Bifunctional Phosphine-Enabled Regioselective Cycloisomerization of Enynyl Esters En Route to Bicyclo[2.2.1]heptenes

Abstract

Expedient access to bridged enol benzoates is realized via a tandem gold-catalyzed cycloisomerization of enynyl esters and the Diels–Alder reaction. The gold catalysis permits the use of enynyl substrates without additional propargylic substitution and achieves the highly regioselective formation of less stable cyclopentadienyl esters. The regioselectivity is enabled by a bifunctional phosphine ligand, the remote aniline group of which facilitates α-deprotonation of a gold carbene intermediate. The reaction works with different alkene substitution patterns and various dienophiles.

Graphical Abstract

Over the past several years, we have exploited metal–ligand cooperation¹ in homogeneous gold catalysis by employing biaryl-2-ylphosphine ligands such as WangPhos, L1, and L2 featuring a Lewis/Brønsted basic functional group at the bottom half of the nonphosphine aryl ring (Scheme 1A).^2,3 The remote basic group can engage in accelerative interactions with nucleophiles via H-bonds⁴ or permit challenging deprotonation.⁵ In the latter scenario, we have so far demonstrated that with either an aniline⁵ or a tertiary amino group⁶ in this type of ligand, exceedingly mild and catalytic propargylic deprotonation, which otherwise would require NaNH₂ or ⁿBuLi (Scheme 1B), can be realized, allowing isomerization of 1-aryl-1-alkynes into 1-aryl-1,3-dienes⁵ and the asymmetric cycloisomerization of propargylic alcohols into 2,5-dihydrofurans as well as inter- and intramolecular propargylations.² To further exploit this cooperative gold catalysis, we surmised that these aniline-/amino-functionalized ligands could facilitate regioselective deprotonation due to their intramolecular nature and the relatively rigid ligand framework. Herein, we disclose that an aniline-based ligand indeed permits deprotonation of in situ generated gold carbenes with complete regiochemistry control, eventually affording bicyclo[2.2.1]heptenes upon subsequent D-A reactions.

Scheme 1.

Bifunctional Phosphine Ligands and Application in Cooperative Gold Catalysis

In 2008, we reported the gold(I)-catalyzed cycloisomerization of enynyl ester 1 to cyclopentenone 2 in wet DCM (Scheme 1C).⁷ Subsequent DFT studies⁸ supported a mechanism involving gold-catalyzed 3,3-sigmatropic rearrangement of the substrate propargylic ester moiety, a goldpromoted Nazarov cyclization to form the cyclic vinylgold carbene species A, subsequent water-facilitated, rate-determining carbene α-deprotonation to form the cyclopentadienyl ester 3, and the hydrolysis of 3 to deliver the final product. Though the scope and a variety of perturbations of the reactions⁹ have been documented, there is no report of cases with R³ being H. One anticipated issue is the formation of the more stable cyclopentadienyl ester isomer 3′ stemming from competitive γ-deprotonation of A or 1,5-hydride migration of 3 and, upon hydrolysis, the isomeric cyclopentenone 2′. We surmised that by using our designed bifunctional ligands featuring a remote amino group, α-selective deprotonation of A with R³ = H (i.e., A-H) could be achieved upon strategic positioning of a basic group in close proximity to the α-hydrogens (Scheme 1D); moreover, such an intramolecular deprotonation could be facile, thereby leading to accelerated. As such, the less stable cyclopentadienyl ester 3-H would be formed selectively and efficiently, and the substrate scope of this valuable gold catalysis could be substantially expanded.

