Stereospecific cis- and trans-ring fusions arising from common intermediates

Significance

The development of strategies in organic synthesis is an important challenge, from the perspective of creative design as well as application to the pharmaceutical industry. In this paper, we study a promising new solution to a classical problem in organic synthesis: namely, the control of ring junction stereochemistry in a particularly challenging setting, wherein one of the rings composing the junction is four membered. This system was selected because it poses a particularly difficult challenge for the synthesis of trans-fused junctions. We believe that the solution to the problem described herein will have broad applications in organic synthesis.

Abstract

Highly concise and stereospecific routes to cis and trans fusion, carrying various functionality at one of the bridgehead carbons, have been accomplished.

Our group has been studying the synthetic utility of the Diels–Alder (DA) reaction of the parent cyclobutenone, 2 (1). Recently reported results demonstrate that 2 is a highly reactive, endo-selective dienophile (Fig. 1, Eq. 1) (2). We have also developed a series of intramolecular Diels–Alder (IMDA) reactions, wherein the cyclobutenone component is tethered to various conjugated dienes (compare 4); cycloaddition of these substrates delivers adducts of the type 5, which can be readily converted to trans-fused systems bearing iso-DA patterns (Fig. 1, Eq. 2, 5→6) (3). Additionally, we have described the synthesis and DA cycloaddition of an even more powerful dienophile, 2-bromocyclobutenone (Fig. 1, Eq. 3, 7) (4). The direct adducts of this [4+2] reaction (compare 8) are readily converted to norcarane carboxylic acids (9) through exposure to hydroxide base. The research described herein was initially focused on efforts to add carbon-based nucleophiles to DA cycloadducts of the type 8. It might well have been expected that such reactions would give rise to products such as 10, wherein a ketone is appended to the junction of the norcarane system (Fig. 1, Eq. 4).

Fig. 1.

Expanding the scope of the Diels–Alder reaction.

Results and Discussion

With a view toward accomplishing the transformation envisioned in Fig. 1, Eq. 4, compounds of the type 8 were exposed to methyllithium as described in Fig. 2. Interestingly, exposure of substrate 8a to the conditions shown led to an 81% yield of the angularly methylated product 11a, apparently unaccompanied by any part of the corresponding trans-fused diastereomer (Fig. 2, entry 1). Reactions conducted with related DA adducts 8b and 8c resulted in quite similar outcomes (Fig. 2, entries 2 and 3). Although we were confident that the products obtained were cis-fused (Fig. 2, 11a–c), it was the synthesis of the rigorously assignable trans-fused product (vide infra) and its nonidentity with 11a that served to allow for sound assignment of the cis-ring fusion.

Fig. 2.

Reaction of DA adducts with methyllithium.

An obvious mechanistic proposal to account for the formation of this structure envisions lithium–halogen exchange (5–8) leading to 12a and methyl bromide (Fig. 3). This step is followed by angular methylation (9–11) of 12a. It would not be surprising for such a reaction to have a very high preference for producing cis fusion. Clearly, a great deal of ring strain would be introduced in any transition state en route to trans-fused products.

Fig. 3.

Mechanistic proposals.

Whereas the enolate alkylation proposal is presumptive, several alternate formulations merited consideration. In principle, one can envision the possibility that lithium enolate 12a is not actually produced (perhaps due to ring strain). Rather, 11 (Fig. 3) arises from the merger of otherwise high-energy species en route to 12a and 12b. For instance, “radicaloid” (5) character could be generated at both the bridgehead (12b) and methyl carbon centers. Coupling of these species, before their disaggregation, would accomplish the observed overall angular methylation, culminating in 11. Exclusive formation of a cis-ring fusion would again be expected, in terms of a solvent-cage−induced “memory effect,” as well as of the ring strain associated with the transition state leading to a trans junction (for reviews on the concept of “chiral memory,” see refs. 12 and 13). Here, we present only the limiting possibility, i.e., coupling of “enolyl” and “methyl” free radicals. The important distinction between pathways i and ii in Fig. 3 is that the former corresponds to a classic enolate methylation. By contrast, in pathway ii the coupling event occurs somewhere en route to metal–halogen exchange.

Because the cis-specific outcome of the reaction did not serve to answer the mechanistic question, a more illuminating probe would be required. The thought was that, if the reaction were occurring through path i, with the alkylating agent corresponding to methyl bromide, then one could anticipate potential competition between the MeBr generated through metal–halogen exchange and incident MeBr in solution. By contrast, path ii dictates that the angular methyl group in 11 must arise only from the original methyllithium. To probe this distinction in practice, it would be necessary to fashion an experiment that would allow us to discern between the methyl group of solution-based methyl bromide and other species capable of methyl transfer that arise during progression of the metal–halogen exchange.

