Total Synthesis of Hypersampsone M

Abstract

We report the first total synthesis of hypersampsone M, an archetypal member of the homoadamantane polycyclic polyprenylated acylphloroglucinols (PPAPs). Commencing from cyclohexenone, a key cyclopentene annulation followed by ring-expansion results in an elusive hydrazulene that undergoes a series of unexpected late-stage transformations, ultimately enabling completion of the synthesis. The route detailed herein represents a potentially general strategy for the synthesis of related homoadamantane PPAPs.

Since their discovery, the polycyclic polyprenylated acylphloroglucinols (PPAPs) have provided scientists a continuous reservoir of diverse natural products with a broad spectrum of bioactivities and structural architectures.¹ A large part of why the PPAP class of compounds has attracted continued interest from synthetic chemists is that a viable structure–activity relationship (SAR) has not yet been established for them, and in many cases, small structural changes lead to unexpected “turn on” or “turn off” effects on biological activity.² With over 400 isolated compounds as of 2018, it becomes useful to further subdivide the PPAP class of compounds into the bicylic polyprenylated acylphloroglucinols (BPAPs), the caged adamantane and homoadamantane PPAPs, and other spirocyclic or otherwise rearranged PPAPs (Figure 1, top).^1e

Figure 1.

Hypersampsone M and representative PPAP classes with examples of related homoadamantane PPAPs.

This interest has led dozens of researchers to develop synthetic approaches to PPAPs, with many having completed sophisticated total syntheses. The bicyclo[3.3.1]nonane subclass (BPAPs) in particular has enjoyed significant attention in the realm of natural product synthesis, reflecting their overall prevalence within the PPAP family, well-established structural diversity, and medicinal potential.³ Despite these extensive efforts, to date there exist only three syntheses of adamantane PPAP natural products.⁴ Moreover, there are no homoadamantane PPAP syntheses, which is surprising in relation to the subclasses’ substantial scope of at least 70 members.^1e,5

We identified hypersampsone M (1), isolated from Hypericum sampsonii in 2014, to be a prototypical representative of this subclass, making it an excellent target for our initial investigations into this family of PPAPs.⁶ Many homoadamantane PPAPs have been known for decades⁷ and many more have provided glimpses into the promising bioactivity of this class, with several being implicated to have antitumor,^7c anti-inflammatory/immunosuppressive,^5i,8 hepatoprotective,^5c and antiadipogenesis⁹ properties. Like other homoadamantane PPAPs, the phloroglucinol moiety in hypersampsone M (1) is clearly conserved (Figure 1, highlighted in blue), yet the unsymmetrical homoadamantane core and fused cyclopentane make it difficult to utilize an alkylative dearomatization strategy such as was cleverly applied to an adamantane PPAP by Porco and co-workers.^4b

We instead devised a retrosynthetic strategy that centers on the C4-cycloheptanone ring (Scheme 1A, highlighted in blue), electing to employ a late stage Claisen condensation and bridgehead benzoylation on C1 from tricycle 3. Utilizing the C4 ketone as a handle, prenylation and cyclization would allow access to tricycle 3 from simplified hydrazulene 4, which could finally be traced back to cycloheptenone 5 through a key cyclopentene annulation.

Scheme 1. (A) Retrosynthetic Strategy and (B) Cyclopentene Annulation of Enone 7.

We began by targeting cycloheptenone 7 (Scheme 1B). Acylation of known ketone 6(10) with NaH and dimethyl carbonate provides access to the required β-ketoester. Cleavage of the acetal, followed by a Knoevenagel condensation, delivers cycloheptenone 7. While several cyclopentene annulation reactions were considered, the demanding gem-dimethyl moiety excluded most obvious options.¹¹ After significant optimization, our observations indicated that 1,1-dimethylpropargylzinc bromide 8 undergoes facile 1,4-addition to enone 7, providing zinc enolate 9 as a single isomer.¹² Subsequent heating of zinc enolate 9 following conjugate addition was sufficient to effect cyclization, affording the hydrazulene product (10) in a single step.¹³ To our surprise,¹⁴ NOESY studies indicated that the undesired syn-isomer (i.e., C20 and C7) was exclusively isolated as a 4:1 mixture of diastereomers at C3 (i.e., cis and trans ring junctures). Attempts to control the addition temperature or alter solvents did not invert the stereochemical outcome, and we propose that the bulky nature of the nucleophile was likely responsible for the undesired initial syn conjugate addition stereochemistry, perhaps through torsional steering effects.

