An Oxidative Dearomatization Approach to Tetrodotoxin via a Masked ortho-Benzoquinone

Abstract

Progress toward a stereoselective synthesis of tetrodotoxin (TTX) is presented. Oxidative dearomatization of a tetrasubstituted guaiacol arene yielded a masked ortho-benzoquinone that intercepted an acyl nitroso species generated in situ by the copper-catalyzed aerobic oxidation of an acyl hydroxylamine. The subsequent alkene dihydroxylation and reduction of a bis-neopentylic ketone proceeded with perfect diastereoselectivity to reveal advanced intermediates toward the synthesis of TTX.

Graphical Abstract

Tetrodotoxin (TTX, 1) is widely known as the toxic principle of pufferfish and is consumed in Japanese fugu. After the isolation of analytically pure samples in 1909,¹ structural elucidation in the 1960s showed that TTX contains a densely functionalized cyclohexane core bearing nine contiguous stereogenic centers wrapped into a dioxaadamantyl cage bearing a pendant guanidinium group.¹ Recent crystallographic studies have unambiguously shown that the guanidinium cation mimics sodium ions in binding voltage-gated sodium channels (Na_v), resulting in the potent inhibition of nerve signaling.² In addition to the resultant well-reported toxicity, TTX is also under study for the treatment of chemotherapy-induced neuropathic pain^3a and opiate addiction.^3b Previous studies have shown that minute structural variations result in the selective inhibition of different Na_V isoforms.¹

Because of both its remarkable biological activity and highly oxidized and stereochemically congested structure, TTX has been subject to landmark synthetic work. Since the first synthesis of TTX as the racemate by Kishi and coworkers in 1972,^4a asymmetric syntheses have been completed by Isobe,^4b Du Bois,^4c Sato,^4d Fukuyama,^4e,f and Trauner.^4g Additionally, a number of natural and unnatural analogs of TTX have been prepared.¹ Various synthetic studies toward TTX have been reported as well, showcasing the continued interest in generating new congeners and advancing methodological frontiers in organic chemistry.¹

Our retrosynthesis began with the disconnection of the orthoacid and guanidinium, as a carbodiimide equivalent, to reveal the hypothetical precursor “tetrodamine” 2 (Scheme 1). The syn-C7,8-diol⁵ is suggestive of a cis-alkene dihydroxylation of diol 3.^4e Antithetic reconnection of the syn-C6,8a-amino-cyclohexnaol motif leads to a retron for an acyl nitroso Diels–Alder cycloaddition, potentially simplifying the installation of two fully substituted stereogenic centers. Previous studies from these laboratories leveraged the Adler oxidative dearomatization of salicyl alcohols with concomitant N-hydroxycarbamate dehydrogenation to generate a pair of mutually reactive species that participated in an acyl nitroso Diels–Alder cycloaddition.^6a Subsequent manipulations to complete a synthesis of TTX were stymied by promiscuous reactivity during reductive epoxide ring openings.^6b We now propose the use of a dimethyl ketal as a latent ketone for late-stage homologation. This choice would enable the facile synthesis of the requisite diene 4 from feedstock aromatics.⁷

Scheme 1.

Retrosynthetic Analysis

The selective ortho-formylation of o-bromophenol 5,⁸ O-methylation, and Dakin oxidation⁹ provided bromoguaiacol 6 in four steps (Scheme 2a). Directed hydroxymethylation using Et₂AlCl proceeded in good yield to provide salicyl alcohol 8, which was protected as acetonide 9. Treatment of the resultant bromoarene with n-butyllithium generated an aryllithium species that was trapped with diethyl oxalate to afford α-ketoester 10. Ketone reduction, acetonide deprotection, and oxidative dearomatization in the presence of diacetoxyiodo-benzene and methanol provided the racemic dienone 4 in 23% overall yield over 10 steps on decagram scale. Whereas we anticipated that the asymmetric reduction of α-ketoester 10 would provide access to enantioenriched material, this system proved recalcitrant when a CBS catalyst,¹⁰ ruthenium-catalyzed transfer,¹¹ traditional hydrogenation,¹² or enzymatic reductions were employed.¹³ Instead, an oxidative kinetic resolution of (±)-11 using a vanadium-oxo catalyst affords the desired mandelate (−)-11 with 98.5:1.5 er.¹⁴

Scheme 2. Synthesis of Masked ortho-Benzoquinone 4^a.

^aReagents and conditions: (a) MgCl₂ (2.0 equiv), Et₃N (2.0 equiv), paraformaldehyde (4.0 equiv), THF, 75 °C; (b) MeI (2.0 equiv), K₂CO₃ (5.0 equiv), DMF, 50 °C; (c) ^mCPBA (1.8 equiv), CH₂Cl₂, rt; (d) Et₃N (10 mol %), MeOH, rt; (e) Et₂AlCl (2.6 equiv), paraformaldehyde (5.0 equiv), CH₂Cl₂, 0 °C to rt; (f) ^pTSA (10 mol %), acetone/2,2-dimethoxypropane, 50 °C; (g) ⁿBuLi (1.1 equiv), then diethyl oxalate (1.4 equiv), THF, −78 °C; (h) NaBH₄ (0.33 equiv), EtOH, 0 °C; (i) THF/EtOH/3 M HCl, rt; (j) PhI(OAc)₂ (1.1 equiv), MeOH, rt.

