Effect of sulfide structure on enantioselectivity in catalytic asymmetric epoxidation of aldehydes: Mechanistic insights and implications

Abstract

Bridged bicyclic sulfide 1 was originally found to provide high levels of asymmetric induction in sulfur ylide-mediated epoxidations. This sulfide possesses chirality in the [2.2.1] thioether moiety and the [2.2.1] camphor-derived carbocyclic moiety. To determine whether the optimal sulfide had been used, a diastereomer of sulfide 1 in which the stereochemistry of the [2.2.1] carbocycle was reversed (sulfide 5) was prepared and studied as an epoxidation catalyst. This diastereomer gave considerably lower levels of asymmetric induction than the original sulfide 1. From computational and x-ray studies it was found that sulfide 5 gave rise to a more hindered ylide, which reacted more reversibly with aldehydes leading to lower enantioselectivity. Conditions that reduced reversibility were also tested and high enantioselectivities were returned for sulfide 5 (similar to sulfide 1). The implications for the synthesis of chiral sulfides for asymmetric epoxidations are discussed.

The reaction of sulfur ylides with carbonyl compounds to give epoxides (1–8) provides a complementary method with oxidation of alkenes (9–12) for the preparation of these valuable synthetic intermediates. Although sulfur ylide reactions have traditionally operated with stoichiometric amounts of sulfonium salts, we have recently reported a user-friendly, catalytic, and asymmetric process for epoxidation of carbonyl compounds that operated under neutral conditions by using tosylhydrazone sodium salts as diazocompound precursors and substoichiometric amounts (5 mol% in many cases) of chiral sulfide 1 and metal catalyst (Scheme 1) (13–16). This chiral sulfide could be recovered in high yield and reused without any loss of asymmetric induction. A broad range of aldehydes, including aromatic, aliphatic, and certain α,β-unsaturated aldehydes, could be converted into the corresponding epoxides in generally high yield with high levels of enantioselectivity and diastereoselectivity (17). Sulfide 1 could also be used in a stoichiometric variant of this process, which allows access to epoxides that are difficult to form using the catalytic protocol (18).

Scheme 1.

Catalytic cycle for epoxidation of aldehydes and ketones with in situ generation of diazocompounds.

We have recently proposed a model to explain the high enantioselectivity in this system based on both experimental and computational data (Scheme 2) (19). The high enantioselectivity observed resulted from ylide 2A being strongly favored over 2B because of steric interactions between the ylide substituent and methylene unit of the [2.2.1] bicycle, and the high face selectivity in reaction of ylide conformer 2A was due to the camphor moiety effectively blocking attack from the si face. Based on this model, the camphor moiety was simply acting as a sterically large blocking group. But was it? If that was simply its role, similar levels of enantioselectivity should result from the diastereomer 5 where the carbonyl group is placed on the right of the [2.2.1] carbocycle (Fig. 1).‡ In contrast, if the current model is too simplistic and the carbonyl group of the camphor skeleton does play a subtle role in control of enantioselectivity, then one would expect to observe variation in enantioselectivity for substrates 1 and 5. Whereas 1 and 5 could represent matched and mismatched substrates (or vice versa, and so we did not know if we had discovered the optimum chiral sulfide), by removal of the carbonyl group of the camphor moiety we would have the opportunity to study a stereochemically neutral substrate 6 because the group α-to sulfur would be achiral. Based on our model we did not expect much variation in enantioselectivity, but in testing these sulfides we were surprised at the outcome. Although the model was largely correct, variation in enantioselectivity resulted from an unexpected source.

Scheme 2.

Origin of enantioselectivity for epoxidation with chiral sulfide 1.

Fig. 1.

Chiral sulfides differing in chirality at camphor for epoxidation.

Methods

For preparations of tosyl hydrazone salts and sulfonium salts, see Supporting Appendix 1, which is published as supporting information on the PNAS web site.

