Development of a Strategy for Linear-Selective Cu-Catalyzed Reductive Coupling of Ketones and Allenes for the Synthesis of Chiral γ-Hydroxyaldehyde Equivalents

We report the development of a stereoselective method for the allylation of ketones utilizing N-substituted allyl equivalents generated from a chiral allenamide. By choice of the appropriate ligand for the Cu-catalyst, high linear selectivity can be obtained with good diastereocontrol. This methodology allows access to chiral γ-hydroxyaldehyde equivalents that were applied in the synthesis of chiral γ-lactones and 2,5-disubstitued tetrahydrofurans.

C hiral alcohols are ubiquitous in organic molecules prepared both naturally and synthetically for a desired biological function. Therefore, development of synthetic methods to access chiral alcohols has been an intense area of research in the field of organic chemistry. 1 One of the most highly studied areas of stereoselective alcohol synthesis is in the controlled addition of an allylmetal reagent to an aldehyde or ketone electrophile. 2 Pioneering work in stereoselective allylation typically employed the generation of a stoichiometric chiral allylmetal nucleophile in a separate step to be used in the allylation reaction with an aldehyde or ketone to generate the chiral alcohol. 3 Over the years, catalytic methods to generate the reactive allylmetal in situ from an unreactive allyl source and metal catalyst have emerged. 4−7 In particular, reductive coupling strategies 8 that generate the reactive allylmetal from unsaturated hydrocarbons via hydrometalation are extremely powerful, atom-economical approaches for the synthesis of chiral homoallylic alcohols. 9 Recently, an elegant catalytic method for the allylation of ketone and imine electrophiles was developed by Buchwald employing hydrocupration of carbon-substituted allenes (2) 10 or 1,3-dienes 11 by a Cu−H catalyst to generate the reactive allylmetal reagent in situ ( Figure 1A). In the ketone version of this process, 10a the anti-diastereomer of the branched product (anti-b-3) was formed as the major product in high diastereoand enantioselectivity when using chiral bis(phosphine) ligands. However, the linear product l-3 was not formed. Our group became interested in developing a method to generate linear products utilizing this approach, which has not been reported with ketones. While Buchwald demonstrated that both linear and branched products could be obtained when using imine electrophiles 10c by changing the Nsubstituent on the imine, this is not possible with ketone electrophiles.
Our working mechanistic hypothesis for regio-and diasteroselectivity for this reaction is given in Figure 1B.
Hydrometalation of allenes typically occurs trans to the R′substituent of the allene due to steric reasons. 12 Therefore, initial hydrocupration of 2 would be expected to afford the Zisomer of the linear (allyl)Cu species l-Z-5. Buchwald has already determined that the turnover-limiting step is allylation of the ketone, 10a so π−σ−π equilibration of 5 would be expected. Assuming the allylation step proceeds through closed Zimmerman−Traxler 13,14 chairlike transition states (6), the major product obtained in Buchwald's report 10a (anti-b-3) would be derived from anti-b-7 from reaction between l-E-5 and 1 via 6c. The preference for this isomer can be easily rationalized by steric effects. Arguably, l-E-5 would be the least sterically hindered of the three possible (allyl)Cu intermediates (5) causing it to be the dominant species in the reaction. Therefore, if this mechanistic hypothesis is correct, then to obtain the linear product (E-l-7), conditions would need to be designed to either stabilize the branched (ally)Cu intermediate b-5 relative to l-5 or make it more reactive.
Our strategy to stabilize (allyl)Cu intermediate b-5 is shown in Figure 1C. Use of allene 8 containing a heteroatom tethered ligand should initially generate linear (allyl)Cu species l-9 after hydrocupration. The tethered ligand could then help stabilize the branched (allyl)Cu intermediate b-9 through coordination to Cu. Reaction of b-9 with a ketone would then generate linear product l-10 with an enol (X = O) or enamine (X = NR) group representing a masked aldehyde functionality to provide useful chiral γ-hydroxyaldehyde equivalents. Additionally, use of a chiral tethered ligand in allene 8 could enable stereocontrol of the newly formed stereocenter of 10. Overall, we envisioned that this methodology could be a valuable entry into chiral lactone 15 or tetrahydrofuran 16 containing natural products (Scheme 1) through lactol 11 obtained by hydrolysis of the enol or enamine functionality of 10. Herein, we report our findings on the development of a diastereoselective copper-catalyzed reductive coupling of a chiral allenamide with ketones to access the linear isomer (10) of product.
