Stereodivergent Attached-Ring Synthesis via Non-Covalent Interactions: A Short Formal Synthesis of Merrilactone A

Abstract

A strategy to control the diastereoselectivity of bond formation at a prochiral attached-ring bridgehead is reported. An unusual stereodivergent Michael reaction relies on basic vs. Lewis acidic conditions and non-covalent interactions to control re- vs. si- facial selectivity en route to fully-substituted attached-rings. This divergency reflects differential engagement of one rotational isomer of the attached-ring system. The successful synthesis of an erythro subtarget diastereomer ultimately leads to a short formal synthesis of merrilactone A.

Graphical Abstract

Introduction

Attached-rings with tetrahedral bridgeheads are typical loci in natural product chemical space. ¹ Nevertheless, the planar equivalents—biaryls—over-populate synthetic libraries (12– 15% of total content), a result of facile synthesis by crosscoupling.² Establishment of point chirality at either bridgehead of an attached-ring system does not always benefit from clear influence of existing ring topology, unlike bridging rings (e.g. outout vs. in-out; endo- vs. exo-), fused rings (e.g. convex vs. concave) or spiro rings (e.g. equatorial vs. axial addition) (see Scheme 1). ^{3, 4} As a consequence, attached-rings ⁵ are not commonly featured as intermediates in synthetic planning. They are, however, embedded in many complex natural products, including the Illicium sesquiterpenes (1–4).⁶ Disconnection to attached-ring intermediates maximized convergency and enabled concise syntheses of 1 ⁷ and 2 ⁸ via the direct heterodimerization of two butenolide starting materials. The surprisingly high rate of this transformation (>5 × 10³ M^–1 s^–1) was enabled by a very low transition state barrier (ca. 4 kcal mol^–1, see Scheme 1) due to exquisite complementarity between reactant orbitals, especially a secondary orbital interaction that favored the (R),(S) (or anti-)⁹ heterodimer (20:1 dr).¹⁰ Here we access the elusive (S),(S) (or syn-) heterodimer representative of 3¹¹ and 4¹² via an alternative strategy that relies instead on acyclic stereocontrol ¹³ operating via non-covalent interactions.^14,15 Whereas prior work established bridgehead stereochemistry by invariable facial selectivity of prochiral rings, this work reports divergent (>20:1 and <1:20) diastereocontrol through selective engagement of one rotational ring isomer over another. Access to the fully-substituted attached-ring system leads to a short formal synthesis of 3.

Scheme 1.

Overview. a. Types of ring combinations; b. Relevance of attached-ring stereochemistry to the Illicium terpenes. c. A solution to attached-ring stereochemistry leads to divergence in high dr and a new model for stereocontrol.

Numerous strategies have emerged for establishment of the vicinal stereocenters of 3 and 4 (Figure 1),^6b-d,16,17 including a beautiful, recent report by Zhang.¹⁸ Interestingly, these authors found (–)-merrilactone A to be devoid of activity in a neurite outgrowth assay (at 0.01, 0.1 and 1 uM). This observation contradicted prior reports that both enantiomers were potent neurotrophic agents^16d and significantly reduced the need to procure this single target. More important was Zhang’s finding that the surrounding chemical space approached the potency of the more useful metabolite jiadifenolide (1), verifying Tiefenbacher’s findings ¹⁹ and supporting the importance of these general motifs.¹⁸ One hypothesis ties enhanced neurite outgrowth to weak, chronic hyperexcitation via antagonism of GABA_A receptors.^{8, 20} Similarly, our interest in the general attached-ring motif arose from the finding that butenolide heterodimers en route to the Illicium terpenes exhibit pathway-selective inhibition of cGAS/STING signaling¹⁰—an important cytosolic DNA sensing mechanism involved in the inflammatory response. ²¹ Previous syntheses cyclized substituents of polysubstituted monocycles (Figure 1) or iteratively cyclized fused rings. An attached-ring approach is absent from prior syntheses due to methodological gaps in attached-ring stereocontrol.

Figure 1.

Prior strategies to address the syn-stereodiad did not intercept attached-ring motifs. Colored dots indicate points of strategic bond cleavage.

