Catalytic enantioselective intermolecular cycloadditions of 2-diazo-3,6-diketoester-derived carbonyl ylides with alkene dipolarophiles

Abstract

Catalyzed cascade reactions that generate molecular complexity rapidly and in an enantioselective manner are attractive methods for asymmetric synthesis. In the present article, chiral rhodium catalysts are shown to effect such a transformation by using a range of 2-diazo-3,6-diketoesters with bicyclo[2.2.1]alkenes and styrenes as reaction partners. The reactions are likely to proceed by formation of a catalyst-complexed carbonyl ylide from the diazo compound, followed by intermolecular cycloaddition with the alkene dipolarophile. It was possible to obtain high levels of asymmetric induction [up to 89% enantiomeric excess (ee) and 92% ee for the two chiral catalysts investigated]. Enantioselectivity is not highly sensitive to substituent variation at the ketone that forms the ylide; however, branching does improve ee. Observations of dipolarophile-dependent enantiofacial selectivity in the cycloadditions indicate that the dipolarophile can be intimately involved in the enantiodiscrimination process.

Some of the best ways of preparing five-membered heterocycles are provided by 1,3-dipolar cycloadditions (1, 2). In recent years, encouraging progress has been made in the catalytic asymmetric synthesis of such heterocycles by using this pericyclic process (3). Carbonyl ylides 1 constitute an important class of 1,3-dipoles that, after cycloaddition with alkynes or alkenes, for example, result in reduced furans 2 (Scheme 1).

Scheme 1.

We have been investigating an enantioselective version of carbonyl ylide cycloaddition (ref. 4 and references therein). This chemistry builds on pioneering research by Ibata and especially Padwa, who established that transition metal-catalyzed decomposition of diazo compounds in the presence of carbonyl functionality generates transient carbonyl ylides that subsequently undergo cycloaddition in the presence of suitable dipolarophiles (5, 6). In our previous studies, we demonstrated that enantioinduction is possible in tandem intramolecular carbonyl ylide formation, intramolecular cycloaddition processes by reaction of unsaturated diazodiketoesters 3 with chiral, nonracemic, rhodium catalysts (e.g., Scheme 2). The best levels of enantiomeric excess (ee) for the cycloadducts 7 were obtained in hexane by using the hydrocarbon-soluble catalysts tetrakis[(S)-N-(4-dodecylphenylsulfonyl)prolinate] dirhodium(II) {Rh₂[(S)-DOSP]₄} 4 (up to 69% ee) and tetrakis[(R)-6,6′-didodecylbinaphtholphosphate] dirhodium(II) {Rh₂[(R)-DDBNP]₄} 5 (up to 90% ee).

Scheme 2.

Several rhodium catalysts are now known to be capable of effecting efficient asymmetric cyclopropanation and C—H insertion by using diazo compounds, in which an intermediate metal–carbene complex is directly involved in the enantiodiscrimination step (7–11). Asymmetric ylide chemistry using diazo compounds places a rather different additional demand on the catalyst (12). Because the catalyst-free ylide is achiral and catalyst association with the simple unpolarized tethered dipolarophiles used in our studies is unlikely, it is reasonable to assume that enantioselectivity requires a catalyst-associated ylide (e.g., 6), formed from the metal–carbene complex, to influence facial selectivity in the cycloaddition. Lower enantioselectivity will be observed if the catalyst (reversibly) dissociates from the ylide and cycloaddition occurs (competitively) from the achiral ylide. Energy is necessary to effect catalyst dissociation, however (13), and part of that energy could be supplied by that produced in bond formation to the dipolarophile. Cycloaddition then would occur on the opposite face of the ylide to the catalyst as the catalyst dissociates. Recent computational studies indicate that in the metal–carbene complex, the Rh—C σ-bond and the adjacent C—O π-bonds tend toward alignment and thus have partial metal enolate-type interactions (14–16). This latter work suggests that a transition state for cyclization to form the metal-complexed ylide might resemble that shown in Fig. 1. In this scenario, enantioselectivity is influenced by which lobe of the vacant 2p orbital at the carbenic carbon the tethered ketone preferentially cyclizes to, under the biasing effect of the chiral information X present in the ligands.

Fig. 1.

Suggested transition state for cyclization to form a metal-complexed carbonyl ylide.

