Total Synthesis of (+)-Rubellin C

Abstract

The rubellins are a family of stereochemically complex anthraquinoid heterodimers containing an unprecedented chemical scaffold. Although the rubellins have been known for over three decades, no total synthesis has been achieved since their discovery. Their topology is characterized by a 6–5–6 fused ring system, five neighboring stereocenters including a quaternary center all in a convoluted core, and an anthraquinone nucleus. The rubellin architecture has been shown to inhibit and reverse the aggregation of tau protein, a therapeutically relevant target for Alzheimer’s disease. Herein, we describe the first stereoselective synthesis of a member of the family, (+)-rubellin C, in 16 steps. Strategic disconnections allow expedient construction of stereochemical and topological intricacy in a short sequence of borylative and transition metal-catalyzed steps.

Graphical Abstract

Anthraquinone-based natural products have provided a variety of medicinally valuable entities, including FDA-approved therapies such as the chemotherapeutic anthracyclines, the chemotherapeutic and immunosuppressant mitoxantrone for multiple sclerosis, the laxative sennosides, and the anti-inflammatory diacerein for osteoarthritis.¹ A large number of naturally occurring anthraquinoids are characterized by homodimeric linkages.² Fewer family members are characterized by heterodimeric structures that contain anthraquinones linked by multiple bonds to other anthraquinone or xanthone partners; these natural products have inspired elegant syntheses to access their molecular architectures.³

The rubellins are a distinct class of anthraquinone dimers that were isolated in the 1980s in Italy from the phytopathogenic fungus Mycosphaerella rubella in enantiopure forms (Figure 1A).⁴ Research surrounding the compounds lay dormant until the isolation of the rubellins in another fungus, Ramularia collo-cygni, the pathogen responsible for barley leaf disease in Europe.⁵ The rubellins potently deaggregate and inhibit the formation of paired-helical filaments of tau protein, a biomarker and viable pharmacological target in tauopathies such as Alzheimer’s disease.⁶ This implicates the rubellin scaffold as a possible tool for neurodegenerative diseases.⁷ Other isolated natural products with the core rubellin structure include the torrubiellins, uredinorubellins, melrubiellins, solanrubiellin, and caeruleoramularin.⁸ Despite numerous biosynthetic and biological studies, none of the rubellin natural products have yet to be prepared through chemical synthesis, presumably due to substantial synthetic challenges presented by their intricate structures.

Figure 1.

(A) Anthraquinone-based heterodimeric rubellin natural products. (B) Retrosynthetic analysis of (+)-rubellin C.

The rubellin core is beset with a myriad of synthetic challenges (Figure 1), including the quaternary center buried in an array of five contiguous stereocenters, the 6–5–6-fused ring system, and the assortment of oxidation levels. We were particularly interested in developing a synthetic approach to rubellin C (1) because of the exceptional challenge of accessing a Csp³–Csp³ hybridized “attached ring system”,⁹ where the quaternary C16 carbon in a ring is joined through a single carbon–carbon bond to C18 via a stereochemically defined lactone.

The challenge is enhanced further due to the presence of three neighboring stereocenters to the Csp³–Csp³ junction, burying the quaternary center in a fused ring system. Given the prevalence of the attached ring motif⁹ and the inability to synthesize the rubellin ring system to date, we envisioned that a stereoselective approach to the core structure would have broad utility in complex molecule synthesis.

Our retrosynthesis of (+)-rubellin C (1) initially disconnects the central C_sp²–C_sp³ C5–C11 bond via a diastereoselective intramolecular Mizoroki–Heck reaction directed by the orientation of the quaternary center (Figure 1B).¹⁰ The anthraquinone moiety could be introduced by a Hauser–Kraus annulation with a p-quinone monoketal intermediate (not shown) derived from selective aromatic oxidation of homoallylic lactone 4. We reasoned that the key buried stereocenter at C16 and hydroxyl stereocenter at C18 could be accessed in a highly selective manner through phthalaldehyde 5 and properly designed synthon allyl anion 6. Stereoselective access to the natural product would require this carbon–carbon bond-forming step to function precisely to produce the C16 and C18 stereocenters. A reasonable stereochemically defined allylic nucleophile corresponding to the allyl anion synthon would have to engage aldehyde 5 through the correct carbon (α- vs γ-addition), on the correct face, and with the correct orientation of the electrophilic partner, thus distinguishing between a total of eight possible isomers (constitutional and diastereomeric) of the coupled product. We proposed a sterically hindered allylboron allylation that would translate stereochemistry through a closed transition state.¹¹ The stereochemically enriched allylating reagent could be prepared from an abundant chiral pool starting material (7).