Here we report a solution to this attached-ring problem that leads to a short formal synthesis of merrilactone A. The fully-substituted attached-ring bears all the necessary functional groups and is formed with remarkable, divergent selectivity (>20:1) for either stereoisomeric series depending on the intermediacy of a siloxyfuran (syn-series) or amidinium enolate (anti-series). This reaction represents a rare example of attached-ring stereocontrol for both stereoisomers. ²² Experimental observations and density functional theory (DFT) calculations performed at the SMD²³(CH₂Cl₂)–ωB97X-D²⁴/def2-TZVPP^25,26//ωB97X-D/def2-SVP level of theory in Gaussian 16²⁷ converge on a reversal of stereoselectivity by alteration of endo-vs. exo- approach of the Michael acceptor π* to the furan ψ-orbital. Enolsilane and coordinated TiCl₄ splay apart from one another to minimize steric clash, but only si-face addition allows endo-approach of the electrophile.

Results and Discussion

Investigation of the attached-ring synthesis began with preparation of the O-ethyl ester and S-benzyl thioester electrophiles 5a/b. Both were prepared in excellent yield in a single step from 2-bromoacetone and the malonic halfesters/thioesters using Et₃N at 0 °C, followed by extraction in the presence of NaOH to effect condensation. The lithium enolate of 2-furanone (butenolide) was generated with LiHMDS at −78 °C to which 5a/b was then added. The butenolide heterodimers 6a/b were generated in good yield and excellent diastereoselectivity (Scheme 2). We previously reported a mechanistic model to understand stereoselectivity involving π– π interactions between the nucleophilic lithium 2-alkoxyfuran as its cubic tetramer and the electrophile butenolide.¹⁰ This bidentate electrophile may induce deaggregation of tetramer to monomer, still as the π–π complex, in which the electron-withdrawing group is positioned endo to the C3 (α) position of the nucleophile (Scheme 2). Electron dispersion effects in the transition state lower the activation barrier to allow facile C–C bond formation at C5 (γ), faster than proton transfer despite a large (ca. 10¹⁰) predicted difference in K_a.¹⁰

Scheme 2.

Quick entry into attached-rings via butenolide heterodimerization lacking fully-substituted bridgeheads.

In both cases, the stereochemistry corresponded to that of 1 and 2, albeit lacking one quaternary carbon in the butenolide that served as the nucleophile. The first center (3’), α to the 1,3dicarbonyl, could be established stereoselectively according to a simple steric model for stereocontrol: the lower demand of the β-methyl versus the β-butenolide substituent. Installation of the remaining quaternary carbon (C5), γ in the second butenolide, however, did not benefit from a precedented control element. As illustrated in Scheme 1, the conjugate base of the butenolide can easily rotate at the nucleophilic bridgehead carbon C5 to expose either the re- or si-face. Although we hoped that remote transition state stabilization might favor si-face alkylation, no influence of the substructure was obvious and initial passes at computational analysis did not return a strong preference for either re- or si-.

To our surprise, a highly stereoselective (92:8, 73% 10a; 93:7 71% 10b) addition of the butenolide anion of 9a/b to methyl vinyl ketone (MVK) occurred in the presence of DBU at −43°C in CHCl₃. ²⁸ While access to the si-face remained our prime objective, the high and inexplicable conversion to 10a/b proved fascinating. The stereoselectivity could not be reversed with solvent, temperature, concentration, or additives, although iminium catalysis using 1,2-trans-diamine cyclohexane removed all stereoselectivity (1:1 dr) and reduced the yield to 12%.

The intransigence of re-face selectivity prompted us to investigate 2-siloxyfurans in a Mukaiyama-Michael reaction, based on the premise that the silyl group might alter the preferred conformation of the lowest energy transition state. This hypothesis has no clear precedent among Mukaiyama-aldol reactions, where stereochemical reversal by silylation converts the closed (pericyclic) transition states of lithium, boron and zinc enolates to the open transition states of enolsilanes.²⁹ Here the predominant interactions include: selection for low-energy staggered transition state conformations,³⁰ Felkin-Ahn control of addition,³¹ and Evans-Felkin-Ahn control for 1,3-stereoinduction. ³² Similarly, models for relatively-rare stereoselective Mukaiyama-Michael additions ³³ invoke reduction of steric repulsion among staggered transition state conformations, ³⁴ minimization of steric clash with γ-substituents, ³⁵ or kinetic effects like acceleration of oxocarbenium desilylation^36,37 and minimization of single-electron transfer.^38,39 We could find no examples where enol silane formation or the Mukaiyama-Michael reaction reversed the apparent rotational preference of an enolate, cyclic or acyclic, to reverse stereoselectivity.