Our previous asymmetric intramolecular cycloaddition studies (Scheme 2) encouraged us to examine intermolecular cycloadditions of dicarbonyl-substituted carbonyl ylides with dipolarophiles that do not contain electron-withdrawing substituents on the reacting π-bond. Such ylides constitute one of the most widely used classes of carbonyl ylides (5, 6) because of the typical stability, storage, and ease of handling of the cycloaddition precursors, readily prepared from 1,3-dicarbonyl compounds by diazo transfer. Before our work, only dimethyl acetylenedicarboxylate as the dipolarophile had been shown capable of efficiently delivering carbonyl ylide-derived intermolecular cycloadducts in high ees: up to 92% ee was observed by Hashimoto and coworkers (17) using 1-diazo-2,5-diketones. Enantioselective carbonyl ylide-type cycloadditions of (aromatic) oxidopyryliums have been reported also (18, 19).

This article reports studies that expand the range of carbonyl ylide substrates and dipolarophiles that are capable of undergoing the asymmetric intermolecular cycloaddition process and provide insight into important factors that influence efficiency and selectivity in this chemistry. A small part of this work has appeared in a preliminary communication (20).

Experimental Procedures

General Procedures. Full experimental details of syntheses, characterization of cycloaddition substrates and cycloadducts, and ee determinations are shown in Supporting Text, which is published on the PNAS web site.

General Cycloaddition Procedure. To a stirred, degassed solution of the diazo compound (0.3 mmol) and dipolarophile (10 eq) in solvent (4.7 ml) was added an Rh^II catalyst (1 mol%). When the reaction was complete (as monitored by TLC), the reaction mixture was concentrated in vacuo, and the residue was purified by column chromatography.

Results and Discussion

Cycloaddition Substrate Synthesis. Examination of Fig. 1 suggests that the nature of R and its interaction with the chiral ligands of the catalyst could influence asymmetric induction. To probe this aspect in the intermolecular cycloaddition process, a range of 2-diazo-3,6-diketoester cycloaddition substrates 10a–10e were prepared from 4-ketoacids 8a–8e by Masamune homologation (21), followed by diazo transfer to the resulting 3,6-diketoesters 9a–9e (Scheme 3).

Scheme 3.

(i) Carbonyldiimidazole, THF, then Mg(O₂CCH₂CO₂-tert-butyl)₂. (ii) 4-AcNHC₆H₄SO₂N₃, Et₃N, MeCN.

Cycloadditions with Bicyclo[2.2.1]alkenes. It is usually found that simple alkenes are not reactive dipolarophiles in cycloadditions with carbonyl-stabilized carbonyl ylides (22). For example, no cycloadduct was observed from the reaction of ester 10a with vinylcyclohexane (10 eq) as dipolarophile under Rh₂(OAc)₄ catalysis in CH₂Cl₂ at 25°C. Cyclopentene, however, under similar conditions, allowed isolation of a 14% yield of cycloadduct after 4 h (see Supporting Text), although no cycloadduct was observed when using Rh₂[(R)-DDBNP]₄ 5. There is precedent for the use of norbornenyl systems as dipolarophiles (23, 24), and we found that with substrate 10a under Rh₂(OAc)₄ catalysis in CH₂Cl₂ in the presence of 1.5 eq of norbornene gave a 45% yield of a single cycloadduct 11a after 1 h (Scheme 4). The structure of the cycloadduct was confirmed by x-ray crystallographic analysis and is consistent with the selectivity observed in other carbonyl ylide cycloadditions with norbornene (23, 24). The yield of cycloadduct 11a could be improved to 82% by using 10 eq of norbornene, the reaction being complete after 30 min. By using Rh₂[(S)-DOSP]₄ 4, a reasonable ee and high yield were obtained (61% ee and 83% yield); Rh₂[(R)-DDBNP]₄ 5 at room temperature improved on this result, giving an ee of 82% and a 74% yield. Reducing the temperature to 0°C and then to –15°C gave additional increases in ee to 87% and 92%, respectively, providing a high asymmetric induction.

Scheme 4.

