Enantiodivergent Pd-catalyzed C–C bond formation enabled through ligand parameterization

Abstract

Despite the enormous potential for the use of stereospecific cross-coupling reactions to rationally manipulate the three-dimensional structure of organic molecules, the factors that control the transfer of stereochemistry in these reactions remain poorly understood. Herein we report a mechanistic and synthetic investigation into the use of enantioenriched alkylboron nucleophiles in stereospecific Pd-catalyzed Suzuki cross-coupling reactions. By developing a suite of molecular descriptors of phosphine ligands, we could apply predictive statistical models to select or design distinct ligands that respectively promoted stereoinvertive and stereoretentive cross-coupling reactions. Stereodefined branched structures were thereby accessed through the predictable manipulation of absolute stereochemistry, and a general model for the mechanism of alkylboron transmetallation was proposed.

Palladium-catalyzed cross-coupling reactions have revolutionized the construction of C(sp²)–C(sp²) bonds. Among these cross-coupling processes, the Suzuki–Miyaura reaction has found particularly broad application due to its extensive reaction scope, as well as the stability, availability, and low toxicity of organoboron reagents (1). The 2010 Nobel Prize in chemistry was awarded, in part, to recognize the transformative impact of the Suzuki cross-coupling reaction on chemical synthesis. However, although C(sp²)–C(sp²) bond construction is now considered routine using the Suzuki reaction, extension of this process to the formation of C(sp³)–C(sp²) bonds using alkylboron nucleophiles remains a significant challenge. Of particular interest, a variant using secondary alkylboron nucleophiles with predictable and controllable stereospecificity would establish a powerful synthetic strategy to access molecular geometries with precise three-dimensional control, expanding the exceptional capabilities of the Suzuki reaction (Figure 1A).

Figure 1:

A) Enantiodivergent Suzuki reactions of secondary alkylboron nucleophiles. B) General mechanism. C) Initial investigation of substrate and ligand influences on stereoselectivity. +% ee = net retention, −% ee = net inversion.

Many efforts have focused on the use of enantioenriched secondary alkylboron nucleophiles in Suzuki cross-coupling reactions (2–4). Significant limitations remain due to slow transmetallation of the highly covalent and sterically congested C(sp³)–B bond in these reagents, as well as the propensity of the resulting Pd-alkyl species to undergo β-hydride elimination/reinsertion sequences, which can result in isomerization of the alkyl group and racemization of the stereocenter. To circumvent prohibitively slow transmetallation, as well as competing β-hydride elimination/reinsertion pathways, most stereospecific Suzuki reactions have required the use of secondary alkylboron nucleophiles that are electronically activated via inclusion of a C(sp²) α-carbon, an a-heteroatom, and/or a strongly coordinating β-carbonyl group (5–17). In addition, alkylboron nucleophiles can undergo transmetallation via either stereoretentive or stereoinvertive pathways depending on the nature of the substrate, catalyst, and/or reaction conditions. In many cases, the factors controlling the dominant mechanism of transmetallation are not understood (Figure 1B). Thus, a predictive stereochemical model for transmetallation of alkylboron reagents remains elusive.

Recently, we reported a stereospecific Pd-catalyzed cross-coupling reaction using unactivated secondary alkylboron nucleophiles (18). With P^tBu₃ as a supporting ligand, enantioenriched arylation products were obtained with transmetallation proceeding primarily via a stereoinvertive mechanism. Whereas several enlightening mechanistic studies have recently been conducted on the transmetallation of arylboron nucleophiles (19–23), these studies have not addressed the transmetallation of alkylboron nucleophiles in C(sp³)–C(sp²) bond-forming processes (24–25). Thus, unactivated alkylboron nucleophiles constitute an attractive starting point from which to investigate the reaction parameters most influential to the mechanism of alkylboron transmetallation. This mechanistic work should simultaneously facilitate the development of new synthetic methods to rationally incorporate/manipulate stereocenters via cross-coupling strategies. To this end, we report a study using predictive statistical models (26–27) to relate phosphine ligand properties to stereochemical outcomes obtained from Pd-catalyzed Suzuki reactions of unactivated enantioenriched secondary alkylboron nucleophiles and aryl electrophiles. With statistical models that rely on a next generation set of molecular descriptors, we achieved a stereoretentive Pd-catalyzed cross-coupling reaction of such nucleophiles. Furthermore, we have identified an improved ligand for the stereoinvertive variant, thus enabling an entirely ligand-controlled enantiodivergent process from a single-enantiomer organoboron nucleophile (28). Our statistical models also provide compelling evidence that each transmetallation pathway is intimately tied to specific electronic properties of the supporting ligand, which serves as a predictive guide to the mechanism of alkylboron transmetallation to palladium.

