Conformationally Locked Pyramidality Explains the Diastereoselectivity in the Methylation of trans-Fused Butyrolactones.

A stereoselectivity model inspired by the total synthesis of stemona alkaloids is developed to explain why enolate-derived 3,4-fused butyrolactones are methylated with a preference for syn alkylation. The model shows how conformational locking present in nonplanar enolate structures favors syn over anti methylation, due to less significant structural distortions in the syn pathway. The developed model was also successfully used to rationalize selectivities of previously documented methylation reactions.

During our work on the total synthesis of stemoamides, we discovered a marked preference for the methylation of the enolate derived from 9a-epi-norstemoamide 1 to give a syn product 2 (Figure a).[1] The same sense of selectivity has been reported on several occasions for a range of trans-fused γ-butyrolactones,[2] but the observed selectivity has not been explained in the literature (for selected lactones, see 3–5 in Figure a). Typically, enolate face selectivity has been rationalized on the basis of either steric effects alone or through a combination of steric and stereoelectronic effects.[3] The trans-fused lactones of Figure a, however, do not appear to have any obvious steric bias, and there are no obvious stereoelectronic effects that would readily explain the observed selectivity. Torsional effects have been found to be of key importance in stereoselective electrophilic addition reactions,[4] including alkylations of various five-membered ring enolates.[5] However, to invoke torsional effects as an explanation requires a full computational analysis of the relevant transition states (TSs), and as such, these TS-based arguments may remain less accessible to a synthetic chemist in planning a stereoselective synthesis.

Figure 1

Selected experimental observations (a) and the proposed stereoselectivity model (b).

Herein, we present a detailed computational analysis of the stereocontrol elements in methylation reactions of 3,4-trans-fused butyrolactones. We demonstrate that the stereoselectivity of the methylation of this class of butyrolactones can simply be deduced from the sense of pyramidality of the ground state (GS) of the enolate, which is constrained by the ring fusion.[6] The pyramidality can be quantified by the O–C–C–H torsion angle ϕ. We predict that the syn attack is favored because it involves a smaller change of ϕ and thereby a smaller distortion in the TS compared with the anti attack.

For 9a-epi-norstemoamide, DFT calculations pointed to a slight thermodynamic preference for the anti-methylated product; however, the formation of the syn product was found to be kinetically favored.[1] No steric effects could be identified in the located transition states, and we hypothesized that the preferential syn attack of the methylating agent is associated with the prepyramidalization of the nucleophilic carbon atom of the enolate (Figure b), potentially providing a less strained transition state for syn methylation.

In our computational analysis, we first employed the energy decomposition scheme offered by the activation strain model (ASM)[7] to the transition states identified for the syn and anti methylation pathways of 9a-epi-norstemoamide (see syn-TS and anti-TS in Figure ). The relatively early nature of these transition states (the reacting carbon atoms are at least 2.4 Å apart in these transition states) justifies the relevance of this approach,[8] which has been widely applied to rationalize reactivities and selectivities.[7,9]

Figure 2

Transition states computed for the syn and anti methylation pathways of 9a-epi-norstemoamide 1. Relative stabilities (ΔΔG) are given in parentheses (in kcal/mol), forming C···C distances in Å. H atoms of the enolate (except those of the butyrolactone ring) are omitted for clarity.

The energy components obtained from the ASM analysis are listed in Table , and the results reveal that the difference in the electronic energies of the two diastereomeric transition states (ΔΔE⧧ = 2.1 kcal/mol) is predominantly related to the distortion (i.e., structural change) of the reactants (enolate en and MeI) with respect to their ground-state structures. The distortion of the reacting partners involves the elongation of the Me–I bond and significant pyramidalization of the enolate C10 carbon atom required for C–C bond formation.[1] The distortion energies ΔEdist(en) and ΔEdist(MeI) are notably higher in the disfavored transition state anti-TS as compared to syn-TS, although the overall effect (ΔΔEdist(en) + ΔΔEdist(MeI) = 3.0 kcal/mol) is somewhat reduced when considering the difference in the interaction energies (ΔΔEint = −0.9 kcal/mol). Altogether, the total energy difference (ΔΔE⧧ = 2.1 kcal/mol) is well captured by the distortion energies of the enolate (ΔΔEdist(en) = 1.9 kcal/mol).[10]

Table 1

Energy Data from ASM Analysis (in kcal/mol)a

energy component	anti-TS	syn-TS	ΔΔEb
ΔEdist(en)	2.9	1.0	1.9
ΔEdist(MeI)	12.9	11.8	1.1
ΔEint	–19.5	–18.6	–0.9
ΔE⧧	–3.7	–5.8	2.1

Energy data are obtained form electronic energies computed at the ωB97X-D/Def2TZVPP level. en and MeITS denote distorted structures of the reactants in the transition states.

Energy difference obtained from the previous two columns.

The ASM analysis thus implies increased distortion (strain) in the enolate along the anti alkylation pathway, which is also apparent from the torsion angles associated with the forming C–C bond. As illustrated in Figure , the vicinal bonds adjacent to the reacting C10 carbon atom are fairly close to the eclipsed arrangement in the anti-TS, whereas no such small torsion angles are seen in the more favored syn-TS or in the ground-state enolate en. It thus appears that the concept of torsional control of stereoselectivity[4] can be applied for the present reaction as well.

Figure 3

Dihedral angles relevant to torsional strain in enolate en and in the transition states of methylation. Newman projections are viewed from the direction indicated by the red arrow.

