Transition State Gauche Effects Control the Torquoselectivities of the Electrocyclizations of Chiral 1-Azatrienes

Abstract

Hsung et al. have reported a series of torquoselective electrocyclizations of chiral 1-azahexa-1,3,5-trienes that yield functionalized dihydropyridines. To understand the origins of the torquoselectivities of these azaelectrocyclizations, we modeled these electrocyclic ring closures using the M06-2X density functional. A new stereochemical model that rationalizes the observed 1,2 stereoinduction emerges from these computations. This model is an improvement and generalization of the “inside-alkoxy” model used to rationalize stereoselectivities of 1,3-dipolar cycloaddition of chiral allyl ethers and emphasizes a stabilizing hyperconjugative effect, which we have termed a transition state gauche effect. This stereoelectronic effect controls the conformational preferences at the electrocyclization transition states, and only in one of the allowed, disrotatory electrocyclization transition states is the ideal stereoelectronic arrangement of substituent achieved without the introduction of a steric clash. Computational experiments confirm the role of this effect as a stereodeterminant, since substrates with electropositive groups like a silyl substituent and electronegative groups have different conformational preferences at the transition state and undergo ring closure with divergent stereochemical outcomes. This predicted reversal of stereoselectivity for the ring closure of a silyl-substituted azatriene has been demonstrated experimentally.

Graphical abstract

Introduction

Hsung et al. have demonstrated that an allylic stereocenter can influence the stereochemical outcomes of the disrotatory 6π electrocyclizations of substituted 1-aza-1,3Z,5-hexatrienes (summarized in Scheme 1).¹ Here we report – on the basis M06-2X/6–31+G(d,p) computations – that gauche effects present in the electrocyclization transition states determine the preferred mode of disrotatory ring closure for these chiral 1-azatrienes. In short, the forming C–N bond prefers to maintain a gauche relationship with the C–O bond at the allylic stereocenter.

Scheme 1.

Torquoselective electrocyclizations of 1-aza-1,3Z,5-hexatrienes

Ring closures of azatrienes have been examined by a number of researchers^2–13 and examples of stereoselective azaelectrocyclizations^14–21 have been reported. Despite the utility of the 6π electrocyclization in natural product synthesis and even in biological contexts (i.e., covalent enzyme inhibition and biolabeling),^22,23 the origins of the torquoselectivities of these electrocyclizations have yet to be understood. Previous experimental work suggests electrocyclizations of 1-azatrienes are often reversible at elevated temperatures,³ unlike the 6π electrocyclization of 1,3Z,5-hexatriene. However, experiments suggest that azatrienes shown in Scheme 1 undergo irreversible ring closure;¹ hence, the stereoselectivities of these reactions are under kinetic control.

Results and Discussion

Figure 1 illustrates the allowed disrotatory transition structures of the ring closures of 1,3,5-hexatriene 1 and those of its 1-aza analogue 2. Whereas the electrocyclization of 1,3,5-hexatriene proceeds through a C_s symmetric boat-like geometry, the transition structures for the corresponding azaelectrocyclization are of lower symmetry; the structures are distorted such that the nitrogen of the azatriene appears to “approach” the C-terminus of the azatriene from above or below the plane of the terminal double bond. According to our previous work, this distortion allows the lone pair of the imine nitrogen to overlap with triene π system, stabilizing the transition state. Lone pair conjugation of this sort is largely responsible for ~ 10⁷-fold (at 298.15 K, see Figure 1 for free energies of activation) increase in reactivity of 1-azahexa-1,3,5-triene relative to 1,3,5-hexatriene.²⁴

Figure 1.

The thermally allowed disrotatory transition structures and activation free energies of the ring closures of 1,3,5-hexatriene and 1-azahexa-1,3,5-triene. Energies are Gibbs free energies in kcal mol⁻¹.

