A Three-Component Coupling Approach to Trachyspic Acid

Abstract

Three-component coupling of the lithium enolate of t-BuOAc, silyl glyoxylate, and an α,β-unsaturated ketone enables the rapid construction of the trachyspic acid carbon skeleton. A 3,4-disubstituted isoxazole is utilized to mask the C7/C9 dicarbonyl. New enolsilane/nitrile-oxide cycloadditions enable the preparation of various 3,4-disubstituted isoxazoles that are challenging to access by other means.

Trachyspic acid (1) was isolated from the culture broth of Talaromyces trachyspermus SANK 12191.¹ Trachyspic acid possesses a substituted citric acid subunit and a hydrophobic sidechain, defining characteristics of the 2-alkyl citrate natural products, and as an inhibitor of tumor heparanase (IC₅₀ = 36 μM), it has inspired several synthetic studies.^2–6 The contiguous C3 and C4 chiral centers and the efficient management of the unusual C6–C9 β,γ-diketoaldehyde have proven to be the principle hurdles to synthesis. For construction of the former subunit, previous syntheses of trachyspic acid have employed either aldol^2,3,6 or Ireland-Claisen rearrangement^4,5 strategies. The occurrence of the fully substituted C3 glycolic acid suggested the opportunity to develop a new silyl glyoxylate-based three-component coupling^7–13 that might permit efficient access to the core substructure. This Letter details experiments directed to this end.

The open chain form (2) of trachyspic acid reveals a 1,4-relationship between the C3 glycolic acid and the C6 ketone, hinting at a potential Michael addition of a silyl glyoxylate with an appropriately functionalized α,β-unsaturated ketone (Scheme 1); however, a C6/C7/C9 tricarbonyl compound would pose an intractable chemoselectivity challenge in the projected three-component coupling. In the interest of streamlining fragment assembly and minimizing functional group and protecting group manipulations,¹⁴ masking the C7/C9-dicarbonyl as an isoxazole emerged as an attractive possibility.¹⁵ This particular packaging of reactive functionality was expected to present the C4 enone as the most electrophilic site on compound 5 and permit downstream unmasking of the C7/C9 keto aldehyde via N–O hydrogenolysis and hydrolysis. The ideal nucleophilic component for the proposed coupling would be an acetate enolate and the needed three-component coupling would in principle complete the full trachyspic acid carbon skeleton.

Scheme 1.

Retrosynthetic analysis of trachyspic acid.

Preliminary studies focused on the synthesis of Michael acceptor 5. Nitrile-oxide cycloaddition with alkynes represents a direct and reliable route to isoxazoles; however, steric effects dictate that terminal alkynes always provide 3,5-disubstituted isoxazoles 9, the incorrect regioisomer needed for the application at hand (Scheme 2).¹⁶ Examples of the preparation of 3,4-disubstituted isoxazoles by nitrile-oxide cycloaddition are scarce.¹⁷ The majority of these approaches rely on a tethered nitrile-oxide¹⁸ or a vinylogous amide or ester directing group.¹⁹

Scheme 2.

Regioselectivity of Nitrile-Oxide Cycloaddition

We sought to dictate the regiochemistry of nitrile-oxide cycloaddition electronically using an electron-rich olefin dipolarophile. For the nitrile-oxide component we selected carboethoxyformonitrile oxide (CEFNO), due to its facile in situ preparation from stable ethyl chlorooximidoacetate (8) (Scheme 2). While nitrile-oxide precursor 8 is commercially available, it can also be prepared in one step from ethyl glycine.²⁰ Initial experiments utilized enol acetate and alkyl enol ether dipolarophiles, which provided the undesired furoxan dimer 11 as the major product. We hypothesized that employing a more electron-rich dipolarophile would facilitate cycloaddition with the electron-poor nitrile-oxide; thus, enolsilanes were tested. Syringe pump addition of triethylamine over 15 h to a solution of oxime 8 and an enolsilane, followed by addition of TsOH˙H₂O and warming to reflux, resulted in the regioselective formation of a 3,4-disubstituted isoxazole.

Since the preparation of 3,4-disubstituted isoxazoles is an unsolved problem in organic synthesis, we evaluated the broader applicability this cycloaddition. Various aldehyde- and ketone-derived enolsilanes were effective dipolarophiles, including both linear and branched alkyl variants (Scheme 3). The regioselective synthesis of isoxazole 10f (from the enolsilane of 3-pentanone) was serviceable where 2-pentyne/CEFNO cycloaddition was not (<10% yield; ~1:1 mixture of constitutional isomers). The preparation of fused-ring isoxazoles is another challenge to organic chemists that was addressed by this method. Small and medium cycloalkynes are unstable due to ring strain, so the preparation of fused-ring isoxazoles has inspired creative methods.^21,22 Five-, six-, and seven-membered ring enolsilanes were found to be competent in the cycloaddition (10g–10j). The modest yields are believed to be due to a combination of product volatility and ester hydrolysis (during the aromatization step). The enolsilanes were used without prior distillation allowing for the operationally simple preparation of isoxazoles in two steps from carbonyl compounds. Prior purification of the enolsilanes did not improve the yields of isoxazole products.