In response, we rationalized that a cyclohexenone substrate would provide a more predictable stereochemical outcome in the cyclopentene annulation and a successive ring expansion would intercept the original retrosynthesis at hydrazulene 4. We began by targeting the annulation substrate (13, Scheme 2). We were able to access ketone 12 from cyclohexenone (11) in a single step using a radical HAT coupling.¹⁵ Acylation and selenoxide elimination provided activated enone 13 in 73% yield over 2 steps. Proceeding to the key cyclopentene annulation, we were surprised to still observe the presence of the syn addition product at C7. Fortunately, further lowering the temperature to −40 °C during the propargylzinc (8) addition resulted in the formation of the desired anti-isomer 14 (6.3:1) with exclusively cis ring fusion. Following hydrogenation of the cyclopentene, we shifted our focus to ring-expansion of the hydroindene system to access the required hydrazulene (4). Significant effort was made on this front, and it was found necessary to use a designer diazoacetate (15).¹⁶ Deprotonation of 15 with LDA enabled nucleophilic addition to the neopentylic ketone on hydroindene 14, furnishing alcohol 16.¹⁷ Catalytic Rh₂(TFA)₄ was found to be uniquely effective in initiating the ring-expansion rearrangement, resulting in β-ketoester 17 as a mixture of three inseparable isomers, which were putatively assigned to the enol and two ketone diastereomers. Commercially available ethyl diazoacetate and tert-butyl diazoacetate undergo similar rearrangements but ultimately proved to be synthetic impasses, as selective decarboxylation of the newly installed esters could not be accomplished in the presence of the neighboring, preexistent methyl ester. The trimethylsilylethyl ester thus proved vital, as following prenylation to form bicycle 18, treatment with TBAF resulted in the selective decarboxylation of the trimethylsilylethyl ester. Concomitant TBS cleavage produces the desired primary alcohol as a single diastereomer, which upon DMP oxidation results in aldehyde 19 in excellent yield (79%, 2 steps).¹⁷

Scheme 2. Total Synthesis of Hypersampsone M.

With a scalable route to aldehyde 19 established, we were optimistic about the aldol reaction proposed to provide tricycle 20. Unfortunately, classic aldol conditions lead to decomposition or, in most cases, return the starting aldehyde without any trace of desired tricycle 20. In fact, TsOH·H₂O in CH₂Cl₂ was the only set of conditions evaluated that resulted in a product of interest, forging caged lactone 21 slowly in moderate yield. We hypothesized that the aldol reaction to form tricycle 20 is likely highly reversible under most conditions, and perhaps even favors aldehyde 19 due to strain induced upon cyclization. Further lactonization of the aldol product (20) provides a powerful thermodynamic driving force for cyclization and highlights the opportune placement of the methyl ester.

Formation of the lactone (21) seemed so favorable that attempts to reopen or saponify it (Table 1, 21 → 24) with any number of acidic or basic conditions resulted in no reaction or decomposition (entries 1–4). We concluded that the lactone would need to be opened in an irreversible fashion and hypothesized that an amide would be suitably stable. While this approach seemed promising, it introduced the eventual synthetic challenge of needing to activate an extremely hindered amide in order to progress with our plans. For example, opening the lactone with AlMe₃/N,O-dimethylhydroxylamine to afford the Weinreb amide proceeded smoothly (entry 5), but following oxidation, we were unable to further transform amide 25 in a useful fashion (25 → 2). Some amides, such as the unsubstituted ammonia-derived amide, were also formed successfully (entry 6), but subsequent oxidation resulted in decomposition. Many disubstituted amides simply failed to form, as was the case with dimethylamide (entry 7). Fortunately, we were able to install aniline (entry 8) and oxidize the revealed alcohol, providing anilide 22 (Scheme 2) in good yield (70%, 2 steps).