1-azatriene 3 is representative of the substrates examined by Hsung and coworkers.¹ It features a number of C–C single bonds capable of free rotation; thus, to reduce the number of conformations that needed to be sampled we examined the simplified trienes, 5a–c, 6, and 7, shown in Scheme 2. In all cases, three rotameric transition states were found for each of the disrotatory modes of ring closure (see Figure 2). We labeled these transition states based on the position of the alkoxy substituent with respect the terminal alkene of the azatriene, nomenclature previously used to describe stereoselective additions to alkenes bearing allylic stereocenters.²⁵

Scheme 2.

A representative azaelectrocyclization and model substrates examined computationally.

Figure 2.

Newman projections of the C5–C6 bond of the M06-2X/6–31+G(d,p)-optimized transition state conformers for the ring closure of 5a. ΔΔG^‡ values are reported in kcal mol⁻¹.

For both disrotatory modes of ring closure of 5a, the transition states in which the alkoxy substituent in the inside position (see Figure 2) are most stable. A gauche effect is responsible for this conformational preference. In the TS5a_in1 and TS5a_in2, the forming C–N bond and the C–O bond between the alkoxy substituent and the allylic stereocenter are antiperiplanar to σ-donating bonds (C–C and C–H bonds, respectively) and gauche to one another. The anti transition state conformers TS5a_anti1 and TS5a_anti2, which do not have these favorable donor-acceptor interactions, and are both about 2 kcal mol⁻¹ higher in energy. The outside-alkoxy transition states (TS5a_out1 and TS5a_out2) also have a gauche arrangement of the C–O and forming C–N bonds and as a result are more stable than the anti transition structures. However, these transition structures are higher in energy than TS5a_in1 and TS5a_in2 due to electrostatic repulsion between the oxygen of the alkoxy group and the azatriene nitrogen. The O–N interatomic distances are shorter in the outside transition state conformer than the corresponding inside transition structures.

The gauche effect described above differs from a ground state gauche effect^26,27 in that one of the C–X σ acceptors is a bond that is only partially formed in the transition state. For this reason, we have termed this hyperconjugative effect a “transition state gauche effect”. For these chiral azatrienes, it is the gauche relationship between forming C–N and the vicinal C–X bonds at the allylic position of the triene that is ultimately responsible for stereoinduction (Scheme 3). Transition state gauche effects of this type are likely general phenomena, and may be responsible for the stereoselectivity of other reactions involving alkenes that possessing an allylic σ acceptor. The “inside-alkoxy” effect invoked by Houk^25,28–31 and others³² to rationalize the stereoselectivities of 1,3-dipolar cycloadditions of nitrones and nitrile oxides and chiral allyl ethers is now recognized to be one case of this more general stereochemicial model.

Scheme 3.

Newman projections illustrating the ground state and transition state gauche effects.

To confirm that these stereoelectronic effects are the principal determinants of the transition state conformational preferences of these azaelectrocyclizations, we also modeled the transition states of related substrates 5b and 5c, bearing a fluorine or a silyl substituent at the stereogenic center, respectively. If these effects were important, then 5b (R = F) should have conformational preferences qualitatively similar to those of 5a at the transition state. Since the SiMe₃ group is a hyperconjugative donor³³, the anti transition state conformers of 1-azatriene 5c should be the most stable: A good donor (C–Si σ bond) is antiperiplanar to electron-deficient forming C–N bond in the anti transition states of 5c. The ΔΔG^‡ of all six transition state conformers of 5a–c are summarized in Figure 3.

Figure 3.

ΔG^‡ differences of the M06-2X/6–31+G(d,p) optimized ring closure transition state conformers of 5a–c. ΔΔG^‡ values were determined relative to inside TS_dis1 for the transition structures of azatrienes 5a and 5b and relative to TS_dis2 for the transition structures of 5c. Energies are reported in kcal mol⁻¹.

