C–C Cleavage/Cross-Coupling Approach for the Modular Synthesis of Medium-to-Large Sized Rings: Total Synthesis of Resorcylic Acid Lactone Natural Products

Abstract

The chemical synthesis of medium (8–11 membered) and large sized (≥12 membered) cyclic systems is often challenging. The introduction of transannular strain and loss of degrees of freedom in forming macrocycles often result in poor reaction kinetics and thermodynamics (i.e., thermodynamically disfavored at equilibrium). To address these challenges, we herein report a strategy for the synthesis of medium-to-large sized rings, which leverages strain-release and metal templating through a palladium-mediated C–C cleavage/cross-coupling. By means of DOSY NMR techniques, we identified an undesired competing β-hydrogen elimination pathway, which was substrate dependent. Using a streamlined synthesis of the requisite precursors, our method enables the rapid generation of complex medium-to-large sized rings in a modular fashion through a C(sp²)–C(sp³) macrocyclization. The transformation enabled the short total synthesis of various resorcylic acid lactone (RAL) natural products and unnatural analogues of late-stage intermediates. A mechanistic proposal for the macrocyclization is supported by computational studies of the reaction using density functional theory.

Graphical Abstract

INTRODUCTION

Medium (8–11 membered) to large (≥12 membered) sized rings are ubiquitous structural motifs found in a wide variety of molecules across numerous fields,^1,2 including pharmaceuticals,³ agrochemicals,^4–6 and natural products (Scheme 1A).^7,8 Macrocycles of varying ring size have found applications in areas such as ligand design for transition metal catalysis,⁹ sensing materials,¹⁰ and synthetic biology.¹¹ Due to their broad relevance and the unique challenges associated with their synthesis, a variety of approaches has been investigated for their production.^12,13 However, general solutions for the generation of medium-to-large sized rings in a modular fashion are still heavily sought after.

Scheme 1.

Representative Macrocyclic Natural Products (A); Strain Energy of Simple Carbocycles as a Function of Size, Referenced to Cyclohexane (kcal/mol) (B); Kinetic and Thermodynamic Challenges Associated with Medium-to-Large Sized Ring Synthesis (C) and Our Approach; Other Approaches Which Use Templating and Strain-Release to Address These Challenges (D)

Cyclization approaches to synthesize medium-to-large sized rings are fraught with kinetic and thermodynamic challenges [see hypothetical potential energy surface (PES) in gold in Scheme 1C]. Specifically, unless irreversible, medium-to-large sized ring synthesis can be thermodynamically unfavorable due to reduction of entropy (i.e., fewer degrees of freedom in cyclic products versus their linear precursors) and the introduction of torsional and transannular strain (Scheme 1B) upon cyclization.¹⁴ Furthermore, on the basis of the Hammond Postulate¹⁵ and the Bell–Evans–Polanyi principle,¹⁶ the loss of degrees of freedom and incorporation of strain in the resulting macrocycle is reflected in the transition state, resulting in a large negative entropy (ΔS^⧧) and large positive enthalpy (ΔH^⧧) of activation. Due to the potentially sluggish rates of cyclization, undesired oligomerization or polymerization can compete, leading to diminished yields of the desired macrocyclic products. Often, high dilution is necessary to kinetically disfavor intermolecular over intramolecular reactions (i.e., achieving high effective molarity).^13,17

These challenges have been addressed by various groups, providing criteria for improving the efficiency of macrocyclizations.¹⁸ For example, one can lower the entropic barrier for cyclization by using metal templating or substrate preorganization (Scheme 1D).¹⁴ Poor cyclization thermodynamics can be addressed by generating stable or volatile by-products to render the transformation irreversible.^19,20 Alternatively, the alleviation of strain en route to the product, or the use of an activating reagent,²¹ can switch the reaction from having an endergonic (i.e., ΔG° > 0) to an exergonic energy profile²² (i.e., ΔG° < 0), due to ground-state raising.