Table 1. Functionalization of Lactone 21: Selected Experiments.

entry	conditions	R	resultsa
1	NaOMe/MeOH	–OMe	decomp
2	Bu3SnOMe	–OMe	0%
3	Al(Oi-Pr)3/MeOH	–OMe	0%
4	KOH/H2O	–OH	decomp
5	NHMe(OMe)·HCl/AlMe3	–NMe(OMe)	85%
6	NH4Cl/AlMe3	–NH2	73%
7	NHMe2·HCl/AlMe3	–NMe2	0%
8	PhNH2·HCl/AlMe3	–NHPh	84%b

^aDetermined by ¹H NMR analysis.

^bIsolated yield.

We next aimed to methanolyze the anilide (22) to the methyl ester, targeting homoadamantane 2 via Dieckmann cyclization. The direct installation of a methyl ester with MeOH/H⁺ was not successful; however, we believed that treatment with Me₃OBF₄ (methyl-Meerwein’s salt) would form the methyl imidate, which could then be hydrolyzed to the methyl ester.¹⁸ Excitingly, we observed direct cyclization to the homoadamantane cage instead, isolating imine 23. Mechanistically, we propose that methyl imidate formation likely occurs first. We then envision that one of two scenarios is likely for bond construction. In one case, enol addition would occur directly to the imidate to form a particularly sterically hindered tetrahedral intermediate. Alternatively, elimination of MeOH or Me₂O could form a nitrilium that can cyclize to the observed imine (23). Following optimization, 2,6-di-tert-butylpyridine (DTBP) was added to suppress decomposition of the prenyl group, and an acidic workup was performed to hydrolyze the imine product and provide access to trione 2 in a single step.

To install the final benzoyl group, we attempted variations of the Danishefsky protocol involving an intermediate bridgehead iodide,^3b as well as Shair’s direct bridgehead acylation protocol.^3p Deuterium incorporation experiments with LiTMP showed that no deprotonation occurs at −78 °C, whereas complete decomposition is observed at 0 °C. A screen of several intermediary temperatures established that 40% deuterium incorporation could be achieved at −35 °C. It has been shown that deprotonation of bridgehead homoadamantane protons is challenging compared to those on bicyclo[3.3.1]nonane systems.¹⁹ Interestingly, the challenge associated with bridgehead acylations in the sister BPAP systems seems to stem from lack of anion reactivity, rather than generation of the stabilized anion itself.^3h This was found to be in stark contrast with homoadamantane 2, as despite the difficulty in generating the required anion, treatment with BzCl directly affords hypersampsone M (1) in moderate yield with an equivalent amount of isolated starting trione (2).

To conclude, we have completed the synthesis of hypersampsone M (1) in 15 steps starting from cyclohexenone, thus achieving the first total synthesis of a homoadamantane PPAP natural product. Additionally, the enabling cyclopentene annulation provides an orthogonal method for installing highly substituted five-membered rings using simple reagents and techniques. This strategy is likely of sufficient generality to allow access to other homoadamantane PPAPs. Biological studies and further development of this route to synthesize additional PPAPs are in progress.

Glossary

Abbreviations

PPAP

polycyclic polyprenylated acylphloroglucinol

BPAP

bicylic polyprenylated acylphloroglucinol

DMP

Dess-Martin periodinane.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c07007.

• Experimental procedures, spectroscopic data (¹H NMR, ¹³C NMR, IR, and HRMS) (PDF)

The authors declare no competing financial interest.