The conformational preferences of the transition states of 5b are, indeed, similar to those preferences observed for the transition states of 5a. The inside transition states TS5b_in1 and TS5b_in2 are most stable, and the anti transition states are highest in energy. By contrast, we find that the anti transition state conformers are preferred for the ring closure of 5c (R = SiMe₃), confirms the nature of C–N forming bond at the transition state as hyperconjugative acceptor. This anti preference is related to Cieplak’s rationalization regarding the origins of π–facial selectivity of nucleophlic additions to cyclic ketones,^34,35 and suggests the nature of bond formation in the electrocyclization, a pericyclic process, is fundamentally different from bond formation in a polar reaction (e.g., nucleophilic acyl addition) for which Cieplak’s rationale has been criticized.^36,37

Experimentally, Hsung et al. have demonstrated that the diastereoselectivities of these azaelectrocyclizations are sensitive to the nature of the N-substituent. To probe this effect, we first examined the transition states conformers of azatriene 6, bearing a N-methyl substituent, and, then, cyclic azatriene 7. The conformational preferences of 6 remain the same as those found for substrates 5a and 5b. According to computations, the two diastereomeric inside transition states of azatriene 6 differ in free energy by 1.0 kcal mol⁻¹ and are shown in Figure 4. This difference in ΔG^‡ is consistent with the sense and level of diastereoselectivity observed experimentally and is caused by a steric clash that destablizes TS6_in2. In TS6_in2, for the methoxy substituent to occupy its preferred inside position, the geminal methyl group must adopt the more sterically demanding outside position; whereas, in TS6_in1 the hydrogen occupies the outside position when the alkoxy is in the inside arrangement. That this steric clash involves the N–methyl group explains the substituent’s importance in stereoinduction.

Figure 4.

Newman projections of the lowest energy M06-2X/6–31+G(d,p)-optimized 6π electrocyclization transition state conformers of 6. ΔΔG^‡ values are given in kcal mol⁻¹.

We next modeled the ring closure of a cyclic azatriene, compound 7, (see Figure 5), which more closely resembles the experimental azatriene 4. The reaction of the cyclic azatriene 7 proceeds with a barrier of almost 18 kcal mol⁻¹ and yields a product that is 16 kcal mol⁻¹ more stable than the reactant. Azatriene 7 is more than 10³-fold more reactive at room temperature (ΔΔG^‡ = 5 kcal mol⁻¹) than 2a due to activation by the carbonyl moiety of the lactone and restriction of a rotational degree of freedom by the fused lactone ring. For azatriene 7, the outside transition states for either mode of ring closure are slightly more stable than the corresponding inside transition states. The subtle change of the conformational preference is due to A^1,3 strain between the methylene group of the lactone and N-methyl of the azatriene in the inside transition state conformers of 7. Nonetheless, a strong preference for a gauche conformation over the anti conformation is still predicted. The lowest energy conformers of either mode of disrotation differ by a ΔG^‡ of ~1 kcal mol⁻¹, consistent with the observed diastereoselectivity for the reaction of 3.

Figure 5.

Newman projections of M06-2X/6–31+G(d,p) optimized transition state conformers of the 6π electrocyclization of azatriene 7. ΔG^‡ are given in kcal mol⁻¹.

Finally, we return to the silyl derivative 5c. As discussed above, the silyl group at the allylic position of the azatriene is predicted to reverse the stereoselectivity of the ring closure. This prediction inspired the syntheses and studies of the ring closures of chiral silyl-substituted 1-azatrienes, including, 10 (see Scheme 4). The synthesis of silyl-substituted 1-azatrienes and the scope of the ring closures of these compounds are reported separately.³⁸ The predicted stereochemical outcome is observed experimentally; however, the ring closure of 1-azatriene 10 is reversible at 130°C, and over time the diastereoselectivity of the reaction comes under thermodynamic control. Here we report the origins of the selectivity of the electrocyclization of 10, using azatriene 12 as a computational model substrate.

Scheme 4.