Although the above-mentioned approaches address the challenges associated with macrocyclizations in specific systems, substantial limitations still exist, precluding a general strategy. Strategies that exploit substrate activation to facilitate macrocyclization increase step count, making a synthetic route less atom economical, and are limited to specific precursor functionality (e.g., in lactone or lactam synthesis). Although ring closing metathesis (RCM) is often used to take advantage of reaction templating, the ring sizes accessible through this approach are limited: ring sizes of 5–7 and >14 are usually formed, whereas the efficiency in formation of ring sizes of 8–13 is usually quite challenging.^23,24 Macrocyclizations that employ cyclopropanol containing substrates (i.e., a ground-state destabilizing approach), often require challenging precursor syntheses and are usually not amenable to the synthesis of a wide range of ring sizes.²²

We hypothesized that modular access to medium- and large-sized rings could be gained through the combination of ground-state destabilization and reaction-templating strategies. Specifically, inspired by previous work from our laboratories,^25,26 we envisioned a cyclobutanol C–C cleavage/cross coupling approach to macrocyclization from readily accessible precursors. In this way, a strained starting material could be leveraged to overcome the thermodynamic challenges of medium-to-large sized ring synthesis, while transition metal-mediated templating could be used to address potentially challenging kinetics (hypothetical PES in blue, i.e., a → f, Scheme 1C). Specifically, we hypothesized that after palladium-mediated cyclobutanol cleavage, a metal-templated intermediate could be accessed (i.e., a metallacycle, c) which would be only slightly energetically uphill compared to the cyclic product (f), but significantly uphill in comparison to a linear precursor (b). With n + 1 members in the ring, and preorganized for cyclization, an irreversible reductive elimination event could generate the macrocyclic product with n members in the ring.

RESULTS AND DISCUSSION

Drawing on prior work from our laboratory, which has its basis in studies by Uemura, Murakami, and their co-workers,^27,28 we hypothesized that a cyclobutanol tethered to an aryl halide could be employed as a modular platform for the synthesis of various medium-to-large sized rings, including those shown in Scheme 1A. We hypothesized that under palladium-mediated C–C cleavage/cross-coupling conditions, these bifunctional precursors could (1) render a macrocyclization exergonic (i.e., ΔG < 0) due to the strain-release associated with the cyclobutanol in the precursor (i.e., ground state destabilization), (2) palladium-templated reactive intermediates could be generated which could lower the entropic barrier of cyclization (Scheme 1C), and (3) the irreversible nature of reductive elimination could furnish the strained macrocyclic product.

After significant optimization, a streamlined four-step process for the modular synthesis of cyclobutanol precursors was achieved (see the Supporting Information for precursor preparation). With ample precursor S1 in hand, we subjected this material to palladium-mediated C–C cleavage/cross-coupling conditions.^25,29 Complete consumption of starting material was observed, resulting in a nearly copolar mixture of two different species, which proved extremely challenging to separate (Scheme 2, entry 1). DOSY NMR studies (see the Supporting Information for details) revealed that these two species had different rates of diffusion in solution, suggesting an isomeric mixture of a cyclic and a linear species. Ultimately, careful purification yielded the desired macrocyclic product (1), alongside linear enone side-product 1b (Scheme 2), in a 2.1:1 ratio, consistent with our observations by DOSY NMR.

Scheme 2.

