Switchable selectivity in Pd-catalyzed [3 + 2] annulations of γ-oxy-2-cycloalkenones with 3-oxoglutarates: C–C/C–C vs C–C/O–C bond formation

Introduction

The development of new strategies for the synthesis of complex carbocyclic and heterocyclic structures remains a general topic for the synthetic chemists [1]. In the past decades, palladium chemistry has gained an important place in the toolbox of chemists, and its use became a privileged strategy for the selective formation of carbon–carbon and carbon–heteroatom bonds [2-5]. Among the different types of palladium-catalyzed transformations, domino – alias cascade – transformations [6-11] as well as annulation reactions [12-15] occupy a special place as they facilitate the synthesis of a variety of complex cycles in a single synthetic operation, through sequential and mechanistically independent bond-forming steps.

In the frame of our long-term project dedicated to domino sequences [16-22] and Pd-catalyzed transformations [23-30], we recently reported a Pd(0)-catalyzed synthesis of bi- and tricyclic structures incorporating pyrrolidone rings (Scheme 1) [31]. In this transformation, resonance-stabilized acetamides and cyclic α,β-unsaturated-γ-oxicarbonyl derivatives are used as bis-nucleophile and bis-electrophile partners, respectively.

This process involves an intermolecular Pd(0)-catalyzed C-allylation (Tsuji–Trost reaction)/intramolecular nitrogen 1,4-addition sequence (Scheme 1, top reaction). The success of this bis-nucleophile/bis-electrophile [3 + 2] C–C/N–C bond-forming annulation is due to the well-defined chronology of the steps and the total chemoselectivity of the initial step (C-allylation). Another non-trivial feature of this process is that the possible undesired stoichiometric intermolecular 1,4-addition step (i.e., the potential initial addition of the nucleophile to the C-β of the bis-electrophile) has to be slower than the intermolecular addition of the nucleophile to the catalytically generated η³-allylpalladium complex, or it has to be at least a reversible process [32]. Furthermore, when the newly formed annulation product contains an ortho-haloaryl moiety at the nitrogen substituent, an additional intramolecular keto α-arylation step can be involved in the cascade, thereby forming two new cycles and three new bonds in the same synthetic operation (Scheme 1, bottom reaction).

We next decided to extend the scope of this strategy to dialkyl-3-oxoglutarates I as the bis-nucleophile partners [33] in the reaction with cyclic α,β-unsaturated-γ-oxycarbonyl derivatives II as the bis-electrophiles (Scheme 2). Interestingly, this new bis-nucleophile/bis-electrophile combination may allow direct access to either fused bicyclic cyclopentanic (pentalene-, indene-, or azulene-type) structures via a γ-C-allylation/β-C-1,4 addition process, or annulated furan-based motifs through a γ-C-allylation/β-O-1,4 addition process, both motifs being incorporated into biologically relevant pharmaceuticals and/or natural products (Figure 1) [34-38]. Indeed, we anticipated that the latter step might occur through either C-addition or O-addition (Scheme 2, products III (or V) vs IV) [32,39-42]. Therefore, an inherent challenge associated to this ambident nucleophile is chemoselectivity control.

Herein we disclose two chemodivergent [3 + 2] annulation reactions taking place between dialkyl-3-oxoglutarates I and α,β-unsaturated-γ-oxycarbonyl derivatives II that differ simply by the reaction temperature adopted. These methods allow the exclusive [3 + 2] C–C/C–C or C–C/O–C annulation at will, thus providing an easy access to annulated cyclopentanic structures III or annulated furan-based motifs IV, respectively. Additionally, in the case of C–C/C–C bond forming annulations, the products III can undergo two decarboxylation steps leading to bis-cycloalkanone derivatives V.

