First Enantioselective Total Synthesis of Amphidinolide F

Over 30 members of the diverse amphidinolide family of biologically active macrolides have been isolated from the dinoflagellate Amphidinium sp.^[1] From this family, amphidinolides C (1–2)^[2] and F(3)^[3] stand among the most complex and densely functionalized members (Figure 1).^[4] These natural products 1–3 contain eleven stereogenic centers embedded within a 25-membered macrolactone including two trans-disposed tetrahydrofuran ring systems, a 1,4-diketone motif and a highly substituted diene moiety at C₉–C₁₁. In addition to the sizable structural challenges present in 1–3, these macrolides have shown significant cytotoxic activity.^[2,3] Consequently, compounds 1–3 have attracted considerable synthetic attention from numerous laboratories^[5] including our own.^[6] Despite these sizable endeavors,^[5–6] neither amphidinolide C nor amphidinolide F have been successfully synthesized in the 20+ years since their isolation. It should be noted that the stereochemical assignment of compound 3 is based on analogy to compound 1 and isolation from the same organism. Herein, we disclose the first total synthesis of amphidinolide F (3), which confirms both the absolute and relative stereochemistry of the natural product.

Figure 1.

Structurally Complex Amphidinolide Natural Products.

Our initial disconnection in the retrosynthesis involved cleavage of the C₁ macrolactone linkage to provide the ketone 4 (Scheme 1). This ketone 4 should be accessible from sulfone 5 and iodide 6 through an umpolung strategy^[7] involving a sulfone alkylation/oxidative desulfurization sequence^{[6a, 8]} which would mask the otherwise challenging 1,4-dicarbonyl functionality. We noticed considerable “hidden” symmetry within the tetrahydrofuran (THF) portions of fragments 5 and 6. Specifically, the C₁–C₈ and the C₁₈–C₂₅ portions contain nearly identical functionalization, oxidation state and stereochemistry. This observation led us to propose that compounds 5 and 6 might be accessible via common intermediate 7. Ketone 7 should provide access to over half the carbon backbone of the macrocycle as well as the majority of the stereochemistry present in amphidinolide F.

Scheme 1.

Retrosynthesis.

Synthesis of the common intermediate 7 is shown in Scheme 2. Starting from the known alcohol 8,^[9] oxidation and Ohira-Bestmann reaction^[10] cleanly provided the alkyne 10. Removal of the benzylidine acetal under acidic conditions followed by protection and Shonagashira cross coupling provided enyne 13. Sharpless asymmetric dihydroxylation yielded the diol 14 in excellent yield and diastereoselectivity.^[11] Building on the work from Gagosz^[12] and Krause,^[13] we had hoped to use a gold-catalyzed cyclization to generate the enol ether 16. The presence of the 1,2-diol moiety complicates any cyclization conditions as both furan and pyran formation might be feasible. Unfortunately, all attempts to facilitate this transformation under Au-catalysis failed to generate the desired product. Fortunately, we found that AgBF₄[14] nicely provided the desired dihydrofuran 16 in good yield and complete stereoselectivity (>20:1 dr). This transformation was routinely performed on 5-gram scale and provided sufficient quantities of 16 which might serve as a building block for a variety of trans-disposed furan-containing natural products. Subsequent silyl protection and removal of the enol benzoate with MeLi•LiBr^[15] produced the common intermediate 7.

Scheme 2.

Synthesis of Common Intermediate.

Synthesis of the C₁–C₁₄ subunit is shown in Scheme 3. We had hoped to directly trap the enolate derived from ketone 7 with methyl iodide to generate the C₄ methyl derivative 20; however, the C₆ stereochemistry appeared to be the dominant stereocontrolling element in the alkylation – yielded the undesired C₄ methyl stereochemistry. Fortunately, we were able to exploit this directing effect to our advantage through hydrogenation of the exo-methylene compound 19 using Wilkinson’s catalyst to provide the correct stereochemical combination 20. Deoxygenation of the C₅ carbonyl followed by deprotection and oxidation at C₈ generated the aldehyde 22. Next, we required the stereoselective addition of a 2-metallo-1,3-diene species to the α-silyloxy aldehyde 22. The prerequisite iodide 27 was prepared in 4 steps from the previously prepared iodide 24^[6a] through a regioselective hydrostannylation of enyne 25. This regiochemistry is counter to what is typically observed with most Pd-catalyzed hydrostannylations.^[16] Treatment of iodide 27 with n-BuLi followed by addition to the aldehyde 22 provided the C₈–C₉ coupled material in good yield and reasonable diastereoselectivity (3:1 dr).^[17] We had been concerned that the organolithium species might undergo 1,3-metallotropic shifts^[18] to generate allenyl metallo species as well as scramble the C₁₀–C₁₁ E/Z olefin geometry; however, we did not see evidence of this rearrangement occurring under the reaction conditions. Generation of a related vinyllithium species via a hydrazone using Shapiro conditions led to extensive decomposition. After C₈ silylation, incorporation of the C₁₄ iodide via a two-step sequence provided fragment 6.