We began by subjecting the enynyl benzoate 1a to gold catalysis with different ligands. As shown in Table 1, the reaction conditions were toluene as the solvent, NaBARF as the chloride abstractor, ambient temperature, and 3 h. Except for the aniline ligand L1 (entry 1) and WangPhos (entry 2), all the other ligands, including JohnPhos (entry 3) and the tertiary amine ligand L2 (entry 4) performed poorly in terms of yields of 2a and 3a′ based on conversion. In the case of the prototypical ligand Ph₃P (entry 6), no desired reaction was detected. In comparison to WangPhos, L1 led to a much faster reaction, and 1a was completely consumed in 3 h. This rate acceleration phenomenon is consistent with the hypothesis that the ligand aniline group may facilitate the rate-determining deprotonation in gold catalysis. Even though no water was purposely introduced to the reaction, the residual water permitted the hydrolytic formation of the more stable cyclopentenone product 2a′, which should be formed at least partly from the corresponding isomeric cyclopentadienyl benzoate 3a′, even though it was not detected. Our attempt to access the less substituted cyclopentenone 2a¹⁰ by hydrolyzing purified 3a under both acidic and basic conditions was foiled by the formation of the more stable 2a′ and the instability of 2a.

Table 1.

Initial Ligand Screening^a

entry	ligand	conversion	% yield of 3a	% yield of 2a′
1	L1	100	46	54
2	WangPhos	53	20	26
3	JohnPhos	60	7	16
4	L2	50	3	7
5	IPr	48	1	1
6	PPh3	16	0	0

^aReactions were run for 3 h at room temperature and then quenched with ⁿBu₄NCl. Yields were calculated via ¹H NMR using diethyl phthalate as the internal standard.

Since it was unclear whether L1 dictates the complete regioselective formation of 3a over its regioisomer 3a′ by the above results, we then decided to trap the cyclopentadienyl benzoate product by N-phenylmaleimide in a tandem Diels–Alder reaction. The results are listed in Table 2. Among the five ligands examined, only L1 resulted in complete selectivity for the formation of cycloadduct 4a over 4a′, which is the D–A product of the previously unobserved 3a′. Moreover, L1 led to an excellent yield (89%). Toluene was found to be a slightly better solvent (entry 1 vs entries 6 and 7), and the addition of drierite and the use of excess N-phenylmaleimide led to nearly quantitative yields of the tetracyclic imide adduct 4a (entries 11 and 12). These results confirmed that the aniline ligand L1 indeed achieves a highly regioselective deprotonation of the carbene intermediate of type A, while JohnPhos (entry 2) lacking the remote basic aniline group led to poor regioselectivity and substantial degradation to the 2a′ enone side product. Of note is the expedient access to the tetracyclic bridged structure of 4a with excellent diastereoselectivity. Its relative stereochemistry was established by 2D NMR, and the face and endo selectivity is consistent with literature reports.¹¹

Table 2.

Reaction Conditions Optimization with N-Phenylmaleimide (5a)^a

entry	ligand	solvent	drying gent	% yield of 4a/4a′/2a′b
1	L1	toluene		89/–/–
2	JohnPhos	toluene		34/16/36
3c	WangPhos	toluene		46/12/5
4	L2	toluene		24/15/–
5	PPh3	toluene		–/–/–
6	L1	DCM		85/–/4
7	L1	PhCF3		83/–/–
8e	L1	toluene		9/–/–d
9	L1	toluene	drierite	91/–/–
10	L1	toluene	3 Å MS	84/–/–
11	L1	toluene	drierite	97/–/–f
12	L1	toluene	drierite	97/–/–g

^aYields were calculated via ¹H NMR using diethyl phthalate as the internal standard. Reactions were run for 16 h at room temperature and then quenched with ⁿBu₄NCl.

^bRegioisomers were determined by an OBz aromatic signal at 8.09 ppm.

^c88% conversion.

^dMostly starting material left.

^eAgNTf₂ (5 mol %) instead of NaBARF (10 mol %) used as the halide abstractor.

^f2 equiv of 5a used.

^g3 equiv of 5a used.

