A biosynthetically inspired synthesis of (–)-berkelic acid and analogs

Abstract

We describe a complete account of our total synthesis and biological evaluation of (–)-berkelic acid and analogs. We delineate a synthetic strategy inspired by a potentially biomimetic union between the natural products spicifernin and pulvilloric acid. After defining optimal parameters, we executed a one-pot silver-mediated in situ dehydration of an isochroman lactol to methyl pulvillorate, the cycloisomerization of a spicifernin-like alkynol to the corresponding exocyclic enol ether, and a subsequent cycloaddition to deliver the tetracyclic core of berkelic acid. Our studies confirm that the original assigned berkelic acid structure is not stable and equilibrates into a mixture of 4 diastereomers, fully characterized by X-ray crystallography. In addition to berkelic acid, C22-epi-berkelic acid, and nor-berkelic acids, we synthesized C26-oxoberkelic acid analogs that were evaluated against human cancer cell lines. In contrast to data reported for natural berkelic acid, our synthetic material and analogs were found to be devoid of activity.

Graphical abstract

1. Introduction

In 2006 Stierle et al. disclosed the tetracyclic chroman/isochroman/spiroketal natural product (–)-berkelic acid (Figure 1) isolated from the fermentation broth of an extremophilic Penicillium fungus encountered in the Berkeley Pit Lake in Butte, Montana.¹ The Berkeley Pit Lake, which formed by groundwater seepage into an abandoned copper pit mine, is currently the United States' largest superfund cleanup site containing approximately 30 billion gallons of highly acidic (pH 2.5), heavy metal-contaminated (arsenic, copper, cadmium, cobalt, iron, manganese, and zinc) water². Berkelic acid was found to be a moderate inhibitor of MMP-3 (1.87 μM) and the cysteine protease caspase-1 (98 μM), and, in testing against the NCI 60 cell line panel, was reported to possess selective activity against the human ovarian cancer cell line OVCAR-3 (GI₅₀ 91 nM). However, a subsequent analysis of fully synthetic (–)-berkelic acid from the Snider group in the 60 cell line panel indicated no activity against any of the cell lines (up to 10 μM) including OVCAR-3.³

Figure 1.

Original and revised structures of berkelic acid.

The initially assigned structure of (–)-berkelic acid (1) was determined by NMR experiments, but did not address the configuration of the C22-quaternary stereocenter or the overall absolute configuration. Subsequent work of the Fürstner group led to a revised structure of berkelic acid (2) through an elegant synthetic, NMR, and crystallographic study culminating initially in the synthesis of both C22-epimers of the corresponding methyl esters of ent-2,⁴ and subsequently (–)-berkelic acid (2).⁵ These efforts established the absolute configuration of five stereocenters and a revision of the configuration at C18 and C19 in the original assigned structure. In fact, Fürstner noticed that an advanced tetracyclic intermediate with configuration reminiscent of the original structure could not be prepared as a single diastereomer. An unfavorable eclipsing interaction interaction between C16 and C25 instead provided a driving force for equilibration into a nearly statistical mixture of four diasteromers epimeric at C15, C17, and C18.^6,7 Unfortunately, insufficient spectroscopic differences between two synthetic methyl berkelates epimeric at C22, lack of an authentic sample, and the inability to selectively saponify the C1-methyl benzoate prevented an unambiguous assignment of the C22 stereocenter. Later, Snider and coworkers reported the first total synthesis of berkelic acid, confirming Fürstner's structural revision and putatively assigning the quaternary stereocenter as C22-S.^3,8 As the C22 stereocenter was introduced via a non-selective Kiyooka aldol reaction,⁹ the stereochemical assignment relied upon correlation of the resultant diastereomers to a model compound. Our subsequent synthesis of both C22-epimers of berkelic acid fully corroborated the Snider assignment.¹⁰ Finally, a gram-scale total synthesis was reported in 2012 by Fañanás and Rodríguez that centered on a silver-catalyzed addition/cyclization cascade of appropriately substituted alkyne and aldehyde precursors to set four new chiral centers in a 2:1 diastereomeric ratio.^11,12,13