With sufficient quantities of dienone 4 in hand, we were positioned to evaluate the key acyl nitroso Diels–Alder cycloaddition.¹⁵ After screening various oxidants, we observed optimal results when using a copper(II) chloride–ethyl oxazoline complex, which catalyzed the formation of an acyl nitroso species under aerobic conditions.¹⁶ Through the use of trifluoroethanol as the solvent and under an atmosphere of O₂, we isolated cycloadduct (±)-12 in excellent yield with excellent diastereoselectivity on decagram scale. Whereas hydrogen-bonding interactions between an exocyclic glycolate hydroxyl and dienophile have been proposed to direct π-facial selectivity in similar systems,¹⁷ the reasons for the high observed diastereoselectivity in the polar protic media of this reaction are at present unclear. Given the retrosynthetic plan previously outlined, alkene dihydroxylation was the next targeted transformation. Typical osmium-catalyzed conditions were surprisingly ineffective, typically resulting in no reaction. A stronger oxidant was called for in this case, and we took inspiration from Plietker’s work with in-situ-generated ruthenium tetroxide (Scheme 3).¹⁸ Minor modifications to the reported conditions (i.e., increased catalyst loading and a second subjection of crude material due to low conversion) provided tetrol (±)-13 in a reasonable yield as a single diastereomer. At this stage, the relative stereochemical outcomes were unknown for both the cycloaddition and the dihydroxylation (vide infra).

Scheme 3. Hetero-Diels–Alder Cycloaddition and Oxidative Remodeling^a.

^aReagents and conditions: (a) CuCl₂·2H₂O (20 mol %), 2-ethyl-4,5-dihydrooxazole (40 mol %), benzyl hydroxycarbamate (4.0 equiv), O₂ (1 atm), TFE, rt; (b) RuCl₃·2H₂O (5 mol %), CeCl₃·7H₂O (20 mol %), NaIO₄ (1.2 equiv), MeCN/EtOAc/H₂O, 0 °C; (c) I₂ (10 mol %), acetone/2,2-dimethoxypropane, 30 °C; (d) ^tBuNH₂·BH₃ (1.1 equiv), AcOH (1.1 equiv), CH₂Cl₂, rt, then silica gel (10:1 w/w), rt.

The requisite C5 ketone reduction was next addressed. Whereas reductions of tetrol (±)-13 resulted in intractable product mixtures, selective cis-7,8-diol protection was achieved to provide the derived acetonide.¹⁹ Reaction with tert-butylamine borane in the presence of acetic acid resulted in a highly diastereoselective reduction at C5.^4c,20 Careful attention to the reaction quench revealed that the formation of undesired borates or oxazolidinones, with a concomitant loss of benzyl alcohol from the Cbz protecting group, could be competitive. The preferred workup involved adsorbing the reaction mixture onto silica gel, which enabled the isolation of triol (±)-14 in excellent yield as a single diastereomer. Nuclear Overhauser enhancement experiments on triol (±)-14 provide strong evidence of the shown structural assignment, indicating that alkene syn-dihydroxylation had occurred on the desired π-face.²¹

We anticipated that global acidic deprotection of both ketal protecting groups would set up the needed ketone homologation at C4a, but we were surprised to find the acetonide recalcitrant to cleavage (Scheme 4a). Instead, ketol (±)-15 was isolated following treatment with trifluoroacetic acid. In evaluating potential homologations under strongly basic conditions, three structural factors of ketol (±)-15 required consideration: (1) three free hydroxyl groups; (2) a fully substituted stereogenic center adjacent to the reaction site, thereby making the ketone neopentyl; and (3) a [2.2.2]-bicyclic framework that should restrict the number of accessible conformers and expose the ketone to nucleophilic additions. Silylation of two hydroxyls was achieved using HMDS/I₂ (protection of the glycolate was never observed),²² and the conditions were screened to generate methyl vinyl ether (±)-16, which we proposed could be deprotected under acidic conditions to afford a β-hydroxyaldehyde. Unfortunately, no product was observed when using the Levine–Wittig,^23a Horner–Wittig,^23b Ohira–Bestmann,^23c or Magnus reagents.^23d The ¹H NMR spectra of the crude material after aqueous workup predominately contained benzyl alcohol, indicating Cbz deprotection to afford a putative secondary amine. Whereas carbamates are not usually susceptible to basic deprotection, the inductively electron-withdrawing oxygen substituent is known to increase the lability of protected hydroxylamines.²⁴ We proposed that the diminished reactivity of the C4a ketone with an adjacent fully substituted stereogenic center at C8a induced competitive carbamate deprotection.