The electrocyclization of TBPDS-substituted 1-azatriene 10.³¹

Just as is the case for the electrocycliczation of 5c, the anti transition state conformers of the ring closure of 12 are lowest in energy. Hyperconjugative stabilization by interaction of the C–Si σ donor orbital with the antibonding σ orbital of the forming C–N bond is likely reinforced by the steric preference of the silyl group for the anti position. Computations of the electrocyclic reaction of 12 are consistent with the experimental stereoselectivity; the computed kinetic product 13b (ΔΔG^‡ = 2.0 kcal mol⁻¹) corresponds (in terms of relative stereochemistry) to dihydropyridine 11a.³¹ The kinetic diastereoselectivity of the azaelectrocyclization of 12 is also controlled by a steric effect. In TS12_anti1, which leads to the minor product, the methyl substituent at the α position must adopt the outside position in order for the TMS group to occupy the anti position. In doing so, this methyl group clashes with the N-substituent. This clash is avoided in the favored transition structure TS12_anti2 as the methyl and TMS groups are able to simultaneously occupy the less sterically demanding inside and anti positions, respectively.

The equilibration of the two diastereomeric electrocyclization products 13a and 13b is observed experimentally, although the major product is still 13b. The computed t_½ at 130 °C for the ring opening (retroelectrocyclization) of either electrocyclization product is approximately 420 hr, too large a value considering experimental equilibration of the diastereomeric products of electrocyclization occurs within 24 hrs.³⁸ The bulky TBDPS group employed experimentally likely accelerates equilibration by destabilizing the products of the electrocyclic ring closure. In any case, the relative energy differences between two the dihydropyridine products indicate that the thermodynamic product is also the kinetic product 13a.

Conclusions

Transition state gauche effects explain the observed stereoselectivities of the electrocyclizations of chiral 1-azatrienes bearing α-alkoxy substitutents. Computational experiments predict that an allylic fluorine substituent at the at the allylic position reproduces the experimentally observed steroselectivity, while a silyl group at the same position changes the preferred transition state conformation and reverses the stereochemical course of the ring closure. The electrocyclization of TBPDS-substituted azatriene 12 confirms this predicted reversal of selectivity. Studies of the generality of the stereochemical influence of α-silyl groups and the synthetic utility of the ring closures of α-silyl azatrienes have been reported seperately.³⁸

Computational Methods

The DFT computations were performed using Gaussian09.³⁹ The M06-2X⁴⁰/6–31+G(d,p) model chemistry was used to optimize the geometries of stationary points and to compute the vibrational frequencies. M06-2X has been demonstrated to reliably reproduce the thermodynamics of π to σ transformations with reasonable accuracy⁴¹ in addition to the kinetics of main group chemistry.⁴² All computations were performed in the gas phase. Normal mode analysis confirmed that the optimized structures of reactants and products were minima (zero imaginary frequencies) and that all optimized transition structures corresponded to first-order saddle points (one imaginary frequency). Thermal corrections were calculated from unscaled M06-2X/6–31+G(d,p) frequencies using the standard state conditions of 1 atm and 298.15 K. Errors in the calculation of vibrational entropy contributions to the free energies by the treatment of low modes as harmonic vibrations was mitigated by raising the frequencies of vibrational modes less than 100 cm⁻¹ to exactly 100 cm⁻¹ as suggested by Truhlar.^43,44 Electronic energies were recomputed at the M06-2X/def2-QZVPP^45,46 level of theory. Reported Gibb free energies were determined using these electronic energies and thermal corrections determined at the M06-2X/6–31+G(d,p) level. The Avogadro^47,48 and Gaussview⁴⁹ software packages were used to build and visualize molecules; the images used in this article were prepared with CYLview.⁵⁰ All energies reported are Gibbs free energies in kcal mol⁻¹.

Figure 6.

M06-2X/6–31+G(d,p)-optimized transition structures, ΔG^‡, and ΔG_rxn for the electrocyclization of 12. Reported energies are in kcal mol⁻¹.