General Procedure for the Epoxidation of Aldehydes with Tosyl Hydrazone Sodium Salts (14, 17). Tetrahydrothiophene (20 mol%), anhydrous acetonitrile (1.0 ml), rhodium(II) acetate dimer (1.5 mg, 1 mol%), benzyl triethylammonium chloride (15 mg, 20 mol%), aldehyde (0.33 mmol, 1.0 eq), and tosylhydrazone sodium salt (0.50 mmol, 1.5 eq) were added sequentially to a 5-ml round-bottomed flask fitted with a nitrogen balloon. The reaction mixture was stirred vigorously at room temperature for 10 min and was then stirred at 40°C for 24 h. The reaction was quenched by the addition of water (1.0 ml) and EtOAc (1.0 ml). The aqueous layer was extracted with EtOAc (2 × 1.0 ml), the combined organic phase was dried over MgSO₄ and filtered, and the solvent was removed under reduced pressure. The crude product was analyzed by ¹H NMR to determine the diastereomeric ratio and then purified by flash-column chromatography to afford the corresponding epoxide.

Epoxidation of Sulfonium Salt 22 and PhCHO Under Protic Solvent Conditions (18). To a stirred solution of the sulfonium salt (31 mg, 0.07 mmol) in a 9:1 acetonitrile/water mixture (0.4 ml) at room temperature was added the freshly distilled benzaldehyde (8 μl, 0.07 mmol) followed by powdered KOH (7 mg, 0.125 mmol). The reaction was then allowed to stir at room temperature for 2 h. Water (2 ml) was added, the aqueous phase was extracted with dichloromethane (3 × 5 ml), and the combined organic layers were washed with water (2 ml), dried over MgSO₄, and concentrated in vacuo. The residual oil was purified by silica gel chromatography (5% EtOAc in petroleum ether) to give stilbene oxide (11.8 mg, 83%, 91:9 trans/cis, 95% ee).

Epoxidation of Sulfonium Salt 23 and PhCHO at Low Temperature (18). To a solution of sulfonium salt 23 (46 mg, 0.11 mmol) in anhydrous CH₂Cl₂ (0.4 ml) was added 1 eq of N,N,N′,N′-tetramethyl-N″-(tris(dimethylamino)phosphoranylidene)-phosphoric triamide ethylimine at –78°C. After stirring for 10–15 min, benzaldehyde (11 μl, 0.11 mmol) was added drop-wise, and the reaction was stirred for 1 h at –78°C. After addition of a saturated solution of NaCl in water (1 ml), the organic phase was separated, and the aqueous phase was extracted with dichloromethane (3 × 5 ml). The organic phases were combined, filtered over MgSO₄, and concentrated in vacuo, and the residual oil was purified by silica gel chromatography (5% EtOAc in petroleum ether) to give stilbene oxide (21 mg, 100%, >98:2 trans/cis, 98.5% ee).

Results and Discussion

Sulfide Synthesis. Sulfide 1 was easily prepared in four high yielding steps from camphor sulfonyl chloride (14) (Scheme 3). The first step of the synthesis was the formation of thiol 7 by reduction of camphor sulfonyl chloride with triphenylphosphine. Alkylation of the thiol with phenacyl chloride in the presence of potassium carbonate afforded phenyl ketone 8. Irradiation of a cooled solution of 8 with a sun lamp in the presence of an excess of cyclopentadiene furnished only the endo-cycloadduct 10 with high diastereoselectivity (20). Finally, hydrogenation of the carbon–carbon double bond with Pd S/C catalyst gave sulfide 1 in 48% overall yield.

Scheme 3.

Synthesis of chiral sulfide 1. Reagents and conditions: (i) PPh₃ (4 eq), 1,4-dioxane/H₂ O (4:1), reflux, 1 h, 90%; (ii) PhCOCH₂ Cl (1.1 eq), K₂ CO₃ (5 eq), tetrahydrofuran, reflux, 20 h, 82%; (iii) sun lamp (hν), CH₂ Cl₂, 20°C, 16 h, cyclopentadiene (20 eq), 70%; (iv) H₂, Pd-S/C, EtOH, room temperature, 3 h, 83%.

Vedejs and coworkers proposed that the hetero Diels–Alder reactions of thioaldehydes, such as 9, with cyclopentadiene proceed by way of an early transition state in which relatively little rehybridization occurs and that steric effects are primarily responsible for the high endo selectivity (20). The cyclopentadiene ring is nearly flat and, of the two possible approaches, the endo approach has the less demanding interaction of the camphor moiety with the alkene portion of the diene, whereas the exo approach results in steric interaction between the camphor moiety and the methylene group (Scheme 4).