With the optimized conditions shown in Table 2, entry 11, in hand, we set out to investigate the reaction scope. We first explored various enynyl benzoate substrates, and the results are shown in Figure 1. Substrates possessing 5- to 7-membered cyclic alkenes were readily allowed, affording 4a–4g in mostly excellent yields. Among the cyclohexene cases, substituents on the ring are allowed, resulting in high product yields and mostly moderate diastereoselectivities (4d–4g). In these cases, the same or similar diastereomeric ratios are observed in the crude NMR of the cyclopentadienyl benzoate intermediates, thus revealing that the D–A reaction is highly face and endo selective. In the cases of 4d–4f, the reactions are much slower, which may be attributed to the inductive deactivation caused by the oxygen-based substituents. In addition, the reaction tolerated O- (4h) and N-heterocycles (4i) and accommodated substrates featuring acyclic alkyl-substituted alkenes, affording 4j–4m in excellent yields. However, with R² = Ph, the reaction was low-yielding and 4n was contaminated with inseparable impurities. In all of these cases, we did not detect the formation of the corresponding regioisomeric products (4′). The D–A reactions were highly endo-selective, as confirmed by the NOESY experiments.

Figure 1.

Scope of enynyl benzoate substrate.

Next, we examined various dienophiles for the reaction, and the results are shown in Figure 2. For highly reactive dienophiles such as dimethyl acetylenedicarboxylate and 6,7-dibromo-1,4-naphthoquinone, the gold catalysis and the D–A reaction were performed in a single step, and the adducts 4o and 4p were formed in good yields. Of note is 4p, the ¹³C spectrum of which revealed its diketone form instead of the anticipated hydroquinone structure. On the other hand, less reactive dienophiles such as methyl acrylate and methyl vinyl ketone required a two-pot sequence, i.e., gold catalysis followed by Me₂AlCl-promoted D–A with crude cyclopentadienyl benzoate 4a at −40 °C, giving products 4q and 4r in good overall yield. The D–A reaction with MVK in the case of 4r also afforded its minor regioisomer, while the reaction of methyl acrylate led to highly regioselective formation of 4q (>30:1). For dienophiles of intermediate reactivities such as diisopropyl azodicarboxylate (4s) and nitrosobenzene (4t), a one-pot, two-step operation was required, where the gold-catalyzed step was allowed to complete prior to the addition of the dieneophile. It was observed that slightly improved yields could be achieved by shielding the reaction from light.

Figure 2.

Scope of dienophiles. Reaction conditions: L1AuCl (5 mol %), NaBARF (10 mol %), CaSO4 (10 eq), toluene, rt, 16 h. ^aThe dieneophile was preactivated with Me₂AlCl in DCM at −40 °C and then treated with the crude product mixture from the first step dropwise. ^bThe addition of a dieneophile followed the completion of the gold catalysis.

Finally, to demonstrate the synthetic utility of this chemistry, we carried out a gram-scale synthesis of 4a and achieved 97% isolated yield with a catalyst loading of only 1 mol % (Scheme 2). Similarly, we scaled up the synthesis of 4t. Due to the difficulty in separating the product from nitrosobenzene, the isolated yield was moderate (Scheme 2). The fused tricycle of 4t readily underwent hydrogenolysis/hydrogenation with H₂ and Pd/C to cleave the N–O bond and reduce the enol ester to deliver the richly functionalized 5,6-fused carbocycle 6 in 87% yield.

Scheme 2.

Gram-Scale Synthesis and Synthetic Transformation

In conclusion, we have developed an efficient route to obtaining bridged polycyclic enol esters in excellent yields from readily accessible enynyl ester substrates. Enynyl benzoates without additional substitution at the propargylic position can be converted to the less stable cyclopentadienyl esters with excellent regioselectivity, which is enabled by a biphenyl-2-ylphosphine ligand featuring a remote, weakly basic dimethyl aniline moiety. This gold-ligand cooperative catalysis tolerates a variety of substitution patterns, and the cyclopentadienyl benzoate products were captured by a range of dienophiles in subsequent D–A reactions, thereby allowing a rapid increase of structural complexity and providing an expedient access to bicyclo[2.2.1]cycloheptenes.

Data Availability Statement

The data underlying this study are available in the published article and its Supporting Information.