|Home | About | Journals | Submit | Contact Us | Français|
The Kishi reduction of a planar oxacarbenium was investigated theoretically. The high diastereoselectivity for hydride transfer to the oxacarbenium intermediate is attributed to the conformation of the transition state that places the allyl side chain in an equatorial position in the major transition state and axial position in the minor. The minor transition state is destabilized by a 1,3-diaxial strain between the attacking hydride and the syn allyl side chain.
One of our groups recently reported an efficient synthesis of two pyranopyran subunits of norhalichondrin B.i A highlight of the assembly is the remarkably diastereoselective Kishi reductionii of hemiacetals 1a and 1b (Scheme 1).
Diastereoselective nucleophilic addition to saturated glycosides has been explained on the basis of anchimeric assistanceiii (2, Scheme 2) as well as selective anti addition to the favored ground state conformation of 3 (Scheme 2).iv However the greater than 20:1 diastereoselectivity obtained in the reduction of pyrananone 1, which does not bear such directing groups, cannot be explained by these models. We now report a computational investigation that provides an explanation for the high reduction selectivity observed. All structures were computed in the gas phase using density functional theory B3LYPv with the 6–31G(d)vi basis set as implemented in the Gaussian 03vii suite of programs. All stationary points were verified as minima or first-order saddle points by vibrational frequency analysis.
Evidence strongly suggests that the reductions of hemiacetals 1a and 1b proceed through discrete oxacarbenium ions. The hemiacetals are formed as a mixture of diastereomers ranging from 3:1 to 9:1 dr, while subsequent reduction yields material with greater than 20:1 diastereoselectivity. If the reaction were to proceed via an SN2 mechanism or ion pairs in which one face is shielded by the leaving group, one of the hemiacetal diastereomers must be selectively destroyed, which is unlikely (although not impossible). Crich and co-workers have performed work which also provides evidence that the transition states for glycosylation reactions have high oxacarbenium character.viii
Oxacarbenium 4 (Figure 1) was used to model the oxacarbenium intermediates obtained from 1a and 1b. The optimized structure of 4 is found to have a nearly perfectly planar six-membered ring. Starting from a variety of nonplanar geometries, calculations always converged to the planar conformation shown in Figure 1. The torsion angles around every ring bond are less than 1°, and a Newman projection along C2-C3 shows nearly perfect staggering around this bond.
Attempts to locate transition structures for reaction of 1a–b with SiH4 resulted in proton abstraction from the oxacarbenium ion. To model the attack of hydride on 4, the energy was scanned as a function of the forming C-H distance (Table 1).ix Because the free energy difference between attack from the favored and disfavored faces remains constant throughout the bond scan (approximately 2 kcal/mol), it is reasonable to conclude that the same factors which cause this energy difference also account for the selectivity observed in the transition state. One imaginary frequency for hydride transfer to 4 Å first appeared at a distance of 1.51 Å for both the major (7d) and minor (8d) faces of attack and continued through distances of 2.31 Å. Structures optimized at distances greater than 2.31 Å resulted again in proton abstraction from the oxacarbenium. Conformational changes were similar throughout the scan (compare 7d and TS1, and 8d and TS2, Figure 2). The geometries obtained at 2.31 Å were selected as representative transition structures for the anti and syn attacks. These structures are represented as TS1 and TS2, respectively (Figure 2).
The C2 substitutent that is syn to the attacking hydride—H in TS1 and methyl in TS2—adopt axial positions. To investigate the origin of this conformational change, the transition structure for hydride transfer to unsubstituted oxacarbenium 9 was modeled (TS3, Figure 3). Like oxacarbenium 4, model 9 optimizes to a planar geometry. And like TS1 and TS2, the C2 substituent that is syn to the attacking hydride (hydrogen) becomes axial in TS3.
The transition state conformation results from the well-known fact that nucleophiles favor anti addition to p-bonds. This has been found for alkenes and alkynes (Scheme 3),x and the preference arises to maintain maximum (anti) overlap of the orbitals corresponding to the π orbitals of the reactant.
In the case of TS3, hydride attack on C6 of TS3 occurs anti to a developing lone pair of the ring oxygen (Figure 4). There is staggering of bonds to C2 with respect to lone pairs on the ring oxygen , causing a 1 , 3-diaxial relationship between the hydride and the syn C2 substitutent. In other words, hydride attack occurs with optimal staggering around the C2-O1 and O1-C6 bonds.
Applying this model to the reduction of 4, favored transition structure TS1 shows perfect staggering, with the C2 methyl in an equatorial position (Figure 5). By contrast, TS2 is distorted away from the ideal geometry because it must overcome very large 1 , 3-diaxial interaction between the axial methyl group and the forming C-H bond (Figure 6). A related structure is 2-methyltetrahydropyran in which the axial-methyl conformer is 2.9 kcal/mol less stable than the equatorial.xi Boons and coworkers have shown that the glycosylation of arabinofuranosyl oxacarbenium ions also favor a staggered mode of nucleophilic attack.xii It is remarkable that our oxacarbenium ions, which do not bear the ground state steric or torsional effects noted by Boons, react with such high diastereoselectivity.
To determine the effect of the pyranosyl ring conformation changes on the energies of TS1 and TS2, single point energy calculations were performed on the frozen transition structures with the hydride removed. In the absence of hydride, the oxacarbenium ion geometry in TS1 is disfavored by 3.7 kcal·mol−1, presumably due to favorable C2-Me and C6-cationic interactions in the TS2 geometry.
In conclusion, the oxacarbenium intermediate of the Kishi reduction is planar, but becomes staggered around the C2-O1 and O1-C6 bonds in the transition state. This places the allyl side chain in an axial position for syn attack. This is less stable than the anti attack, where the allyl side chain is equatorial.
The preference for staggering in this case further extends the role of torsional steering in the control of stereoselectivity.xiii
We are grateful to the National Institute of General Medical Sciences, National Institutes of Health (G36700, KNH) and the National Cancer Institute (CA110246, AJP) for support of our work.