Moving from methyl ketone 10a to n-hexyl ketone 10b did not significantly affect the yield of cycloadduct 11b with Rh₂(OAc)₄ (76%) or the ee with either of the chiral catalysts (Scheme 4). Using the phenyl ketone 10c allowed an examination of the presence of a conjugated substituent on the ylide, which might be anticipated to exert a more significant effect than alkyl substitution. In that event, cycloaddition using 10c proceeded in slightly lower yields, compared with 10a–10b, to give cycloadduct 11c [53% using Rh₂(OAc)₄], the structure of which was confirmed by crystallography. Comparing the effectiveness of the chiral catalysts in the syntheses of bicyclic ketones 11a–11c, a noticeable rise in ee to 77% was seen with phenyl ketone 10c using Rh₂[(S)-DOSP]₄ 4, whereas Rh₂[(R)-DDBNP]₄ 5 was less effective with 10c at ambient temperature (51% ee) but showed a steep rise in ee after lowering the reaction temperature to 0°C (80% ee); in parallel with earlier observations in intramolecular cycloadditions (25), with 10c there was no marked effect on ee with respect to temperature variation when using Rh₂[(S)-DOSP]₄ 4.

We also found norbornadiene to be viable as dipolarophile in cycloaddition reactions of the 2-diazo-3,6-diketoesters (Fig. 2); using substrate 10a under Rh₂(OAc)₄ catalysis in CH₂Cl₂ at 25°Cin the presence of 10 eq of norbornadiene gave cycloadduct 12a in 78% yield. Rh₂[(S)-DOSP]₄ 4 at 25°C allowed isolation of 12a in 65% yield with 79% ee; Rh₂[(R)-DDBNP]₄ 5 at 0°C gave 12a in 53% yield with 82% ee. The improvement in ee obtained by using norbornadiene in place of norbornene with Rh₂[(S)-DOSP]₄ 4 (79% ee compared with 61%) is intriguing and does not have an obvious explanation. In comparison, with Rh₂[(R)-DDBNP]₄ 5 the ee falls slightly from 87% to 83% after moving from norbornene to norbornadiene. Moving to the hexyl ketone substrate 10b gave similar results using Rh₂[(S)-DOSP]₄ 4 compared with methyl ketone 10a.Rh₂[(R)-DDBNP]₄ 6 proved to be more sensitive to the increased substitution at the ylide carbonyl: both yield and ee of 12b were reduced, to 35% and 69% respectively, at 0°C. The phenyl ketone substrate 10c underwent cycloaddition with norbornadiene in 30–64% yield to give cycloadduct 12c (structure confirmed by x-ray analysis). In this case, the enantioselectivity of the cycloaddition was reduced slightly compared with that when using norbornene as dipolarophile: with Rh₂[(S)-DOSP]₄ 4 at 20°C, the ee was reduced from 77% to 64%, and with Rh₂[(R)-DDBNP]₄ 6 at 0°C, the ee was reduced from 80% to 63%.

Fig. 2.

Cycloadducts 12a–12d obtained from 10a–10d by using norbornadiene and catalysts 4 and 5.

With norbornadiene (10 eq) as the dipolarophile, the effect on the cycloaddition chemistry of branching α- to the ketone functionality in the cycloaddition substrate was explored further by using isopropyl ketone 10d and tert-butyl ketone 10e. Less efficient cycloaddition was observed, with cycloadduct 12d (structure confirmed by x-ray crystallographic analysis) being obtained in only 25–44% yields and no cycloadduct being seen when using 10e. However, the asymmetric cycloadditions using isopropyl ketone 10d were accompanied by an ee improvement relative to the other cycloaddition substrates examined, and up to 89% ee was observed for 12d with both chiral catalysts. Increased interactions between R and the chiral ligand systems X (Fig. 1) when R is branched may be contributing to the reduced yield (or no yield when R = tert-butyl) as well as to the raised levels of asymmetric induction.

Methyl ketone 10a was examined also with 10 eq of benzonorbornadiene as dipolarophile (Scheme 5). Under Rh₂(OAc)₄ catalysis in CH₂Cl₂ at 20°C, cycloadduct 13 was obtained in 44% yield; this is somewhat lower than the corresponding yields seen with norbornene and norbornadiene (82% and 78%, respectively). The stereochemistry of cycloadduct 13 was assigned by analogy with the cycloadditions discussed above. Both chiral catalysts were more effective than Rh₂(OAc)₄ with 10a and benzonorbornadiene, giving similar yields at room temperature to those obtained with norbornene and norbornadiene. Using Rh₂[(S)-DOSP]₄ 4, a similar ee (74%) for 13 was obtained to that for 12a (79%) from 10a and norbornadiene. Intriguingly, however, with Rh₂[(R)-DDBNP]₄ 5,a significant lowering of asymmetric induction was seen (to 12%, compared with 68% ee when using norbornadiene), and the reaction proceeded in low yield at 0°C with no improvement in ee.