Scheme 3.

Synthesis of C₁–C₁₄ Subunit.

The construction of the second major fragment was also accomplished using common intermediate 7 (Scheme 4). As before, deoxygenation at C₂₂ easily provided tetrahydrofuran 28. Removal of the pivolate at C₁₈ followed by oxidation generated aldehyde 29. Addition of the organolithium species derived from the known iodide 30^[19] provided the 2° alcohols 31 and 32 as an inseparable mixture of stereoisomers. We initially had hoped to convert the C₁₈ alcohol into its corresponding dimethyl ketal via oxidation followed by ketalization under Noyori conditions. This approach had proven productive in our prior model system.^[6] While the oxidation was effective, we were never able to ketalize the corresponding ketone under a diverse array of conditions. Consequently, we selected protection of alcohol(s) 31 and/or 32 as its ethoxyethyl (EE) ether as a viable alternative. While we believe both C₁₈ alcohols 31 and 32 are viable compounds for the synthetic sequence, we proceeded forward with the 18S isomer 31^[17] for practicality reasons including simplification of NMR spectra. Oxidation of the mixture 31 and 32 with TPAP, NMO followed by reduction with L-Selectride gave the 18S isomer 31 in high distereoselectivity [15:1 (31:32), 85% yield over two steps]. After EE protection of alcohol 31, debenzylation and incorporation of the sulfone moiety at C₁₅ provided the compound 34. Selective 1° TBS deprotetion using HF•pyr followed by Swern oxidation revealed the key α-oxy aldehyde 35. We had initially planned to exploit Julia-Kocienski-Blakemore olefination^[20] of this aldehyde with the known PT sulfone;^[21] however, this reaction showed a preference for the undesired cis alkene. While alternate, multi-step solutions have been developed to circumvent this problem,^[5c,5j] we continued to look for a direct solution. Fortunately, use of the Vedejs-type tributyl phosphonium salt 36^[22] cleanly generated the desired E alkene 37 in good selectivity (97% yield, 11:1 E:Z). Exchange of the C₂₄ silyl protecting groups provided the C₁₅–C₂₉ fragment 5.

Scheme 4.

Synthesis of C₁₅–C₂₉ Subunit

The completion of the total synthesis of amphidinolide F is shown in Scheme 5. The key coupling of the major fragments was accomplished by treatment of sulfone 5 with LHMDS and HMPA followed by the addition of alkyl halide 6 smoothly formed the C₁₄–C₁₅ coupled material 38.^[6a] The nucleophilicity of sulfone carbanions was instrumental in the success of this challenging coupling between an α-branched alkyl iodide and an α-branched nucleophile.^[23] Next, oxidative desulfurization was accomplished using LDA/DMPU followed by treatment with Davis’ oxaziridine to provide the desired ketone 4 along with the Piv deprotected ketone 39 in a combined 65% yield (94% BORSM). While this type of oxidation has been known for some time,^[24] it is only recently starting to gain attention as a viable method for the incorporation of carbonyl moieties in synthesis.^[6,8] Interestingly, the Davis oxaziridine proved superior to our previous TMSOOTMS conditions.^[6a,8a] Both compounds 4 and 39 were easily converted to the seco acid 42. In contrast to our synthesis of amphidinolide B,^[25] macrolactonization proved to be an effective way for construction of the cyclized product 43 with Yamaguchi conditions^[26] being optimum. Next, careful deprotection at C₁₈ under aqueous acidic conditions followed by oxidation yielded the sensitive C₁₅, C₁₈-diketone. Finally, global desilylation using Et₃N•3HF^[27] provided synthetic amphidindolide F (3) which matched with the natural material (¹H, ¹³C, [α]_D).^[2]

Scheme 5.

Total Synthesis of Amphidinolide F.

In summary, the total synthesis of amphidinolide F has been accomplished in 34 steps (longest linear sequence). Highlights to the synthetic sequence include a silver-catalyzed dihydrofuran formation, use of a common intermediate 7 to access both the C₁–C₈ and C₁₈–C₂₅ fragments, regioselective hydrostannylation of enyne 25, diasteroselective addition of a 2-lithio-1,3-diene species to aldehyde 22 and the sulfone alkylation/oxidative desulfurization sequence to couple the major subunits and incorporate the C₁₅ carbonyl moiety.