As shown in Fig. 4, Eq. 5, reaction of 8a with CD₃Li was conducted for 15 min at –78 °C. Following this, CH₃Br was added to the reaction mixture. Deuterated compound 13 was obtained in 64% yield, along with 16% of 11a. For control purposes, the experiment was also conducted in the opposite sequence, wherein CH₃Li was used in the metal–halogen exchange, and CD₃Br was subsequently added to the reaction mixture (Fig. 4, Eq. 6). In this mode, 11a was obtained in 59% yield, along with a 14% yield of 13. These observations suggest substantial angular methylation in the initial 15-min interval. However, because the reaction was not complete after 15 min, the remaining methyl bromide (CD₃Br or CH₃Br) was forced to compete with the newly introduced alkylating agent (CH₃Br or CD₃Br) in a 1:1 ratio.

Fig. 4.

Mechanistic evaluation of methylation.

Whereas these experiments are certainly consistent with enolate formation followed by conventional alkylation, they did not fully document the source of the angular methyl group before introduction of the external methylating agent (CD₃Br or CH₃Br). In a more discriminating experiment, CD₃Li was added to a solution already containing an equivalent of both CH₃Br and 8a (Fig. 4, Eq. 7). This reaction gave rise to 25% of recovered 8a as well as ca. 20% each of 11a and 13. (The somewhat reduced yield of angular methylation in this experiment presumably reflects some degree of competitive reaction between methyllithium and methyl bromide.) Thus, at least under these conditions, incident CH₃Br is fully competitive with methylating agent generated in the metal–halogen exchange. This experiment also serves to rule out any “cage effect” between methyl bromide generated in the immediate sphere of the enolate and solution-based methyl bromide. We emphasize that the combined results of these studies rule out the intermediacy of species 12b, insofar as it implies a special affinity to join the bromine-bearing bridgehead carbon to the methyl group in the original methyllithium species. Also excluded are possibilities arising from initial “nucleophilic methylation” of the ketone followed by formation of a highly strained epoxide, culminating in suprafacial migration of the methyl group to provide overall stereochemical retention (Fig. 4, hypothetical path iii, Eq. 8). The results shown in Fig. 4, Eqs. 5–7 clearly exclude path iii because, in the latter mode the methyl group of solution-based “methyl X” would not appear in the product. Finally, one additional control experiment was conducted, wherein cyclohexanone was subjected to reaction with CD₃Li, in the presence of CH₃I. We observed generation of ethane and addition of CD₃ to the ketone. Importantly, no exchange event between CH₃I and CD₃Li and subsequent attack with CH₃Li were observed.

It will be recognized that the chemistry disclosed thus far corresponds to what would be possible directly from intermolecular DA cycloaddition to the unknown 2-methylcyclobutenone. Next, we wondered whether 8 could serve as a precursor to a trans-fused bicyclic [4.2.0] system. Because a compound of the type 8 can be viewed as a bifunctional electrophile, it was of interest to explore its chemistry upon reaction with a dianion (14). In our inaugural experiment, we studied the reaction of 8a with the previously established geminal dianion 14, derived from bis-deprotonation of phenyl methyl sulfone (for the chemistry of α,α-dilithiosulfones, see refs. 15–22). Remarkably, reaction of 8a with 14 afforded an angular phenylsulfonylmethyl product in 84% yield, as a single stereoisomer (Fig. 5). At this stage, the stereochemistry of the junction could not be assigned with confidence. Reductive cleavage of the phenylsulfonyl moiety was accomplished through reaction with Raney Ni in ethanol. Interestingly, the resultant angular methyl compound, although broadly similar to 8a, was different in detail. Given our confidence in the assignment of 11a as bearing a cis junction, it was surmised that both the initial angular phenylsulfonylmethyl compound generated from the dianion chemistry and its reductively derived angular methyl product should be formulated as trans-fused systems 15a and 16a, respectively.

Fig. 5.

Dianion addition to DA adduct 8a.

To verify the junction configuration, the known hydrindenone 17 (23, 24) was converted to diazoketone 19, through a seldom-used Forster reaction (25, 26) of intermediate α-oximinoketone 18 (Fig. 6). Ring contraction of 19 via Wolff rearrangement (27) provided ester 20 and, thence, acid 21. The latter, upon Barton–McCombie-inspired homolytically induced decarboxylation (28), afforded 22. This compound was also obtained, albeit in only 27% yield, by subjecting our compound 16a to microwave Wolff–Kischner conditions (29).

Fig. 6.

Verification of trans fusion.

Similarly, adducts 8b and 8c were converted to their trans-fused angularly functionalized counterparts, 15b and 15c (Fig. 7). This overall transformation formally corresponds to what could be accomplished from suprafacial DA cycloaddition to trans-2-methylcyclobutenone or by an antarafacial DA cycloaddition to cis-2-methylcyclobutenone (vide supra). We note that it is our experience that in DA cycloadditions, α-halocyclenones are far more effective dienophiles than are their α-H or α-Me counterparts. This improvement in dienophilicity is still another advantage of this chemistry.