To perturb attached-ring stereocontrol, a Lewis acid might activate MVK for addition by the siloxyfuran, rotated in a different conformation than the enolate transition state relative to the remote γ-lactone. Treatment of the corresponding trimethylsilyl (TMS) and triethylsilyl (TES) siloxyfuran with MVK and various Lewis acids did not deliver significant quantities of product due to the lability of these species. However, the more robust TBS and TIPS siloxyfuran (11a/b and 11c/d, respectively) underwent addition to MVK in the presence of TiCl₄ (Scheme 4). Remarkably, these conditions led to clean generation of syn stereoisomer 12b from S-benzyl thioester 11d in >95:5 dr and 58% yield. O-ethyl ester 11c proved equally stereoselective but lower yielding, and the TBS siloxyfurans 11a/b did not perform well (see SI).

Scheme 4.

Initial (incorrect or incomplete) hypotheses for re- versus si-facial biases.

Why did the Mukaiyama-Michael reaction of 11c/d lead to almost exclusive si-face addition, whereas the DBU-mediated addition favored the re-face? We first distinguished between two possible nucleophile structures in the Mukaiyama-Michael: a titanium enolate via transmetalation⁴⁰ of the labile siloxyfuran (closed transition state), which might chelate with remote functional groups, or the parent enol silane (open transition state) which would require non-covalent interactions to differentiate diastereomeric transition states.

The viability of transmetalation of the siloxyfuran to a TiCl₃-enolate was first evaluated experimentally. Prolonged incubation of 11d with TiCl₄ at −78 °C generated a red solution; quench with Et₃N and aqueous workup led to recovery of 9b as its β,γ-unsaturated isomer and no siloxyfuran 11d. Nakamura reported that enolsilanes transmetalate TiCl₄ in halomethane solvents to generate titanium enolates with a telltale red color,⁴⁰ in contrast to Lewis acids like SnCl₄, which form the colorless α-stannyl ketones.⁴¹ Most trichlorotitanium enolates decomposed within minutes or seconds, whereas others were characterized by ¹H NMR; mesityl ethyl ketone TiCl₃-enolate was isolated as a stable brown solid.⁴⁰ Whereas brown-red solutions formed from mixtures of 11d, MVK and TiCl₄, the independently-formed titanium-enolates (Et₃N, TiCl₄) were less competent to undergo productive reaction and preferred re-face addition, not si. Simultaneously, various calculated coordination modes of a titanium enolate found the si-face addition transition state to be disfavored by 2.3 kcal/mol (Figure 2), which did not agree with the experimental outcome.⁴² Several rounds of conformational searches using both molecular mechanics and metadynamics were not able to locate lower-energy transition state conformers (see Supporting Information). A titanium enolate was not compatible with either computation or the role of silane in diastereoselectivity.

Figure 2.

Theory and experiment exclude a closed titanium enolate pathway.

Titanium enolate formation had been entertained due to the deprotection of siloxyfurans under the reaction conditions and development of a red color. Despite its speed, deprotection appears competitive to productive reaction, not necessary.⁴³ We probed the ability of TiCl₄ and other Lewis acids to promote siloxyfuran addition to MVK without desilylation. Neither Ti(OiPr)₄, SnCl₄, nor BF₃•Et₂O yielded 12a, but EtAlCl₂ mediated si-selective addition like TiCl₄ (Scheme 4), albeit with reduced dr (25:75) and very low yield (14%).⁴⁴ Incubation of 11d with EtAlCl₂ at −78 °C and exclusion of MVK from the reaction, however, led to full recovery of the TIPS-siloxyfuran, indicating that production of 12 can occur without transmetalation. Thus both experiment and calculation excluded the intermediacy of a titanium enolate as an intermediate leading to 12 (si-face).

