mmol scale with 19 (3.0 equiv) and 20 (1.0 equiv) at 0.025 M. Yields reflect an average of two isolation experiments. For all isolations, dr = >20:1 and ratios of 34:21 range from 6.3:1 to >20:1; see SI for details. ^a Reaction was performed at 100 °C. ^b Reaction was performed in MeCN (0.0125 M) at 30 °C for 14 h, without Bu₄NOTf additive.

Plausible mechanisms for the transformations we have developed are depicted in Figure 4. The Pd(0) catalyst coordinates to 45. Oxidative addition of the Pd(0) catalyst into the allylic C–O bond of 46 followed by decarboxylation affords zwitterionic π-allylpalladium intermediate 12.⁵¹ Ligand-controlled migratory insertion of cyclic allene 6 (formed from fluoride-mediated 1,2-elimination of silyl triflate 20) gives π-allylpalladium intermediates 47 and 48.⁶⁰ These intermediates arise from C–C bond formation at either of the two sites of reactivity of π-allylpalladium species 12. The branched or linear selectivity of the migratory insertion is proposed to be a result of the ligands employed.⁶¹ Cyclizations of 47 and 48 afford tricycle 13⁶² and tetracycle 14, respectively, through the formation of one or two new bonds.

Figure 4.

Proposed catalytic cycle.

Given the notable impact of ligands on selectivity in these transformations, the feasibility of an enantioselective variant was investigated using vinyl benzoxazinone 49 and oxacyclic allene precursor 50. Notably, the use of two chiral racemic starting materials would represent a departure from previous studies that use an achiral trapping partner.^37,38 Select key results are shown in Table 2, with additional information being available in the SI.⁶³ The use of PHOX ligand L1 and Trost’s ligand L2, which are commonly employed in π-allylpalladium chemistry,⁵⁰ provided no desired reactivity (entries 1 and 2, respectively). However, using phosphoramidite ligand L3 (entry 3), annulation product 51 was formed as the major isomer, thus demonstrating the feasibility of rendering the annulation to form both products enantioselective. Given the greater structural complexity of isomer (−)-52, we elected to assess other chiral diphosphine ligands in pursuit of (−)-52. The use of Phanephos L4, Segphos ligand L5, and Josiphos ligand L6 each gave the desired product (−)-52 in varying yields and enantioselectivities (entries 4, 5, and 6, respectively). The employment of Walphos ligand L7 and Mandyphos ligand L8 each resulted in increased yields and enantioselectivities (entries 7 and 8, respectively). Of these, use of L8 provided superior results and delivered (−)-52 in 54% yield and 70% ee.⁶⁴ We surmise that the absolute configuration of (−)-52 is as depicted, on the basis of an X-ray crystal structure obtained for carbocyclic derivative (−)-43. Although improving the enantioselectivity was challenging in our initial efforts,⁶⁵ the results shown provide the most structurally complex products accessible from the merger of strained cyclic allenes and asymmetric catalysis to date. As enantioselective transformations of fleeting strained cyclic intermediates remain exceedingly rare, we hope these results will promote further efforts in this area.

Table 2.

Evaluation of Select Chiral Ligands in the Annulation Reaction

^aReaction conditions: 49 (3.0 equiv), 50 (1.0 equiv), Pd₂(dba)₃ (5 mol%), ligand (20 mol%), CsF (10 equiv), H₂O (9.0 equiv), MeCN (0.0125 M), 30 °C, 14 h.

^bYield determined by an isolation experiment.

^cSee SI, section F for details and crystallographic data.

We have developed the first strategy for engaging strained cyclic allenes with π-allyl metal species. These are rare examples of cyclic allene reactions in transition metal-catalyzed processes. By the judicious choice of ligands, selectivity can be controlled, leading to the preferential formation of tri- or tetracyclic heterocycles with significant structural complexity. Of note, the transformations proceed by formation of either two or three new bonds and two or three new stereocenters. We also demonstrate the feasibility of an enantioselective variant, thus providing a crucial proof-of-principle study for further asymmetric reaction development. Overall, the union of strained cyclic allenes and π-allylpalladium chemistry expands the types of organometallic intermediates that can be used in cyclic allene reactions to access complex sp³-rich scaffolds. We hope this study promotes the further development and understanding of transition metal-catalyzed reactions of strained cyclic allenes.