Results and Discussion

Optimization

We started our investigation using dimethyl 3-oxoglutarate (1a) and 2-cyclohexenone 4-benzoate (2a) [43,44] as bis-nucleophile and bis-electrophile model substrates, respectively (Table 1). In line with our previously developed standard reaction conditions for [3 + 2] C–C/N–C bond-forming annulation [31], the first tests were performed with the following two catalytic systems: [Pd(η³-C₃H₅)Cl]₂ (5 mol %), dppf (15 mol %)], and [Pd(OAc)₂ (10 mol %), dppb (15 mol %)] in THF at room temperature (Table 1, entries 1 and 2). These conditions promptly (ca. 1 hour) generated annulated product 4a arising from C-allylation (C–C bond forming)/intramolecular O-Michael addition (O–C bond forming) sequence. The purification of the crude reaction was easier using the Pd(OAc)₂ and dppb system (Table 1, entry 2), which was thus chosen to continue the optimization. The effect of the solvent was then assessed. While no reaction took place in CH₃CN (Table 1, entry 3), DMF (Table 1, entry 4) and DMSO (Table 1, entry 5) allowed the formation of 4a in 66% and 75% yield, respectively. The use of another bidentate phosphine such as dppe gave no improvement (Table 1, entry 6), showing only traces of 4a. The influence of the temperature in DMSO was examined next. The reaction performed at 75 °C gave a similar yield for compound 4a (compare entries 5 and 7 in Table 1), whereas the reaction carried out at 100 °C and 130 °C during 6 hours generated solely bicyclo[4.3.0]nonane-3,8-dione (5a) with 50% and 69% yield, respectively (Table 1, entries 8 and 9). This product plausibly arises from a C-allylation step followed by an intramolecular C-conjugate addition/decarboxylation sequence, although throughout this study the putative intermediate 3a proved always elusive. Using DMF and DMA as solvents at 130 °C (Table 1, entries 10 and 11), and [Pd(η³-C₃H₅)Cl]₂, dppf] as the catalytic system in DMSO (Table 1, entry 12) did not allow further improvements for the formation of compound 5a. Furthermore, after 1 h at 130 °C under microwave irradiation, the desired compound 5a was isolated in 69% yield (Table 1, entry 13). Finally, separated control experiments carried out by omitting the Pd source or the ligand resulted in the exclusive recovery of the starting materials, which confirmed the need of the catalytic system for the catalytic process (Table 1, entries 14–16).

The above work allowed obtaining the optimal reaction conditions for the generation of the furocycloalkanone 4a [Pd(OAc)₂ (10 mol %), dppb (15 mol %) in DMSO, 1 h at rt (conditions A)] as well as for the formation of the bicyclo[4.3.0]nonane-3,8-dione (5a) [Pd(OAc)₂ (10 mol %), dppb (15 mol %) in DMSO at 130 °C, 6 h (conditions B), or 1 h under microwave irradiation (conditions C)].

Scope

The scope of the C–C/O–C [3 + 2] annulation between dimethyl 3-oxoglutarate (1a) and six- (2a), five- (2b) and seven-membered (2c) cyclic α,β-unsaturated-γ-oxycarbonyls was next studied (Scheme 3). Under the optimized conditions A, at room temperature in DMSO, the six- (2a) as well as the seven-membered (2c) bis-electrophiles reacted smoothly giving the furocycloalkanones 4a and 4c in good yields (Scheme 3). Furthermore, the protocol could be scaled up to 1 mmol without significant yield erosion. Treatment of cyclopentenone 4-benzoate (2b) under conditions A did not allow the formation of the corresponding C–C/O–C annulated product. Instead, elimination of the benzoate anion, very likely from the transiently formed η³-allylpalladium complex, gave cyclopentadienone, which underwent the known self-Diels–Alder cycloaddition to form dimer 6 [45].

We next turned our attention to the [3 + 2] C–C/C–C annulation by using the conditions B and C in DMSO at 130 °C (Scheme 4). The 2-cyclohexenone 4-benzoate (2a) afforded the expected bicyclo[4.3.0]nonane-3,8-dione (5a) in 69% yield (64% from 1.0 mmol of 1a) under either thermal conditions or microwave irradiation. Surprisingly, application of these protocols to the seven-membered bis-electrophile 2c led to the corresponding bicyclo[5.3.0]decane-3,9-dione (5c) with a low yield of 24% under thermal conditions B, while microwave irradiation was ineffective. Moderate yields of bicyclo[3.3.0]octane-3,7-dione (5b) [46] were obtained from five-membered bis-electrophile 2b under both conditions. This result suggests that the undesired Pd-catalyzed elimination from a γ-acyloxycyclopentenone (to give cyclopentadienone which in turn promptly dimerizes; see above), can be, at least in part, alleviated by performing the reaction at high temperature.

However, in the specific case of this latter [bis-nucleophile/bis-electrophile] couple, the desired bicyclo[3.3.0]octane-3,7-dione structure can be better obtained under basic conditions (and in the absence of Pd catalysis), as reported by Winterfeldt and Osterthun in the seventies [47].

The reaction between the 1,3-activated 1,3-propanones bis-nucleophiles other than 1a was then considered in the reaction with 2-cyclohexenone 4-benzoate (2a, Scheme 5). Thus, under conditions A, diacetylacetone 1b afforded the expected product in 43% yield as a 1:9 keto/enol (7:7’) mixture. Passing from DMSO to THF (conditions A’) led to a slight increase of the yield (53%) and eased the work-up. Similarly, under conditions A', moderate to good yields of the annulated products were also obtained with dibenzoylacetone 1c (products 8:8’) and 1,3-bis-(benzenesulfonyl)propan-2-one (1d, product 9), respectively. On the other hand, despite several trials using various conditions (temperature, time, and solvent) attempted conversion of the two triketones 1b,c or the ketodisulfone 1d into the corresponding 1,3-disubstitued bis-cycloalkanones 3 met always with failure.