To identify an allene that fit the requirements of 8, we initially chose to investigate allenamide 14 derived from Evans' auxiliary because it has been synthesized previously (Table  1). 17 Furthermore, we had hoped that the carbonyl group of the oxazolidinone would serve as a sufficient coordinating group for Cu. 18 Additionally, based on our design in Figure  1C, we focused on reaction conditions where Cu would have a low coordination number to facilitate potential coordination of the oxazolidinone carbonyl group (i.e., noncoordinating solvents, monodentate ligands).
Trialkyl monodentate phosphines favored the formation of linear product l-15a with modest l/b selectivity (entries 1, 3, and 4). Furthermore, use of dcpe, a bidentate ligand commonly employed in Cu−H catalyzed reductive coupling reactions, 10,11 also gave preferentially the branched product (entry 2). A further survey of monodentate phosphine ligands of varying electronic 19 and steric 19a properties revealed that the linear selectivity was largely influenced by the electrondonating ability of the ligand employed with less electrondonating ligands affording higher linear selectivities (compare entries 1, 3, 4, and 8−14). There was a rough correlation between ligand cone angle and diasterocontrol with larger ligands affording higher diastereoselectivity (compare entries 1, 3, 4, 9, 10, 12, and 13). Ultimately, phosphoramidite ligand 16 afforded the highest reaction yield with excellent linear selectivity and good diastereoselectivity.
The substrate scope for the linear-selective reductive coupling of ketones and allene 14 is given in Scheme 2. In general, high linear selectivity was obtained in good to excellent reaction yield for halogenated (l-15f,i), electronrich (l-15b,c,j,k), and electron-poor (l-15e,m) arenes. Hindered ketones bearing ortho-substitution on the aryl group required heating to achieve full conversion; however, this did not severely reduce the diastereoselectivity (l-15c,h,i). Additionally, diastereoselectivity and linear selectivity were reduced when the steric bias between the two R-groups of ketone 1 was reduced (e. g. l-15d,l,n,o,p,u). Notably, a nitrile Scheme 1. Chiral Lactone and THF Containing Natural Products group was not reduced under the reaction conditions (l-15q), and both amino (l-15s) and a free hydroxyl group (l-15k) was also tolerated. In regards to factors dictating branched/linear selectivity and stereocontrol in these reactions, studies employing achiral allenamide 18 were informative (Scheme 3). Use of 18 lacking substitution on the oxazolidinone ring afforded reduced linear selectivity in the reaction when the optimized ligand 16 was used. Additionally, use of PCy 3 in the reaction employing allenamide 18 led to a turnover in the reaction selectivity and favored formation of the branched product b-19. In contrast, with chiral allenamide 14, use of PCy 3 as a ligand afforded linear selectivity (Scheme 3 and Table 1, entry 1). Based on these observations, and the results of our ligand optimization survey (Table 1), a model to rationalize regio-and stereocontrol in these reactions could be developed ( Figure 2).