Optimization of the Reaction Conditions

At this stage, we sought to tune the conditions to favor the cyclization product (1) over the enone side product (1b) by varying ligands. Even though tert-butyl Xantphos L2 (Scheme 2, entry 2) gave an improved isomeric ratio, significant decomposition was observed. In contrast, biaryl phosphine L3 (entry 3) or BINAP L4 (entry 4) resulted in diminished selectivity or led to complete decomposition, respectively. Ultimately, we found that by increasing the catalyst and ligand loading (entry 5), along with rigorously degassing the reaction solution with argon, desired macrocycle 1 could be obtained in 89% isolated yield. These results suggest that catalyst deactivation or change in speciation through oxidation of the ligand to the corresponding phosphine mono-oxide may contribute to the change in selectivity and reactivity, as recently elucidated by Blackmond, Engle, and co-workers.³⁰ The structure of 1 is unambiguously supported by single crystal X-ray diffraction analysis (Scheme 3). Additionally, we found that the competing enone side product is entirely disfavored for nonfused substrate S2, which, upon subjection to the optimized conditions, provided the desired macrocycle (2) in 72% isolated yield (entry 6), using only 10 mol % catalyst. A slightly higher yield (88%) could be obtained with higher loading, and these conditions were chosen for the investigation of our scope. Other bases such as Na₂CO₃ or K₂CO₃ proved inferior (entries 7 and 8), and Pd(OAc)₂ was essential for the reaction (entry 9). Lastly, given existing precedent for cyclopropanol-mediated cleavage cross-coupling, despite being limited to the synthesis of 5-³¹ and 7-membered³² rings, we tested an analogous cyclopropanol precursor (S3) under the optimized conditions. However, in this case, we observed an inseparable 1.3:1 mixture of the corresponding macrocycle and enone products. Notably, under the optimized conditions, a higher fraction of the enone product is formed in this case, reflecting a higher propensity for the competing β-hydrogen elimination with cyclopropanol substrates.

Scheme 3.

Scope of the C C Cleavage/Cross-Coupling Macrocyclization^a

^aYields shown are isolated yields ^aReaction free energies calculated at the B3LYP-D3(BJ)/DEF2-SVPD (CPCM, 1,4-dioxane, 373 K) level.

With optimized conditions in hand, we set out to explore the functional group tolerance and modularity of the palladium catalyzed C–C cleavage/cross-coupling macrocyclization. A variety of cyclobutanols were productive in the cyclization reaction, generating macrocyclic products bearing a wide array of functional groups in synthetically useful yields (Scheme 3). In alignment with our previous observations (vide supra), ring fusion (see 3) diminished the yield of the desired product and led to more of the β-hydrogen elimination pathway. However, substrates bearing mono- (6, 7, 10, and 12) and disubstituted (4, 5, 8, 9, and 11) backbones were well tolerated providing macrocycles with various substitution patterns in good yields, illustrating that sterically hindered cyclobutanols are tolerated under the reaction conditions. A methyl ester (7, 8, and 9), silyl ether (10), protected spirocyclic azetidine (11), and tertiary amide (12) participated in the reaction to give products in moderate to good yields, presenting different exit vectors to facilitate further functionalization. Notably, heteroatoms like oxygen (13 and 18) or nitrogen (14) could be incorporated into the backbone of the macrocycle, enabling potential access to peptidomimetic chelators or crown ethers.^33,34 An all-carbon macrocycle (see 22), the structure of which was confirmed by single crystal X-ray diffraction, could also be readily accessed, showcasing potentially broad applicability to carbocyclic frameworks.

Additionally, 9–11-membered medium sized rings (15–17) are accessible using this approach, albeit in low to modest yields. In contrast, 13- to 15-membered macrocycles (18–23) are generated efficiently. It is well-known that geminal substituents can accelerate cyclization reactions (Thorpe–Ingold effect).³⁵ Indeed, cyclization of a substrate containing a gem-dimethyl group (see 20) performed better than its methylene analog (see 19). For substrates containing shorter chains, we hypothesized that the low yields for cyclization were due to the greater strain energy in the product, leading to less exergonic reactions and attendant rate deceleration of the desired pathways (see prior discussion on the Bell–Evans–Polanyi principle¹⁶). A slower reaction could then be outcompeted by off-cycle decomposition (e.g., protodebromination). In contrast, the larger macrocycles would be less strained, leading to a more favorable kinetic and thermodynamic outcome.