Initial investigations using electronically differentiated aryl chlorides with enantioenriched ^sBuBF₃K revealed a trend correlating diminished stereofidelity with the use of more electron-deficient coupling partners (Figure 1C, I). This observation suggested that subtle electronic effects could influence the mechanism of transmetallation and the resulting stereochemical outcome. Additionally, when the phosphine ligand was varied in an initial screen with a common aryl chloride electrophile, a considerable change in the reaction outcome from stereoinvertion to stereoretention was found (Figure 1C, II). No obvious correlation was observed between these results and the steric properties (solid angle) of the ligand. Taken together, these outcomes were difficult to interpret and inspired the use of ligand parameterization tools to provide a platform for both predictive ligand performance and mechanistic interrogation.

An expanded inventory of common phosphines with varied properties was evaluated in the Suzuki reaction of enantioenriched ^sBuBF₃K and ethyl 4-chlorobenzoate. This dataset was then subjected to correlation analysis of phosphine structural features with the stereochemical outcomes as well as the ratios of branched:linear products in these reactions. We devised a workflow and universal parameter set to describe the catalyst properties from the phosphine itself (29–32). The workflow was initiated by performing a molecular mechanics (MM) conformational search to reveal representative low energy conformers. (Figure 2A). Next, geometry optimization of the conformers using DFT was followed by parameter collection. Subsequently, four descriptor subsets were defined to capture the conformational dynamics of the ligands by including the mathematical extreme descriptor values (minimum and maximum), the lowest energy conformer values, and the Boltzmann weighted averages. We viewed the unique treatment of representative conformers as a crucial means of describing ensemble properties such as chemical shift while also probing structural flexibility during catalysis.

Figure 2.

Phosphine parameterization. A) Workflow of parameter generation and statistical modeling. B) Application of phosphine parameterization to ligand optimization of the reaction shown in Figure 1C with Z = CO₂Et. b:l = branched to linear ratio. +% ee = net retention, −% ee = net inversion.

The final step in the workflow involved the analysis of both the stereofidelity and the branched:linear product ratio. These two readouts presumably describe two stages of the reaction mechanism (Figure 1B): i) the competing stereoretentive and stereoinvertive transmetallation mechanisms that determine the final stereochemistry of the cross-coupling product and ii) the competitive β-hydride elimination/isomerization sequences that follow transmetallation. A correlation of the branched:linear ratio with the final enantiopurity of the product reveals that β-hydride elimination is responsible for both racemization and isomerization to the linear side product. Furthermore, a modest trend is observed relating the minimum width B₁ of the phosphine ligand to the branched:linear ratio (Figure 2B). This is consistent with reports of large ligands facilitating reductive elimination over β-hydride elimination (33), and suggests the use of a parameterization approach to take into account the conformational flexibility of ligands.

Since the inherent selectivity of the transmetallation mechanism is masked by deleterious racemization as a consequence of β-hydride elimination, only ligands providing high selectivity were further investigated (>30% ee, Figure 2B). The molecular electrostatic potential minimum in the phosphorus lone pair region (V_min) has been shown to correlate with the classical Tolman electronic parameter (34). Thus, V_min serves as an easily computable measure for the overall ligand electronics. A correlation between enantioselectivity and V_min was observed within the abridged dataset indicating that electronic properties of the ligand determine the mechanism of transmetallation. Specifically, electron-rich trialkylphosphines promoted stereoinvertive reactions, whereas the electron-poorer triarylphosphines provided modest selectivity for stereoretention. Use of the bulky, electron-rich ligand PAd₃ (9), which was recently reported by Carrow (35), resulted in a particularly large preference for the stereoinvertive outcome. Based on these data, we hypothesized and virtually evaluated ligands for improved stereoretentive outcomes with the following features: i) large ligand bulk to prevent β-hydride elimination and racemization, and ii) electron-deficient aryl substituents at phosphorus to promote the stereoretentive mechanism and to accelerate reductive elimination (Figure 2B). Among the proposed ligands was a set of biaryl phosphines (11-15), as pioneered by Buchwald (36), featuring various electron-deficient aryl groups at phosphorus. Gratifyingly, ligands 11 and 14 promote the alkyl Suzuki cross-coupling reaction with significantly enhanced selectivity (up to 90% ee) and minimal alkyl isomerization. Thus, parameterization-driven optimization facilitated development of a stereoretentive Suzuki reaction involving unactivated alkylboron nucleophiles. When considered alongside the introduction of 9 to achieve stereoinvertive couplings, complete control of the absolute sense of enantioselectivity (retention or inversion) can be engendered by simply selecting the appropriate ligand.