Mechanistically, hydrocupration of allene 14 or 18 is expected to initially form the Z-linear (σ-allyl)Cu complex (l-Z-20; vide supra) that will be in equilibrium with the branched (σallyl)Cu complex (b-σ-20) through the intermediacy of the (πallyl)Cu complex π-20. The π-allyl geometry and coordination of the oxazolidinone group to Cu in complex π-20 are proposed based on structural information determined by X-ray crystallography and NMR spectroscopy for related anions of this type found in the literature. 20 Considering the turnoverlimiting step in Cu-catalyzed reductive coupling of ketones and allenes is believed to be the addition of the (allyl)Cu reagent to the ketone electrophile, 10 this would allow for a pre-  equilibrium between l-Z-20 and b-σ-20 to be established before reaction with the ketone. Therefore, a model to rationalize regioselectivity in the reaction could be developed based on considering the stability of these two intermediates whereby a preference for l-Z-20 would result in a branched selective process while preference for b-σ-20 in the reaction would result in linear selectivity. 21 Due to the Z-olefin geometry formed in the initial hydrocupration event, reaction regioselectivity could be explained by a competition between the strength of the oxazolidinone coordination versus the size of the A 1,3 -strain present in l-Z-20 ( Figure 2). The high linear selectivity obtained as the electron-donating ability of the phosphine ligand decreases (Table 1) can be explained by an increase in the preference for complex b-σ-20 due to the enhanced electrophilicity at Cu. Additionally, the magnitude of the A 1,3strain in l-Z-20 would be expected to affect the overall equilibrium between the (allyl)Cu complexes. As a result, when the poorly electron-donating ligand 16 is employed, coordination of the oxazolidinone to the electrophilic Cu atom leads to a preference for b-σ-20 leading to linear selectivity when using either allenamide 14 or 18. A reduction in linear selectivity with ligand 16 when using allenamide 18 in place of 14 can be rationalized by the presence of increased amounts of l-Z-20 due to the reduction in the magnitude of the A 1,3 -strain present in l-Z-20 when R = H. In contrast, when the electronrich ligand PCy 3 is used, the branched product (b-19) is preferred when the unsubstituted allenamide 18 was used. This may result from a shift in the equilibrium of the (allyl)Cu complexes to favor l-Z-20 because of the reduced coordinating ability of the oxazolidinone to the more electron-rich Cu atom. When the magnitude of the A 1,3 -strain present in l-Z-20 is increased by utilizing the chiral allenamide 14 with PCy 3 as ligand, the oxazolidinone coordination is presumably enhanced by destabilizing l-Z-20 leading to preferential linear selectivity in the reaction. Furthermore, it is important to point out that if the A 1,3 -strain present in l-Z-20 is involved in governing regiochemical control in this reaction, then the alkene moiety in b-σ-20 likely remains coordinated to Cu. 22 If the alkene of bσ-20 were to disassociate from Cu, isomerization of the Zalkene to the E-isomer l-E-20 could occur that would remove this A 1,3 -interaction that is proposed to be important. Additionally, it is possible that b-σ-20 may not be a discrete intermediate in these reactions, and rather, π-20 may be the dominant species that reacts directly with ketone 1a to afford linear product l-15a/19. 22 However, this scenario is also consistent with the model described above for regiocontrol. Finally, the absolute stereochemistry and the Z-olefin geometry of the linear product l-15a can be rationalized by the reaction of b-σ-20 or π-20 with ketone 1a through chair-transition state 21 with the oxazolidinone group in an axial position and complexed with Cu for selective reaction to the Si-face of 1a. Transition state model 21 is supported by literature precedent 20b,c and further supports oxazolidinone coordination in these processes.
Demonstration of the synthetic potential of the reductive coupling products is outlined in Scheme 4. Reduction of the oxazolidinone of l-15a with excess DIBAL afforded lactol 22 after hydrolysis of the resultant eneamine formed in the reduction to unmask the chiral γ-hydroxyaldehyde equivalent. Lactol 22 could then be converted to chiral γ-lactone 23 by oxidation with TPAP/NMO or converted to the 2,5substituted tetrahydrofuran 24 in good yield albeit with poor diastereocontrol in the Et 2 Zn addition. 23 Finally, the linear selective reductive coupling reaction was performed on a 1.0 g scale to complete a three-step asymmetric synthesis of the natural product (S)-(−)-boivinianin A 15a−c starting from 4′methylacetophenone.
In conclusion, we have disclosed a strategy for the stereoselective reductive coupling of ketones and a chiral allenamide to selectively generate the linear reaction products providing useful chiral γ-hydroxyaldehyde equivalents. This method employs simple starting materials and a readily available catalyst system to furnish chiral products with increased complexity in an efficient manner. Further development of this reaction to enable stereocontrol by a chiral catalyst is currently under investigation and will be reported in due course.

Notes
The authors declare no competing financial interest.

Organic Letters
Letter ■ ACKNOWLEDGMENTS Startup funding from the Virginia Commonwealth University and the Bill and Melinda Gates Foundation (The Medicines for All Institute, Grant Number OPP#1176590) is gratefully acknowledged. We thank Dr. Joseph Turner (VCU) for assistance in collecting HRMS data.