Among several computed coordination models involving chelation of TiCl₄ among the Lewis basic heteroatoms of 11a-d, only addition of siloxyfuran to a TiCl₄-coordinated MVK was found to be viable. Experiment supported this mechanism as well: addition of 11d to a precomplexed mixture of MVK and TiCl₄ provided a higher yield (65%) and superior impurity profile than addition of TiCl₄ to a mixture of 11d and MVK (58%). An analogous effect was observed in Reetz’s study of TiCl₄-mediated chelation-controlled Mukaiyama-aldol reactions where complexation of α-benzyloxy aldehydes with TiCl₄ precedes addition by enolsilanes.⁴³ We observed MVK to degrade in the presence of TiCl₄ at −78 °C in a matter of minutes, whereas addition of 11a–d outcompeted this degradation.

The calculated free energy difference between these two transition states (Figure 3) is in excellent agreement with experimental dr: ΔΔG^‡_(calc) = 2.0 kcal/mol vs. ΔΔG^‡:_(expt) >1.8 kcal/mol. The TS leading to re-face addition splays substituents outward to avoid the worst steric clashes: the thioester forces the TIPS away from the furan ring oxygen, and in turn, the TIPS forces the coordinated TiCl₄ away from the furan (thioester ⇅ Si(i-Pr)₃ ⇅ TiCl₄, using the symbol ⇅ for anti-parallel). This re-face addition then requires MVK to adopt an exo-transition state (analogous to Diels-Alder cycloaddition) that prevents favorable secondary orbital interactions between the furan and MVK.¹⁰ In contrast, the TS leading to si-face addition exhibited similar splaying of substituents but, by virtue of furan rotation, allows secondary orbital overlap of MVK in an endo-transition state leading to diastereomer 12b. This transmission of stereochemistry via conformational relay by multiple substituents resembles that observed in Claydon’s oligoxanthenes,¹³ but with the added wrinkle of endo- / exo- transition states to differentiate diastereomeric transition states.

Figure 3.

Theory and experiment converge on open transition states, which fit stereoselectivity / relative rate. Only si-face addition can accommodate an endo-transition state, whereas the re-face enforces an exo-transition state.

A more subtle interaction, not benefitting from the immensity of the TIPS group, governed the reversal of selectivity (re-face addition) in the DBU-mediated reaction. Low-energy transition state conformers were located that corresponded to re- and si-face additions (Figure 4), differing by 1.2 kcal/mol and in excellent agreement with experimental outcome (experimental ΔΔG^‡: 1.5 kcal/mol). Here, the separation of bulky substituents was absent, so that proximal steric interaction dominated. The origin of the energy difference again resided in secondary orbital interactions of MVK with the alkoxyfuran. Both diastereomeric transition states position the methyl ketone endo- to the furanyl-diene, but exposure of the furan si-face forces the MVK β-protons into the butenolide substituents at closer proximity than in the re-face.

Figure 4.

An alternative, open transition state explains the stereoselectivity of the naked enolate Michael addition.

To summarize, the two endo-transition states of the DBU-mediated additions are differentiated by steric repulsion proximal to the site of bond formation. The TiCl₄-mediated additions, in contrast, comprise a low-energy endo transition state and a high-energy exo transition state, which minimize steric repulsion from bulky substituents splayed apart in alternating anti-parallel orientations (see Figure 5). Both low-energy re- and si- transition states are acyclic and differentiated by non-covalent interactions, whereas past stereochemical reversals have, in general, relied on silylenol ethers to open the closed (pericyclic) transition states of alkali metal, zinc and boron enolates.²⁹ These stereodivergent reactions serve as strong incentive to consider prochiral attached-ring intermediates as viable intermediates in retrosynthetic analysis.

Figure 5.

Analysis of re- and si-face addition transition states for siloxyfuran addition. Alternation of bulky substituents (thioester ⇅ Si(i-Pr)₃ ⇅ TiCl₄, see also Figure 3) prevents MVK from adopting an endo transition state.