Scheme 4.

Endo/exo selectivity in thio Diels–Alder reaction.

The facial selectivity of the Diels–Alder reaction can be explained by the preference for conformation 11b over 11a, the latter being destabilized by electronic repulsion between the oxygen and sulfur lone pairs (Scheme 5). The cyclopentadiene approaches from the less hindered si face of the thioaldehyde in conformation 11b.

Scheme 5.

Facial selectivity in thio Diels–Alder reaction.

Sulfide 6 was prepared by Wolff–Kishner reduction of the keto group of sulfide 1 in moderate yield (Scheme 6).

Scheme 6.

Wolff–Kishner reduction of sulfide 1. Reagents and conditions: (i) NH₂ NH₂ ·H₂ O, KOH, tri(ethylene glycol), reflux, 3 h, 30%.

To obtain sulfide 5, the opposite facial selectivity in the photochemical thio Diels–Alder reaction was required. We believed that this selectivity could be achieved through stabilization of the previously disfavored conformer 11b by metal chelation. Furthermore, it would allow both sulfide diastereomers 1 and 5 to be prepared from the same precursor in a direct and simple manner. To test the viability of such a procedure, we performed the photolysis/thio Diels–Alder reaction of 8 in the presence of a variety of Lewis acids. However, rather than a reversal of diastereoselectivity, a small reduction in diastereoselectivity was observed together with a substantial decrease in yield.§

As an alternative to metal chelation, we considered the possibility of using hydrogen bonding to control the conformation of the thioaldehyde (13a, Scheme 7). Indeed, it had previously been shown that hydroxy groups α- to thioaldehydes gave opposite diastereomers to substrates bearing α-alkoxy groups (20). This strategy was finally successful.

Scheme 7.

Facial selectivity in the thio Diels–Alder reaction of hydroxythioaldehydes 16.

Alcohol 14 was readily prepared in three steps from camphor sulfonyl chloride (Scheme 8). Reduction with lithium aluminum hydride gave hydroxythiol 16 as a 5:1 mixture of the endo and exo diastereomers, which were separated by column chromatography (21). Alkylation of the endo-hydroxythiol with phenacyl chloride under basic conditions afforded phenacyl sulfide 17, which was the precursor of thioaldehyde 13. Photolysis in the presence of excess cyclopentadiene at 20°C gave two endo cycloadducts 14 and 15 in 71% combined yield and 2.7:1 ratio. Conducting the photochemistry at –50°C gave a similar yield of the cycloadduct but with improved diastereoselectivity (5.4:1; Table 1). Oxidation of the alcohol with chromium trioxide and hydrogenation of the alkene by using Pd-S/C catalyst afforded 5 in 47% yield over two steps.

Scheme 8.

Synthesis of sulfide diastereomer 6. Reagents and conditions: (i) LiAlH₄, Et₂ O, 0°C for 3 h, then reflux for 24 h, 59%; (ii) phenacyl chloride, K₂ CO₃, tetrahydrofuran, 20 h, reflux, 74%;(iii) cyclopentadiene (20 eq), hν, CH₂Cl₂, –50°C, 6 h, 69% (14:15 = 5.4:1); (iv) CrO₃/Py, 0°C → room temperature, 24 h, 50%; (v)H₂, Pd-S/C, room temperature, 3 h, 94%.

Table 1. Diastereoselectivity for Diels—Alder reactions of 13 at low temperature.

Entry	Temperature, °C	Yield, %	15:16
1	20	71	2.7:1
2	-30	48	4.0:1
3	-50	69	5.4:1

Asymmetric Epoxidation by Using Sulfides 1, 5, and 6. Sulfides 1, 5, and 6 were tested as catalysts for the epoxidation process in the standard test reactions (Table 2).¶ Sulfide 1 gave the highest enantioselectivity, whereas the camphor diastereomer 5 gave the lowest enantioselectivity. The substrate bearing the achiral substituent (sulfide 6) gave intermediate selectivity. Thus, our study showed that the stereochemistry of the camphor moiety did have an impact on the enantioselectivity of the epoxidation process and that this group was not simply a large blocking group. It also showed that the stereochemistry of the [2.2.1] bicyclic sulfide moiety and the stereochemistry of the camphor moiety in 1 are matched, whereas in sulfide 5 they are mismatched, and sulfide 6 represents a neutral system (Fig. 2). But how does the stereochemistry of the camphor moiety influence the enantioselectivity of the epoxidation process?