Mechanistic studies

Considering that the change of reaction outcome of the dimethyl 3-oxoglutarate (1a) only depends on the variation of temperature in DMSO, we surmised that in the above annulation the C–C/O–C adducts 4 are the kinetic products, while the C–C/C–C adducts 5 are the thermodynamic ones. Indeed, heating compound 4a in DMSO at 130 °C gave compound 5a (Scheme 6). This result strongly supports the hypothesis that 4a is the kinetic C–C/O–C annulated product, which, upon heating, rearranges to the thermodynamic C–C/C–C [3 + 2] annulated compound before undergoing two decarboxylation to give 5a. This type of thermal 1,3-oxygen-to-carbon rearrangement was already described by Trost in the early 80’s [48,49]. In view of the high temperature needed (130 °C for several hours or under microwave irradiation), this decarboxylative rearrangement appears to require a rather high activation barrier. This activation barrier might be even greater in the case of the triketones 1b,c and 1,3-bis(benzenesulfonyl)propan-2-one (1d), which would explain the impossibility to access the corresponding 1,3-disubstitued bis-cycloalkanone derivatives.

With these considerations in mind, a plausible mechanism for this annulation is proposed for the reaction between dimethyl 3-oxoglutarate (1a) and 2-cyclohexenone 4-benzoate (2a). The reaction starts with an oxidative addition of the bis-electrophile 2a onto the Pd(0) complex to generate η³-allyl complex B from the transient η²-alkene complex A (steps a and b). Deprotonation of the pro-nucleophile 1a by the counter-anion of the η³-allyl-Pd complex exchanges the benzoate for the enolate anion (step c) [50], and following C–C bond formation from the resulting anion-scrambled complex C leads to the Pd(0) complex D (step d). Pd(0) decoordination closes the catalytic cycle delivering intermediate E_(keto) (step e) (Scheme 7).

The following spontaneous intramolecular O-conjugate addition of one of the two possible enol tautomers of E_(keto) affords the kinetic C–C/O–C adduct 4a through steps (f, g), or (h, i, j) (Scheme 8, top and middle lines) [51]. At room temperature and standard reaction times, the reaction stops at this level. However, at 130 °C the reversibility of the sequence leading to 4a becomes important and the system has enough energy to rapidly undergo the irreversible intramolecular C-conjugate addition of enol E_{(enol 2)} followed by double decarboxylation to give the final bicyclo[4.3.0]nonane-3,8-dione (5a, steps k, l, Scheme 8, bottom line).

Conclusion

In conclusion, we have successfully developed two totally selective and chemodivergent methods for the palladium-catalyzed [3 + 2] annulation between dialkyl 3-oxoglutarates and cyclic α,β-unsaturated-γ-oxycarbonyl derivatives differing simply on the variation of the reaction temperature. These new domino transformations allow a switchable (C–C/O–C to C–C/C–C) [3 + 2] annulation through an [intermolecular Pd-catalyzed C-allylation/intramolecular (oxygen or carbon) 1,4-conjugate addition] sequence. In particular, the conjugate addition becomes reversible if the temperature is increased, allowing to pass from the O-alkylation to C-alkylation product. Overall, the success of these [3 + 2] annulations is due to the total chemoselectivity of the initial step (C-allylation) as well as to the well-defined chronology of the following steps. Further work is currently ongoing to develop enantioselective versions of these new transformations.

Experimental

Conditions A – [3 + 2] C–C/O–C bond-forming annulations. In a Schlenk tube, under argon atmosphere, were added Pd(OAc)₂ (0.10 equiv), dppb (0.15 equiv) and anhydrous DMSO (0.1 M). After 10 minutes stirring, the cyclic electrophile 2a–c (1.3 equiv) and dimethyl 3-oxoglutarate (1a, 1.0 equiv) were added, and the reaction was stirred at room temperature. After 1 hour stirring, the reaction mixture was filtered on a plug of silica and washed with EtOAc. The filtrate was washed with a 10% aqueous solution of NaHCO₃. The aqueous phase was extracted with EtOAc, and the combined organic phases were washed with brine, dried over anhydrous MgSO₄, filtered and concentrated under reduced pressure. Purification by column chromatography on silica gel afforded the corresponding product 4a or 4c (or compound 6 in the case of 2b).

Conditions B – [3 + 2] C–C/C–C bond-forming annulations. In a sealed tube, under argon atmosphere, were added Pd(OAc)₂ (0.10 equiv), dppb (0.15 equiv) and anhydrous DMSO (0.1 M). After 10 min, the cyclic electrophile 2a–c (1.3 equiv) and dimethyl 3-oxoglutarate (1a, 1.0 equiv) were added, and the reaction was stirred at 130 °C. After 6–8 hours stirring, the reaction mixture was filtered on a plug of silica and washed with EtOAc. The filtrate was washed with a 10% aqueous solution of NaHCO₃. The aqueous phase was extracted with EtOAc, and the combined organic phases were washed with brine, dried over anhydrous MgSO₄, filtered and concentrated under reduced pressure. Purification by column chromatography on silica gel afforded the corresponding product 5a–c.