Our stereochemical investigations of secondary alkylboron transmetallation in the Suzuki reaction suggested that both enantiomers of a cross-coupling product could be selectively accessed through use of a single enantioenriched alkylboron reagent with the proper selection of the phosphine ligand. The scope of this process is depicted in Figure 3. Using enantioenriched, unactivated alkyltrifluoroborate nucleophiles, ligand-controlled stereoselectivity was broadly achieved in cross-coupling reactions with aryl electrophiles. Strongly π-accepting ligands bis- CF₃PhSPhos (11) and bis-CF₃PhXPhos (14), which emerged from our parameterization-guided optimization, preferentially promote the stereoretentive pathway, while strongly σ-donating ligand PAd₃ (9) preferentially promotes the stereoinvertive pathway. Because electron-poor palladium catalysts commonly undergo slow oxidative addition with aryl chlorides, we also evaluated aryl bromide and triflate electrophiles in reactions involving 11 and 14. A particular highlight of this protocol is the uniformity of the conditions used for both the stereoinvertive and stereoretentive reactions: each operates in a toluene/water mixture as solvent, with a carbonate base, and no additional additives. Both reaction variants tolerated the use of electron-rich and electron-deficient aryl electrophiles, as well as an aryl electrophile bearing an ortho-substituent. High stereofidelity was achieved for all of these reactions, including those involving alkylboron nucleophiles bearing thiophenyl and phenoxide substituents. Use of an alkylboron nucleophile containing a larger substituent at the stereogenic center was also well-tolerated (16i). Diastereomeric products 17a and 17b could be generated from a single alkylboron diastereomer (37) using 14 and PAd₃, respectively (Figure 3B). In these reactions, replacement of ligand 14 with PAd₃ resulted in a change in diastereoselectivity from 30:1 to 1:5, a 3.6 kcal/mol free energy of activation difference dependent only on the ligand identity. No erosion of specificity was observed for electron-deficient aryl substrates in stereoinvertive Suzuki reactions using PAd₃, in contrast to analogous reactions using P^tBu₃. Furyl and thiophenyl electrophiles are also compatible with our system (Figure 3C). As an additional mechanistic probe, trans-2- methylcyclopentyltrifluoroborate was subjected to the stereodivergent reaction conditions. Because trans-2-methylcyclopentyltrifluoroborate is sterically impeded from undergoing stereoinvertive transmetallation, only the stereoretentive process using 14 should be mechanistically viable. Indeed, we observed that use of ligand 14 smoothly generates 18 with stereoretention, while use of PAd₃ results in low alkylboron conversion.

Figure 3.

Stereodivergent Pd-catalyzed cross-coupling reactions using enantioenriched alkylboron nucleophiles. % es = % ee (final product) / % ee (starting material). † 44% yield, 84% ee (86% es) when run at 60 °C.

To further probe the origin of the ligand-dependent enantiodivergent process, we interrogated the mechanism of transmetallation using the parameterization strategy described above (Figure 4). To accomplish this, phenyl-substituted substrate 20 was selected due to enhanced performance and thus a greater output range. Additionally, 24 ligands were tested, excluding smaller ligands to reduce the complexity associated with β-hydride elimination. Multivariate linear regression revealed that most of the outputs can be expressed in two readily interpretable terms that discriminate the transmetallation pathways: the average energy of the P−C antibonding orbitals E_σ*(P−C), representative of π-back bonding, and the energy of the lone pair orbital of phosphorus ELP(P), a measure of the ligand’s σ-donation capability (Figure 4B). This outcome suggests that the stereoinvertive pathway is dependent on strong σ-donation from the ligand, which may stabilize a two-coordinate, cationic palladium complex. Conversely, the stereoretentive pathway is enhanced by π-back bonding, which may stabilize the coordination of a π-donor ligand X (presumably OH⁻) to Pd. Including two steric descriptors such as B₁ and L improves the model fit by treating the competitive β-hydride elimination that occurs using smaller ligands and decreases the observed specificity. This becomes evident when the four smallest ligands in this dataset are removed from the analysis, which results in an excellent correlation using just the two electronic descriptors with the experimentally observed stereochemical outcomes (Figure 4D). Multivariate regression analysis thereby provides compelling evidence for the electronic factors favoring each transmetallation mechanism and thus a guideline for future developments in stereospecific cross-coupling reactions.

Figure 4.

Mechanistic investigation by multidimensional regression modeling. A) Data from the arylation of 20 was used. [PdL] is the precatalyst as shown in Figure 3 with varying ligands. B) Regression model containing all 24 ligands of this data set. E_{σ*(P−C)avg} ^Boltz: Boltzmann-weighted average across the conformers of the average energies of the three P−C σ* antibonding orbitals in each phosphine. E_LP(P) ^Boltz: Boltzmann-weighted average of the energy of the phosphorus lone pair orbital. Sterimol B₁ ^Boltz is the least width and Sterimol L^minconf is the length of the lowest-energy conformer as seen from opposite the P substituents. Red points in the diagram: validation data (EV) not used in the model training. LOO = leave-one-out cross validation score. K-fold: average leave-three-out cross-validation score. C) Illustration and interpretation of the model terms. D) Regression model after removing the four smallest ligands in this data set to exclude the influence of competitive β-hydride elimination on the data.