Table 2.

Yield, enantioselectivity, and diastereoselectivity of stilbene oxide from benzaldehyde and benzaldehyde tosylhydrazone salt by using sulfides 1, 5, and 6

Entry	Sulfide	Yield, %*	Trans/cis†	ee %‡
1	1	87	>98:2	94 (R, R)
2	6	80	>98:2	86 (R, R)
3	5	83	>98:2	78 (R, R)

^*Yield of isolated product.

^†Determined by ¹H NMR spectroscopy of crude reaction mixture.

^‡Determined by chiral HPLC with a Chiralcel OD column.

Fig. 2.

Stereochemically matched, neutral, and mismatched sulfide diastereomers.

Four factors influence the enantioselectivity for epoxidation of carbonyl compounds by using semistabilized sulfonium ylides; to achieve high enantioselectivity it is necessary to (i) form a single diastereomeric sulfonium ylide, (ii) achieve high levels of control in ylide conformation, (iii) achieve high levels of control in face selectivity of the ylide, and (iv) ensure that anti-betaine formation is nonreversible (19). These four criteria are largely met with sulfide 1, which gave 94% ee. However, one of these criteria is no longer met with diastereomeric sulfide 5. It was potentially possible that ylide conformation (criterion ii) was responsible for the reduced enantioselectivity. However, quantum mechanical calculations on ylides 2 and 18 indicated that the ratio of conformers A:B should be similar in both cases [18.3 kJ/mol vs. 17.7 kJ/mol [calculated at the B3LYP/6–311G**+(CH₃CN)//6–31G* (CH₃CN) level by using jaguar (22)] (Scheme 9), and therefore based on our simple model the enantioselectivity of these reactions should be similar. It might be argued that the calculated energy differences we are discussing are of the same magnitude as the error expected from the B3LYP method (23) and that we should therefore not attach too much value to these numbers. However, we are comparing two diastereoisomers, i.e., species with the same chemical bonding, and we can therefore expect very favorable error cancellation, such that the computed differences should be at least semiquantitatively reliable.

Scheme 9.

Calculated energy differences for ylide conformations (B3LYP/6–311G**+(CH₃CN)//6–31G* (CH₃CN).

Although the calculations did not show any significant energy differences between the ylide conformations of the two diastereomers, they did reveal an interesting conformational difference between the two ylides in another part of the molecule, which provided a clue to the origin of the difference in enantioselectivity. In their lowest energy conformations, the camphor moiety of 18 lies in a position such that the gem-dimethyl bridge is orientated toward the reacting ylide center, whereas in ylide 2 the ethylene bridge occupies this position. Presumably the orientation of the camphor group is controlled by a combination of two factors: (i) minimization of nonbonded steric interactions and (ii) minimization of molecular dipole moment. The x-ray crystal structure of the parent sulfonium salts also showed the same conformational preference for the two diastereomers (Figs. 3 and 4 and Supporting Appendices 2 and 3, which are published as supporting information on the PNAS web site).

Fig. 3.

Calculated geometries from ylides 2A and 18A [B3LYP/6–31G* (CH₃CN)].

Fig. 4.

X-ray crystal structures of benzyl sulfonium salts derived from sulfides 1 and 5. (Counterion and other sulfonium species in the unit cell were removed for clarity.)