Scheme 4. Attempted Homologation and Guanidinylations^a.

^aReagents and conditions: (a) TFA/H₂O, 50 °C; (b) Pd/C (10 mol %), H₂ (1 atm), EtOH, rt; (c) KO^tBu (4.5 equiv), ^tamylOH, 45 °C; (d) 18 (1.5 equiv), HgCl₂ (2.0 equiv), Et₃N (5.0 equiv), DMF, rt.

Given the challenges associated with homologation, we next considered the possibility of changing the order of required late-stage transformations, that is, installing a protected guanidine prior to homologation. In practice, hydrogenolysis of ketal (±)-14 gave an intermediate aminoalcohol that was lactonized in the presence of excess potassium tert-butoxide to provide lactone (±)-17 (Scheme 4b). Unfortunately, traditional guanidinylation conditions using mercuric chloride to generate a reactive carbodiimide from S-methylisothioureas 18 (R = Bn or ^tBu) did not result in the formation of any desired guanidine (±)-19.⁴ We hypothesized that the amine nucleophilicity is diminished by the adjacent dimethyl ketal at C4a, making the amine formally neopentylic. Given the difficulties arising from sterically congested local environments and the remarkable stability of the protecting acetonide, we began exploring alternative protecting group strategies.

We next postulated that if a more acid-labile protecting group was employed, then the C7,8 diol could be unveiled to facilitate the formation of an orthoester prior to guanidinylation. To achieve this goal, we deployed the para-methoxy-benzylidene acetal and, by analogy to the sequence deployed above, synthesized triol (±)-20 as a single diastereomer in two steps from tetrol (±)-13 (Scheme 5). After extensive optimization, we found that acidic resins mediated double deprotection to provide pentol-ketone (±)-21; however, the material rapidly decomposed under ambient conditions and could not be parlayed through downstream transformations. Selective deacetalization under milder conditions was employed to generate pentol-ketal (±)-22; however, hydrogenolysis of the N–O bond to give hexol-amine (±)-23 was not successful.

Scheme 5. p-Methoxyphenyl (PMP) Bis-Acetal Route^a.

^aReagents and conditions: (a) I₂ (10 mol %), 4-MeO-C₆H₄-CH(OMe)₂ (2.0 equiv), CH₂Cl₂, 40 °C; (b) ^tBuNH₂·BH₃ (1.1 equiv), AcOH (1.1 equiv), CH₂Cl₂, rt, then silica gel (10:1 w/w), rt; (c) Dowex D50WX8 (10:1 w/w), 1,4-dioxane/H₂O, 55 °C; (d) ^pTSA (5.0 equiv), EtOH/H₂O, rt.

During studies conducted previously in these laboratories, cyclization of the C9-hydroxyl onto the C5 ketone to generate lactols was observed in some intermediates.⁶ We hypothesized that if the N–O bond embedded in tetrol (±)-13 could be cleaved, then cyclization to form a lactol would occur spontaneously and position the carboxylate ester proximal to the axial C7-hydroxyl, potentially leading to lactonization (Scheme 6). In practice, this cascade sequence was achieved under mild conditions to provide lactone–lactol–amine (±)-24. An X-ray crystallography experiment was performed on this material, which confirmed the connectivity and stereochemistry. These data provide strong support for previous NMR spectroscopic studies that suggested the relative stereochemical outcomes of the nitroso Diels–Alder cycloaddition at C9 and of the dihydroxylation at C7,8. Attempts to guanidinylate this material were unsuccessful, again showing the propensity of fully substituted carbons to dramatically reduce the reactivity at adjacent sites.

Scheme 6. Lactone–Lactol–Amine^a.

^aReagents and conditions: (a) Pd/C (10 mol %), H₂ (1 atm), MeOH, rt. Thermal ellipsoids were reported at 50% probability.

In conclusion, we have advanced 2-bromophenol in 16 steps to the complex lactone (±)-17, among other highly functionalized compounds, en route to a proposed synthesis of tetrodotoxin. Six of the seven contiguous stereogenic centers present in “tetrodamine” were forged with high stereoselectivity, including both fully substituted stereogenic centers embedded within the syn-1,4-aminocyclohexanol motif. This synthesis was enabled by a copper-catalyzed aerobic oxidation of acyl hydroxylamines to trigger an intermolecular acyl nitroso hetero-Diels–Alder cycloaddition. Alkene dihydroxylation and reduction of a bis-neopentylic ketone were then achieved with perfect diastereoselectivity. Unfortunately, two significant challenges were encountered that limit the possibility for this plan to achieve a total synthesis of TTX: (1) homologation of the neopentylic ketone at C4a and (2) guanidinylation of the primary amine at C8a in the presence of an adjacent dimethyl ketal. Efforts to engineer a new C4a aldehyde surrogate and complete a synthesis of TTX are ongoing in our laboratories.