Fig. 7.

Dianion addition to DA adducts 8b and 8c.

The mechanism by which 8 is converted to 15 cannot be asserted in detail. At first glance, this reaction could be viewed as a direct S_N2 displacement of the angular bromide, although this is mechanistically unlikely (8). Another more precedent-supported possibility envisions 1,2-addition of 14 to the keto group of 8. The path from phenylsulfonyl-stabilized anion 23a to 16, which corresponds overall to a 1,2 migration of a carbanion, could be viewed from the perspective of a 1,2 migration of the σ bond (Fig. 8, path i). Alternatively, one could also envision formation of a high-energy cyclopropane by intramolecular alkylation, (compare 23b) followed by ring opening of the edge cyclopropane bond to 16 (Fig. 8, path ii) (30–32). Although pathways i and ii are distinct, discerning between them would be a complex matter and we presently have no distinguishing experimental evidence that speaks to these lines of conjecture. We note that by either mechanism, the immediate preworkup species should correspond to anion 24. Indeed, when the reaction conducted between 8 and 14 was worked up with D₂O, the deuterated sulfone 25 (as a mixture of H,D stereoisomers) was obtained. We note, parenthetically, that this latent anion was also successfully reacted with methyl iodide, although, in that case, the desired compound was not separable from 15a, which was generated during the workup.

Fig. 8.

Mechanistic proposals.

We note that these previous mechanistic proposals centered on pathways involving nucleophilic addition to the sterically hindered concave face of the molecule. We acknowledge that an alternative mechanistic pathway could also explain the observed stereoselectivity. In this case, lithium–halogen exchange by the sulfone dianion could, in theory, generate enolate 12a and a bromosulfone anion. After α elimination, generation of a carbene and addition of this species across the double bond of 12a would give rise to intermediate 23b. The delineation of the precise pathway to the trans junction is the subject of current computational analysis.

By combining the high-quality dienophilicity of 7 (4) with the capacity to generate bicyclo[4.2.0]octane ring junctions from angularly brominated cis-DA products, we have entered into the elusive structural space of trans-fused cyclobutanones. We wondered whether these findings could be extended to generate 16-norsteroids bearing four-membered D rings. Although such systems are known, they were previously obtained by ring contraction of standard steroids (33–37).

We first selected a target whose structure, were it to be synthesized, could be assigned with great confidence. As seen in Fig. 9, DA cycloaddition between 7 and Dane’s diene (26) (38–40) gave rise to 28. Clearly, 28 resulted from double-bond isomerization of the primary DA product 27. This type of acid-induced isomerization is well precedented in earlier Dane’s-diene–based routes to steroids (39–43). It should be noted that when conducted thermally, DA reactions of cyclenones with 26 are not significantly regioselective (44, 45). Thus, one would have expected, by precedent, that the thermally driven DA reaction of 7 with Dane’s diene would have also produced 29 as a seriously competing primary product. The absence of detected 29 again points to the special character of the dienophile. Thus, 7 undergoes DA cycloaddition with a level of regioselection that has only previously been observed under Lewis acid catalysis (46).

Fig. 9.

Synthesis of rac-D16-norestrone methyl ether.

Reaction of 28 with dianion 14 provided the angular phenylsufonylmethyl compound 30 in 80% yield. Reduction of the 8,9-double bond led to the dihydro product 31. Next, reductive clevage of the sulfone was accomplished to provide the angular methyl product 32. This assignment can be offered with complete confidence because the chromatographic and spectral properties of rac-32 are identical to those of the enantiomerically homogeneous adduct produced from the known ring contraction of estrone methyl ether 33 (22).

Ordinarily, binding of steroids to estrogen receptors requires a free phenolic hydroxyl in the A ring (47). However, attempts to cleave the methyl ether function of rac-32 to produce the 16-norestrone were not successful, no doubt reflecting the lability of the trans-fused cyclobutanone substructure to the conditions used. In the end, it proved more straightforward to synthesize the triisopropyl silyl-protected version of Dane’s diene, i.e., 34 (Fig. 10) (synthesis details in SI Appendix). The latter was moved through the same sequence as was 26, leading eventually to 37. Deprotection of the A-ring phenol and reductive cleavage of the carbon–sulfur bond led to rac-D16-norestrone (39) (48).

Fig. 10.

Synthesis of rac-D16-norestrone.

Conclusion

It has been shown that a variety of sequences, initiated by DA reactions of various conjugated dienes with 2-bromocyclobutenone, set the stage for stereospecific angular functionalization of the resultant cycloadducts. Through a remarkably straightforward experimental sequence of lithium–halogen exchange and angular methylation, access to the cis-fused series is gained. By contrast, reaction of the cis-fused bromocycloadducts with the presumed dilithiomethylphenyl sulfone leads to the trans-fused angularly substituted bicyclo[4.2.0]octane systems.