To better quantify this relationship, we calculated the free energy of reaction (ΔG_rxn) as a function of ring size (see the Supporting Information for computational details). We found a direct relationship between reaction exergonicity and ring size. As shown in Scheme 3 (bottom right inset), the reaction becomes more favorable for products with larger rings. In this case, even though cyclic products with varying degrees of strain (when compared to cyclohexane, see Scheme 1B) are formed, this is offset by strain-release upon opening of the cyclobutanol, providing overall exergonicities similar to those of various linear (i.e., strain-free) palladium-mediated cross-couplings.^36–38 Because these reactions are under kinetic control, and isolated yield can be subject to variations, only a qualitative relationship can be drawn, and in some cases reaction efficiency is not well correlated to the reaction thermodynamics. Nevertheless, when isolated yield is plotted against ring size, the ΔG_rxn correlates with isolated yield across the series, in support of our hypothesis.

Given the modularity and general functional group tolerance of this macrocyclization method, we sought to apply it to the total synthesis of natural products (Scheme 4). Resorcylic acid lactone (RAL) natural products are an intriguing class of polyketide macrocycles with varying biological activity³⁹ including antimalarial, anticancer, and Hsp90 inhibition.^40,41 Even though we targeted 13-O-methylbotryosphaeriodiplodin (31), an aryl macrolactone isolated from a Sicilian fungus which causes disease in grapevines,⁴² we recognized that access to divergent ketone intermediate 30 would enable access to several RAL natural products in a rapid fashion. While precursors of this type have been prepared through Mitsunobu etherification of cresol derivatives, we hypothesized that other nucleophiles, such as commercial benzoic acid derivative 28, would enable modular entry into new macrocyclization precursors. Mitsunobu esterification of diol 27, prepared on gram scale in 67% yield over a three-step sequence, and benzoic acid 28 furnished macrocyclization precursor 29 in 45% yield. Subjection of this material to the optimized conditions provided macrocyclic ketone intermediate 30 in 78% yield on 50 mg scale, showcasing the amenability of the macrocyclization to not only ether and carbon linkages (vide supra), but also esters as well. Subsequent reduction of the ketone using sodium borohydride provided 13-O-methylbotryosphaeriodiplodin (31) in 55% yield, along with 41% of the β-epimer, constituting a 6-step total synthesis of the natural product in 12.9% overall yield. X-ray crystallographic analysis of the natural product enabled its structural elucidation for the first time, namely the cis relative configuration of the C3 methyl and C7 hydroxy substituents. The divergent intermediate 30 could also be deoxygenated at the ketone moiety to access known macrocycle 32, constituting a 7-step formal synthesis of the natural products lasiodiplodin and des-O-methyllasiodiplodin.^43,44 We have also applied this approach to the synthesis of heteroaromatic derivatives of ketone 30, which could prove useful in structure–activity relationship studies.^45,46 Indeed, thiophene-and furan-containing 33 and 34 were accessed in 34% and 72% yield, respectively, underscoring the broad modularity of the method.

Scheme 4.

Divergent Total Synthesis of RAL Natural Products through C–C Cleavage/Cross-Coupling Macrocyclization

Computational Investigation.

As detailed in the introduction, we hypothesized that the challenges associated with medium-to-large size ring synthesis could be addressed using a C–C cleavage/cross-coupling approach, as it would combine several key mechanistic advantages (see Scheme 1). First, ground state raising of the precursor cyclobutanol and its cleavage under palladium-mediated conditions could render the reaction thermodynamically favorable, (2) generate a metal templated, palladacycle reactive intermediate which is only slightly uphill from product, and (3) result in an irreversible reductive elimination to generate a strained cyclic product. We provided initial support for the first of these three aims by correlating ΔG_rxn with reaction efficiency (Scheme 3, inset)—suggesting the importance of ground-state raising by having a cyclobutanol substrate. We have now used density functional theory (DFT) to compute the potential energy surface (PES) of the reaction in order to develop a deeper understanding as to the reaction mechanism.