To assess the relation between the torsional strain induced in the enolate along the two methylation pathways and the degree of pyramidalization of the enolate C10 carbon atom, we examined the energy variation of the bare enolate en (i.e., without MeI) as a function of the O–C=C–H dihedral angle (see Figure ). This torsion angle ϕ is a natural choice to quantify the deviation from the planar enolate structure and the pyramidality of the C10 atom. Computations predict ϕ = 12.6° for the ground-state structure of enolate en. The potential energy curve depicted in Figure indicates that the energy penalty of reaching the dihedral angle measured in the anti-TS transition state (ϕ = −22.1°) via the inversion of pyramidality is 1.8 kcal/mol larger than the energy change accompained by the pyramidalization at syn-TS (ϕ = 25.6°). This energy difference is almost identical to that obtained in the strain analysis for the distortion of the enolate (ΔΔEdist(en) = 1.9 kcal/mol), suggesting that the dihedral angle ϕ is a reasonable indicator of the torsional strain induced on the two methylation pathways.

Figure 4

Potential energy curve derived by constrained geometry optimization of enolate en varying the O–C–C–H dihedral angle (highlighted in red). Dihedral angles measured in the optimized structures of en and the two transition states are marked by arrows.

The role of ring fusion and pyramidality in stereoselectivity control was evaluated next by carrying out DFT calculations for selected model enolates (Figure a).[11] The simple γ-butyrolactone enolate en exhibits a nonplanar equilibrium structure with a pyramidalized carbon center (ϕ = 9.6°). The pyramidalization arises from minimization of torsional strain of the five-membered ring. The C4–C5 bond adopts a noneclipsed conformation (Figure b, left), and the resulting nonplanar ring (twist, 4T5 conformation) is further stabilized by pyramidalization at C3 as the C3–H bond adopts a staggered arrangement with the vicinal C4–H bonds (Figure b, right).

Figure 5

(a) Selected model enolates with characteristic torsion angles (ϕ and θ; as defined by the O–C=C–H and O–C–C–C units of the cyclic enolates). Data obtained for enolate en are also given for reference. Computed stereoselectivities (ΔΔG data; in kcal/mol) are given in parentheses. Most of the hydrogen atoms (except those of the butyrolactone ring) are omitted for clarity. (b) Rationalization of the sense of pyramidalization, using enolate en as an example.

Interestingly, trans-fusion with a cycloheptane ring does not alter the pyramidality of the nucleophilic carbon atom (en; ϕ = 9.7°); however, a more constrained cyclohexane ring exhibits increased pyramidality (en; ϕ = 14.8°). The pyramidality of the C10 carbon atom in enolate en is somewhat between those in en and en, implying that the azepane ring and the cyclic amide unit in the trans-fused system impose additional structural restraint in the enolate. The dihedral angle defined by the O–C–C–C unit of the five-membered ring (θ in Figure a) represents another characteristic structural parameter of the cyclic enolates, and they also point to an enhanced distortion effect in en. Transition states for the syn and anti methylation pathways for bicyclic enolates en and en were computed, and the predicted selectivities (see ΔΔG values in Figure a) correlate fairly well with the ϕ dihedral angle.[12]

We note that the ground-state structure of enolate en is flexible because the nonplanar five-membered ring can easily flip. The ring inversion occurs via a planar transition structure, so it involves the variation of the θ dihedral angle as well (see Figure ). The rotational flexibility is restricted in fused bicyclic enolates. This is apparent from the ΔE(θ) potential energy curves computed for trans-fused enolates en, en, and en, which all display a single energy minimum, and a substantial amount of energy is required to reach the planar (θ = 0°) region. Consequently, the nonplanar conformations of these enolates are confined (locked), and the innate pyramidality of the nucleophilic carbon provides an energetic bias for the syn methylation pathway as discussed above.

Figure 6

Potential energy curves derived by constrained geometry optimization of various enolates varying the θ dihedral angle.

We have also examined a number of methylation reactions reported for bi- and tricyclic trans-fused butyrolactones,[2] and our DFT calculations confirm the syn selectivity observed in all these cases (Figure ).[13] For the complex 5 + 7 + 5 tricyclic fused systems A(2a) and B,[2b] we find a similar degree of selectivities as for our target reaction with 9a-epi-norstemoamide 1. The syn selectivity increases slightly when unsaturation is introduced in the seven-membered ring (product C).[2c−2e] In accordance with the results obtained for model en, calculations predict further enhanced selectivities for all 5 + 6 bicyclic systems (D–H),[2f−2k] which are particularly high if heteroatoms are involved in the six-membered ring (G and H).[2j,2k] In these latter cases, the bicyclic systems bear bulky substituents as well (OBn and Cbz), but no steric effects could be identified that would alter the selectivity.

Figure 7

Syn products observed in various alkylation processes and selected computed properties (ϕ and θ as torsion angles of corresponding enolates; ΔΔG as predicted selectivities).

In summary, we have disclosed how methylation of 3,4-trans-fused butyrolactones occurs in a syn-selective fashion. The nucleophilic carbon center of the five-membered ring enolate is pyramidalized. The direction of the pyramidalization is essentially determined by the stereochemistry of the trans-ring fusion. The nonplanar enolate structure prefers alkylation from the direction of the pyramidalization, affording a kinetically more favored pathway for the formation of syn products. The emerging model implies that the stereoselectivity of this class of methylation reactions can be inferred simply from the ground-state structure of the enolate, i.e., without the inspection of the diastereomeric methylation transition states. The results obtained herein are a testimony to how total synthesis efforts can help to identify gaps in both synthetic methodology as well as in our understanding of stereoselectivity.