Initially we developed a synthesis of the core structure of rubellin C to analyze our stereochemical hypotheses (Scheme 1). Our synthesis began with D-(−)-quinic acid 7, which was elaborated to enone 8 in a few steps.¹² Benzylic Grignard reagent 9, derived from o-cresol, was transformed into the cuprate and underwent 1,4-addition to enone 8 in the presence of trimethylsilyl chloride. The resulting silyl enol ether, obtained as one diastereomer, was converted to substituted enone 10 through Saegusa–Ito oxidation. To access the synthetic equivalent of synthon 5 (Figure 1), we chose to examine recently reported borylation chemistry from the Szabó group that would enable the stereospecific conversion of chiral allylic alcohols into allylboron reagents.¹³ Selective reduction of the carbonyl center followed by allylic borylation would give our desired stereochemistry by inversion. A variety of reductants were screened, with L-Selectride (lithium trisec-butylborohydride) giving the best chemo- and diastereoselectivity for addition into the convex face of bicyclic enone 10. We utilized the aforementioned borylation to convert the resulting free allylic alcohol 11 to the bench-stable allyl nucleophile. Boronic acid pinacol ester 12 was formed in high diastereoselectivity using bis(pinacolato)diboron (B₂pin₂) and a palladium(II) tetrafluoroborate catalyst (Pd-(MeCN)₄(BF₄)₂).

Scheme 1. Asymmetric Synthesis of Chiral Rubellin Core Structure^a.

^aSee the Supporting Information for reagents and conditions; L-Selectride = lithium tri-sec-butylborohydride, DMSO = dimethyl sulfoxide, TBAF = tetra-n-butylammonium fluoride, DMAP = 4-dimethylaminopyridine, DMF = dimethylformamide.

With the allyl nucleophile prepared, we then pursued the key allylative and Mizoroki–Heck transformations in our core system (Scheme 1). Thermal allylation of allyl pinacol boronic ester 12 with benzaldehyde furnished homoallylic alcohol 13 as a single isolable diastereomer. Presumably, the high stereo-control was imparted through a closed Zimmerman–Traxler transition state on the convex side of the bicyclic system. This transformation installed the quaternary and hydroxyl stereocenters at C16 and C18 respectively in high yield and remarkably high diastereoselectivity. Removal of the tert-butyldimethylsilyl (TBS) protecting group with tetra-n-butylammonium fluoride (TBAF) generated free phenol 14, whose X-ray crystal structure confirmed our stereochemical hypotheses in the preceding allylation, borylation, and reduction reactions. Triflation and subjection to palladium-catalyzed Mizoroki–Heck conditions gave one diastereomer of the final 6–5–6 fused system 15.¹⁴ The stereochemistry was confirmed by nuclear Overhauser effect (NOE) correlations to be the cis-ring junction as desired in the natural product scaffold.

After gaining supporting evidence for our stereochemical hypotheses in constructing the core structure of the natural product, we began the synthesis of the full rubellin C system with the allylation of phthalaldehyde 5, which was accessible from commercial materials in seven steps.¹⁵ Unfortunately, conventional allylation protocols utilized in complex molecule synthesis proved to be ineffective for the assembly of the buried quaternary carbon center in the rubellin C structure, presumptively due to steric congestion in the transition state (see the Supporting Information for details). For example, treatment of pinacol boronic ester 12 with phthalaldehyde 5 gave no yield of homoallylic lactone 4 under extreme thermal (>220 °C), ligand exchange, Lewis acidic, or Brønsted acidic conditions.¹⁶ Conversion of allylic alcohol 11 to allyl metal reagents derived from metals known to proceed through Type I allylation reactions provided low reactivity with a mixture of isomers.¹⁷