Geometry optimization and frequency calculations were conducted at the B3LYP-D3(BJ)/Def2-SVPD level of theory⁴⁷ using the asymmetric SMD implicit solvation model in 1,4-dioxane at 373 K, and single point energy calculations were conducted at the B3LYP-D3(BJ)/Def2-TZVP level⁴⁸ using the same solvation model. DFT calculations were conducted using Gaussian ‘16, with the assistance of Spartan ‘20 and Orca5 for minima and saddle point starting structures, respectively (see the Supporting Information for computational details and additional citations). Thermodynamic values (ΔG) shown are with respect to palladium alkoxide I1, while kinetic values (ΔG^⧧) are with respect to the preceding minimum (Scheme 5), and some atoms are omitted for clarity. Since oxidative addition and palladium-alkoxide formation in similar systems have been studied in detail prior to this work,^49,50 we focused on the kinetics and thermodynamics of cyclobutanol C–C cleavage and reductive elimination versus the competing β-hydrogen elimination.

Scheme 5.

Density Functional Theory Computations of the Key Intermediates

Consistent with the previously discussed investigation into reaction thermodynamics as a function of ring size (vide supra), we found in this model system that the conversion of alkoxide I1 to macrocycle P1 is exergonic (ΔG = −44.2 kcal/mol). Initial ligand dissociation to form the less thermodynamically stable but kinetically more reactive tricoordinate⁵¹ κ¹phosphine-ligated palladium alkoxide I2 (+15.0 kcal/mol) sets the stage for cyclobutanol C–C cleavage via β-carbon elimination to generate alkyl palladium I3. This process (i.e., I1 → 13 via TS1) is rate determining (ΔG^⧧ = 30.2 kcal/mol with respect to I1) and is exergonic (−12.9 kcal/mol), in support of our initial hypothesis. Alkyl palladium I3 undergoes initial ligand dissociation to form the more reactive tricoordinate alkyl palladium species I4, which is then poised, through preorganization, to undergo rapid reductive elimination (via TS2, ΔG^⧧ = +15.4 kcal/mol with respect to I3), which is kinetically favored over β-hydrogen elimination (ΔΔG^⧧ = +3.1 kcal/mol), to form product P1, which is significantly downhill (ΔG = −31.3 kcal/mol) with respect to the templated intermediate I3. This difference in the transition state energies corresponds to a theoretical product ratio of 65:1, in favor of reductive elimination at 100 °C, in good agreement with experiment (i.e., only a single product is observed). These computational results support that strain-release and reaction templating can be used in concert to address the challenging kinetics and thermodynamics of medium-to-large sized ring synthesis.

CONCLUSION

In conclusion, we have developed a strategy for medium-to-large ring synthesis using a cyclobutanol C–C cleavage/cross-coupling approach. By combining strain-release with reaction templating, this macrocyclization method enables access to a broad array of macrocycles and is functional group tolerant. Due to the modularity of this approach, natural products in the RAL family of polyketides were prepared in high overall yields in rapid fashion. In addition, unnatural heteroaromatic analogues of late-stage intermediates could also be prepared. We used DFT to investigate the free energies of reaction as a function of ring size and found that reaction exergonicity qualitatively correlates with both product ring size and isolated yield. Additionally, computation of the PES revealed that β-carbon elimination is slow but provided a thermodynamically favorable alkyl palladium species which was preorganized for kinetically favored reductive elimination. We anticipate that the strategy developed here will enable broader access to valuable medium-to-large sized rings, and will find utility in the pharmaceutical and agrochemical industries.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c00801.

Additional experimental details, materials, and methods, and spectral data of all reported compounds (PDF)

ABBREVIATIONS

RAL

resorcylic acid lactone

DFT

density functional theory

Data Availability Statement

The data underlying this study are available in the published article and its Supporting Information.