Although these differences in conformation do not change the face selectivity in the ylide reaction [which is believed to be complete in both cases (criterion iii)], they do create different steric environments around the ylide. On the basis of a large amount of experimental, literature, and computational data, it is known that sterically more hindered sulfur ylides react more reversibly with aldehydes because of their higher barriers to torsional rotation of the intermediate cisoid betaine (19). For this reason 18 could react more reversibly (in betaine formation) than 2 and, as a result, suffer diminished enantioselectivity (criterion iv). This outcome is best explained by considering the reaction of the two ylide conformers with the aldehyde to give the intermediate betaines and the degree of reversibility of the two processes (Scheme 10). Reaction of the minor conformer of the ylide 18B is likely to be less reversible than the major conformer 18A, because it is less stable and, thus, it is less likely for betaine 20 to revert to starting materials. As such, a greater proportion of product will inevitably be derived from the less reversible pathway. Thus, partial reversibility in betaine conformation results in overexpression of the less favored ylide conformer and therefore reduced enantioselectivity.

Scheme 10.

Rationalization of origin of reduced enantioselectivity with sulfide 5.

To test the hypothesis that the ylide derived from sulfide 5 (ylide 18) was reacting more reversibly than the ylide derived from sulfide 1 (ylide 2), we needed conditions where betaine formation is known to be less reversible, because this should result in the enantioselectivity better reflecting the high ratio of ylide conformers. Betaine formation can be rendered less reversible by either increasing the barrier to reversion to ylide and aldehyde or by reducing the barrier to bond rotation of the cisoid conformer to the transoid conformer. It had been established that the use of protic solvents or metal ions solvated the charges in the intermediate betaine and made their separation, which occurs during bond rotation, more facile (17, 19). More facile separation results in reduced reversibility in betaine formation. Thus, reactions of the two ylides were performed in the presence of water, and the results are summarized in Table 3. Reactions were also performed at –78°C in aprotic media because it had been established that betaine formation was again less reversible at low temperature (19). Reactions were performed with sulfonium salts 21 and 22, because the yields of our catalytic chemistry are diminished by the presence of significant amounts of water (17).

Table 3.

Epoxidation with diastereomeric sulfonium salts under nonreversible conditions

Entry	Conditions	Yield, %*	Trans/cis†	ee %‡
1	21, KOH, CH3CN/H2O (9:1), rt	75	98:2	98 (R,R)
2	22, KOH, CH3CN/H2O (9:1), rt	83	91:9	95 (S,S)
3	22, EtP2, CH2Cl2, -78 °C§	100	>98:2	98.5 (S, S)

^*Yield of isolated material.

^†Determined by ¹H NMR of crude reaction mixture.

^‡Determine by chiral HPLC on Chiracel OD column.

^§EtP₂ = N,N,N′,N′-tetramethyl-N″ -(tris(dimethylamino)phosphoranylidene)-phosphoric triamide ethylimine.

As predicted, under less reversible conditions, the enantioselectivity of the reaction of sulfonium salt 22 was significantly improved and much closer to the enantioselectivity observed with salt 21 (compare entry 2 with entry 1) and higher than that observed in the catalytic process for both sulfides 1 and 5 (95% ee compared with 94% ee for sulfide 1 and 78% ee for sulfide 5), thus implicating the reversion of betaine 19 to starting materials as the cause for the lower levels of enantioselectivity observed with sulfide 5 (criterion iv).

Conclusion

Four factors influence the enantioselectivity in reactions of semistabilized sulfonium ylides with aldehydes. Although it was expected that the four factors would be well controlled by sulfide 5 because of its structural similarity to 1, which appears to meet these criteria, the reduced enantioselectivity observed showed that this was not the case. In sulfide 5, it is believed that criterion iv (nonreversible formation of anti-betaine) was not completely fulfilled, probably because the preferred conformer of the ylide placed the bridging gem-dimethyl group in close proximity to the ylide carbon. This made the ylide more hindered, which in turn made betaine formation more reversible. Criterion iv could be met by adopting conditions (protic solvent) that made betaine formation less reversible, and this resulted in high enantioselectivity for both 1 and 5. This study shows how small changes in sulfide structure can have a significant effect on enantioselectivity, because the degree of reversibility in betaine formation (criterion iv) is particularly sensitive to the steric encumberance of the ylide. To ensure that criterion iv is satisfied, the reactions are simply carried out in protic media where reversibility in betaine formation is largely eliminated. This study highlights the importance of criterion iv for controlling the enantioselectivity and provides a simple test to determine whether this criterion is being fulfilled.