Aside from its idiosyncrasy, divergent attached-ring stereoselectivity also holds great importance for the exploration of cGAS/STING antagonism and stereochemical structure-activity relationships.¹⁰ This reaction could also be leveraged toward a short formal synthesis of merrilactone A.^16,18

The heterodimerization between thioester butenolide 5b and the parent butenolide anion achieved partly-unsubstituted attached-ring adduct 6b with high stereoselectivity according to our previously established model, confirmed by crystallography (9b x-ray, Scheme 5)⁴⁵. In order to improve the low yield of methylation (Scheme 3), we rearranged choreography to form a siloxyfuran first, which mitigated base-related decomposition of the butenolide. Methylation yield improved substantially to 80% to provide the substrate for long-range stereocontrolled bridgehead alkylation. The S-benzyl thioester now served a third and final purpose: first as an activating group for butenolide heterodimerization without promoting deprotonation (5b to 6b), second as a stereochemical control element in attached-ring bridgehead alkylation (11d to 12b), and third as an acyl-metal equivalent for a palladium-mediated Heck reaction (12b to 13).^46,47 This is the first example of a thioester Heck reaction in total synthesis and serves to highlight the tactical advantages of thioester incorporation. No other acyl metal or acyl radical equivalent proved effective. The new, highly strained butenolide decomposed upon numerous attempts to close the final ring, but its electrophilicity could be attenuated by Luche reduction of the keto function (13 to 14). Finally, taking inspiration from Zhai’s pioneering synthesis,¹⁶ⁱ we closed the final ring via SmI₂-mediated 5-exo-trig radical cyclization. Thus, in 7 steps from butenolide 5b (synthesized in one step), we established the merrilactone A core with full stereocontrol; 4 known steps¹⁶ⁱ complete the target 1.

Scheme 5.

Short formal synthesis of merrilactone A using attached-ring stereocontrol.

Scheme 3.

DBU-mediated γ-selective Michael addition is highly selective for the anti-stereoisomer. DBU: 1,8-diazabicyclo(5.4.0)undec-7-ene

Conclusion

Here we disclose a surprising and, to the best of our knowledge, unprecedented model for divergent attached-ring stereocontrol via base- versus Lewis acid-mediated γ-selective butenolide Michael addition. The differentiation of diastereomeric transition states appears to depend on a change from short-range repulsion to secondary orbital interactions (endo- vs. exo- transition state geometries by analogy to Diels-Alder reactions). The naked enolate that gives rise to re-face selectivity avoids a steric clash encountered between the incoming electrophile and the rotational isomer that leads to si-face reaction. Both geometries are enforced by endo-approach of the electrophile to the alkoxyfuran diene. In contrast, the siloxyfuran paired with a TiCl₄-coordinated MVK splays bulky components outward in alternating fashion (benzyl thioester ⇅ triisopropylsilyl ⇅ TiCl₄) and therefore cannot accommodate the endo- transition state leading to re-face reaction. This attached-ring stereochemical reversal parallels the anti- to syn- reversal of, for example, boron enolate to enolsilane aldol reactions. This aldol stereochemical divergence is caused by a shift from a closed transition state governed by the Zimmerman-Traxler model to an open transition state governed by the Felkin-Ahn model.²⁹ In the attached-ring system reported here, however, both transition states are open. We are not aware of any comparable models for divergent stereocontrol in Michael reactions or attached-ring synthesis. Furthermore, precedented models of enolsilane stereocontrol rely on short-range effects,^{14,15,29–36,43} whereas attached-ring rotational isomers may require longer-range interactions,¹³ as disclosed here in the alternate splaying of substituents. One consequence of this stereocontrol is a short entry to the fused, tricyclic core of merrilactone A (1), which possesses the opposite relative stereochemistry of the neurotrophic sesquiterpenes jiadifenolide (1) and O-debenzoyltashironin (2).

The chemical space around 4¹⁹ (although not 4 itself)¹⁸ possesses potent neurotrophic agents which approach the activity of 1. Since we found 1 and 2 to selectively antagonize GABA_A receptors and hyperexcite neurons^8,20 without convulsion,^20b access to this chemical space should allow greater exploration of these privileged, neuroactive chemotypes. Perhaps more importantly, this work establishes a novel and possibly general model for divergent control of stereochemistry at attached-ring bridgeheads.