Scheme 3.

Enolsilane Dipolarophiles in Nitrile-Oxide Cycloaddition

Yields are of isolated material after column chromatography, average of at least two trials.

Having developed a reliable, scalable (6.4 g) method for the preparation of isoxazole 10a, we next pursued the completion of enone 5. Claisen condensation with lithiated methyl dimethyl phosphonate provided β-ketophosphonate 12 (Scheme 4). t-Butyl glyoxylate underwent olefination (LiCl/Et₃N), but downstream operations were problematic with this substrate. Thus, the 2-furyl group was selected to mask the C4 carboxylic acid. Olefination of 2-furfural proceeded smoothly with aqueous K₂CO₃ at 100 °C to afford the desired enone 13 (>20:1 E:Z).

Scheme 4.

Synthesis of the C4–C9 Enone

Studies commenced on the heretofore unknown three-component glycolate Michael reaction. Based on prior studies in our group,^9,11 initial experiments employed Reformatsky reagents as nucleophilic triggers; however, unselective mixtures of three-component aldol and Michael addition to the enone 13 were found to be the major products. In contrast, use of lithium enolate 14 provided the Michael addition product 15 (Scheme 5), with no trace of undesired 1,2-addition. Solvent choice influenced the diastereoselectivity: THF provided 2.7:1 dr favoring the desired isomer, while the diastereoselectivity was inverted when toluene (1:1.7), hexanes (1:1.5), or diethyl ether (1:2.0) were utilized. The ability to switch the diastereoselectivity by changing solvents could be an asset for analogue synthesis. This coupling generates two C–C bonds, the critical C3 and C4 stereogenic centers, and completes the full carbon skeleton of trachyspic acid. A reaction conducted with 3.7 g of enone 13 provided 4.9 g of diastereomerically pure 15.

Scheme 5.

Three-Component Coupling

With ketone 15 in hand, remaining synthetic operations included isoxazole N–O cleavage, furan oxidation, and spiroketalization. Exposing ketone 15 to either oxidative or reductive conditions was ineffective at cleaving the furan or the isoxazole, respectively. Decomposition of 15 during isoxazole reduction attempts was believed to be a consequence of the reactivity of the C6 ketone. Treating ketone 15 with acidic methanol resulted in transesterification and mixed methyl ketal formation. Ketal 16 underwent successful N–O reduction with Mo(CO)₆ to provide enamine 17 (Scheme 6).²³ With the C7–C9 dicarbonyl equivalent unmasked, the stage was set for formation of the 5,5-spiroketal.

Scheme 6.

Progress Toward Trachyspic Acid

We expected that Brønsted acid-promoted ketalization of 17 would furnish C6 spiroketal 18. Following the model of Rizzacasa⁵ and Hatakeyama,⁶ enamine 17 was treated with 3 M HClO₄; however, rather than the expected spiroketal 18, ketone 22 was generated as the major product (Scheme 7). Ketone 22 is presumably formed by enamine hydrolysis and retro-Claisen fragmentation of the C8–C9 bond. Other aqueous Brønsted acids were also found to trigger the same undesired fragmentation.

Scheme 7.

Unexpected Retro-Claisen Fragmentation

In an effort to avoid retro-Claisen fragmentation, we tested a series of less nucleophilic ketalization conditions. A number of Lewis acids were found to induce only tautomerization or no reaction, but TiCl₄ demonstrated unique efficacy in promoting spiroketalization. Quenching the reaction with 1 M HCl resulted in hydrolysis of the imine intermediate and hydration of the intermediate C9 enol ether to provide spiroketal 18 as a mixture of diastereomers (Scheme 6). Multiple repetitions of the synthetic sequence revealed sporadic diastereoselectivity in the spiroketalization, which ranged from as high as 5:1 dr to as low as 1.5:1 dr The major diastereomer was found to correspond to the natural isomer in each case.

Oxidative furan deprotection furnished acid 19, revealing the citric acid motif (Scheme 6). Subsequent hemiacetal dehydration delivered trachyspic acid dimethyl ester 20. Methylation of the lone carboxylic acid gave the known trachyspic acid trimethyl ester 21, as confirmed by ¹H NMR and ¹³C NMR analysis.¹ Compounds 20 and 21 have proven to be unexpectedly recalcitrant toward numerous conventional and unconventional saponification conditions.

In summary, the trachyspic acid dimethyl ester was prepared in 10 linear steps from commercially available undecanal. Key steps include a silyl glyoxylate three-component coupling, as well as a new nitrile-oxide cycloaddition of an enolsilane dipolarophile. The latter enables the synthesis of various 3,4-disubstituted isoxazoles that are challenging to access by other means. The aim of future work is to evaluate the scope of the glycolate Michael reaction in the context of other alkyl citrate natural products.