We then examined a similar procedure for formation of allyl boronic acids instead of allyl boronic esters from allylic alcohol 11 utilizing tetrahydroxyboron B₂(OH)₄.¹⁸ We hypothesized the higher allylative reactivity of allyl boronic acids, stemming from a lack of a hindered steric environment around the empty p-orbital of boron, would possibly allow for this congested key carbon–carbon bond-forming event. In addition, allyl boronic acids are thermally stable and thought to form trimeric allylic boroxines upon removal of water, containing only one oxygen per boron atom in a remarkably Lewis acidic system.¹⁹ Use of Pd(MeCN)₄(BF₄)₂ with B₂(OH)₄ in a DMSO/MeOH mixture followed by extraction and treatment with Na₂SO₄ provided postulated allylating reagent 16 as either the boronic acid or allylboroxine, which allylated phthalaldehyde 5 to give homoallylic lactone 4 as one isolable diastereomer (Scheme 2). Notably, allylboron reagent 16 was the only reagent observed to undergo allylation with phthalaldehyde 5 to selectively afford the desired homoallylic lactone product. A large screening campaign was undertaken to optimize the yield of the boronic acid/allylation sequence. We realized in our studies that phthalaldehyde 5 could be used as a limiting reagent to produce 80% yield of lactone 4 based on limiting reagent with partial recovery of starting allylic alcohol 11, demonstrating the powerful reactivity of allylboronic acids even in sterically crowded chemical environments. The requirement for excess allylic alcohol 11 may arise from the aggregated structure of the putative allylating species, boroxine 16. Through the course of this transformation, we formed both the key quaternary and hydroxyl stereocenters in the Csp³–Csp³ attached-ring system with high control of stereo- and regioselectivity, forming one out of eight possible isomers.

Scheme 2. Synthesis of (+)-Rubellin C^a.

^aReagents and conditions can be found in the Supporting Information; DMSO = dimethyl sulfoxide, TBAF = tetra-n-butylammonium fluoride, THF = tetrahydrofuran, Tf₂O = trifluoromethanesulfonic anhydride, DMF = dimethylformamide.

Assembly of the core Csp³–Csp³ attached-ring system allowed us to pursue the late-stage Hauser–Kraus annulation and Mizoroki-Heck reaction (Scheme 2). Removal of the silyl group with TBAF provided the free phenol of 4. Higher equivalents (>1.5) of TBAF epimerized the C18 lactone center, hinting toward its propensity to undergo stereochemical reconfiguration, possibly through either an addition to the C18 lactone center or deprotonation at the C18 position of the vinylogous ester. Oxidation of the free phenol with (diacetoxyiodo)benzene (PIDA) formed the p-quinone monoketal 17 in 60% yield over two steps.

Initial examination of the Hauser-Kraus annulation with typical bases such as lithium diisopropylamide (LDA) or the dimsyl anion with matching equivalents of p-quinone monoketal 17 and sulfone 18, accessible in six steps from commercially available starting materials,²⁰ gave poor yields of the bright red anthraquinone solid 3 (see the Supporting Information for more details).²¹ Use of lithium tert-butoxide (LTB) as a base gave an increase in yields; however, at higher base equivalents (>3) or upon heating, scrambling of the C18 stereocenter was observed to give undesired epimer 19 through either addition of the butoxide and opening of the ring system or through a deprotonation event to give an alkoxyfuran species that reprotonates on the opposite face. Careful examination of base equivalents led us to employ 2.65 equiv of LTB to achieve high yield of anthraquinone 3 with preservation of stereochemistry. The absolute stereochemistry of the undesired C18 epimer 19 formed in the Hauser–Kraus reaction with excess base was confirmed by X-ray crystallography, which also further established our stereochemical assignment of the allylation preceding this step.

To complete the synthesis, we attempted triflation, subsequent Mizoroki–Heck reaction, and deprotection of the near-complete rubellin system (Scheme 2). Treatment of anthraquinone 3 with excess sodium hydride in CH₂Cl₂ followed by addition of trifluoromethanesulfonic anhydride (Tf₂O) and pyridine supplied the triflate with minimal loss of stereochemical fidelity at the C18 center, which was immediately used in the subsequent Mizoroki–Heck reaction to deliver one diastereomer of the rubellin architecture 20, with its stereochemistry confirmed by NOE correlations. Surprisingly, O-demethylation proved challenging; after vast screening of conditions, the mild nucleophilic iodide source magnesium iodide etherate suitably removed all of the methyl groups at higher temperatures.²² Acidic hydrolysis of the acetonide produced one enantiomer of the natural product (+)-rubellin C (1), whose characterization data matched that described in the literature.⁴ Our synthetic efforts therefore also confirm the absolute stereochemistry reported of natural (+)-rubellin C (1).

The first total synthesis of a member of the rubellin family of natural products has been accomplished with high stereo-selectivity in 16 steps (longest linear sequence) from readily available D-(−)-quinic acid (7). The expedient construction of the core ring system relied on an efficient sequence of steps to install topological and stereochemical complexity, which may provide the means to explore structure–activity relationships. Exploration of these methods, the syntheses of the remaining rubellins, and biological studies are currently underway in our laboratory.