Scheme 5.

Cycloadditions with Styrenes. Aside from enantioselectivity, the use of an unstrained intermolecular dipolarophile such as styrene introduces issues of reactivity, regioselectivity, and exo/endo-stereoselectivity in the cycloaddition process. Indeed, Rh₂(OAc)₄-catalyzed tandem carbonyl ylide formation–cycloaddition of 1-diazo-5-phenyl-2,5-pentanedione with styrene proceeds rather unselectively, producing all four possible cycloadducts (22). Nevertheless, because the use of styrenes could provide additional insight into factors that influence the (asymmetric) cycloaddition, Rh₂(OAc)₄-catalyzed reaction of methyl ketone 10a with styrene (10 eq) in CH₂Cl₂ at 20°C was examined. Pleasingly, the reaction proceeded in 53% overall yield and in a regiocontrolled manner to give a separable epimeric mixture of 7-substituted bicyclic cycloadducts 14 (exo/endo, 2:1; Scheme 6, where exo refers to the aryl group being syn to the oxido bridge). Structural assignments are based on extensive spectroscopic (nuclear Overhauser effect) studies (see Supporting Text). Both chiral catalysts 4 and 5 also favored formation of exo-14 in up to 62% ee; these results suggest the possibility of using para-substituted styrenes to begin to probe the electronic effects of the alkene dipolarophile in the cycloaddition.

Scheme 6.

A comparison of styrene with 4-chlorostyrene and 4-methoxystyrene was carried out. Reaction of methyl ketone 10a with the two latter dipolarophiles under Rh₂(OAc)₄ catalysis gave similar yields of cycloadducts 15 (50%) and 16 (43%) and exo/endo-stereoselectivities (2:1) to those observed with styrene. The structure of exo-16 was confirmed by x-ray crystallographic analysis. With the chiral catalysts, 4-chlorostyrene also gave similar yields and levels of asymmetric induction to those seen with styrene (Scheme 6). 4-Methoxystyrene was found to be a less effective dipolarophile in terms of yield, but higher ee values were obtained with both catalysts. These latter results suggest a clear electronic effect on ee.

One surprising observation with all three styrenes is that catalysts 4 and 5 deliver opposite senses of asymmetric induction in the cycloadducts 14–16. Cycloaddition of 10a with phenylacetylene using catalysts 4 and 5 gave a similar result (Scheme 7), and no significant reduction in ee of cycloadduct 17 was found after conducting the reaction with 10a and 5 in neat phenylacetylene (86% yield and –57% ee).

Scheme 7.

The above observations (Schemes 6 and 7) on the sense of asymmetric induction with catalysts 4 and 5 contrast with all our previous enantioselective intramolecular and intermolecular cycloadditions using 2-diazo-3,6-diketoester cycloaddition substrates with alkene and alkyne dipolarophiles. The preferred absolute sense of asymmetric induction with catalysts 4 and 5 with substrate 3 (R = H), as well as with the corresponding tethered terminal acetylene, have been established (25) and is as shown in Scheme 2. The current results indicate that an unusual dependence on the dipolarophile for the predominant sense of asymmetric induction in the cycloadditions is exhibited by either catalyst 4 or 5 (efforts are underway to determine which) and suggests that the two types of catalyst induce asymmetry by different processes, at least with styrenes and phenylacetylene.

Conclusions

We have achieved highly enantioselective intermolecular carbonyl ylide cycloadditions of 2-diazo-3,6-diketoesters, with levels of ee correlating well with those observed in intramolecular cycloadditions with structurally related dipoles, indicating that a tethered dipolarophile is not necessary for high enantioselectivity. The sensitivity of the reaction to substitution at the ylide carbonyl has been explored, and it has been found that enantioselectivity is relatively insensitive to the alkyl substituent introduced. Nevertheless, branching does improve ee, albeit at the expense of yield, and no cycloaddition was observed with a tertiary alkyl substituent. An aryl substituent is accommodated also, with only a small reduction in ee compared with alkyl functionality. Finally, examples of dipolarophile-dependent enantiofacial selectivity in carbonyl ylide cycloadditions have been observed. These have potentially important implications concerning the origin(s) of asymmetric induction in these processes.