Constitutionally, the originally proposed structure of (–)-berkelic acid (1) emerges as an amalgamation of two other natural products, namely pulvilloric acid (4)¹⁴ and spicifernin (3). The co-isolation of berkelic acid with spiciferone A (5),¹ which previously had been isolated alongside spicifernin (3) from a Cochliobolus fungus,¹⁵ is indeed suggestive of a proposal that berkelic acid could be the end-product of such a biosynthetic merger (Scheme 1). Biosynthetic studies utilizing ¹³C- and ²H-labeling have specified that both spicifernin (3) and spiciferone A (5) are generated from a common hexaketide precursor.^15,16 Thus, the presence of spiciferone A (5) in the berkelic acid culture medium suggests that the Berkeley Pit Lake Penicillium sp. might also possess the biosynthetic machinery to produce spicifernin (3). To account for the stereochemical discrepancy between natural spicifernin (3) and (–)-berkelic acid (2), which display an epimeric relationship at C18 and C19 (berkelic acid numbering), Snider proposed that spicifernin (3) or an immediate biosynthetic precursor could undergo a double epimerization (C18 and C19) followed by a reduction to provide a correctly configured bis-epi-deoxyspicifernin (6) for addition to pulvilloric acid (4).^3,17 We proposed an alternative hypothesis where spicifernin (3) would be unified with pulvilloric acid (4) first, perhaps via its dehydrated enollactone 7 in a formal cycloaddition with the ortho-quinone methide tautomer of pulvilloric acid, followed by a double epimerization (at C18 and C19) and final reduction.^18,19

Scheme 1.

Biosynthetic proposal for the synthesis of (–)-berkelic acid (2) from the natural products spicifernin (3) and pulvilloric acid (4).

Given berkelic acid's unique structural features, stereochemical questions, biological activity, and potential biosynthetic origin from combination of two other natural products, we embarked on a synthetic program that culminated in the total synthesis of (–)-berkelic acid (2) communicated several years ago.¹⁰ Here, we present a full account of our extensive research program including the development of a novel silver-catalyzed dehydration/cycloisomerization/cycloaddition cascade to forge the tetracyclic chroman/isochroman/spiroketal structure embedded within berkelic acid, the synthesis and full characterization of diastereomers of the originally proposed structure, C22-geminal dimethyl substituted analogs, and C26-oxoberkelates to explore the proposed biosynthesis, and biological evaluation of synthetic berkelic acid and analogs.

2. Results and discussion

Initially, we became interested in a synthetic campaign toward berkelic acid due to a unique constellation of structural features that we perceived as a fruitful platform to implement methodology developed in our group. In our initial decisively non-biomimetic analysis, we thus perceived linear polyhydroxy alkyne 9 as a suitable direct precursor to berkelic acid (2) if conditions could be identified to forge a one-pot cycloisomerization / dehydrative cycloetherification as depicted in Scheme 2. The dividend pay-off would be even higher if thus identified reaction conditions could operate on a protecting group free substrate with all final functionality at the correct oxidation state (i.e. substrate 9). Based on studies from our group, we had considerable confidence in executing such a strategy. Indeed, as shown in the box in Scheme 2, we developed methodology to exploit internal alkynes as a potential nucleus for metal-catalyzed remolding of linear substrates to cyclic/polycyclic structures. In 2006, we reported on the metal-catalyzed cycloisomerization of dihydroxy-alkynes to form spiroketals under mild conditions (relevant to path a, 10 → 11)²⁰ followed by propargylic cycloetherification in 2008 (relevant to transformation 10 → 12)²¹ and a tandem combination that would yield 13 directly from 10 (via 12; propargylic substitution preceding cycloisomerization).²² En route to berkelic acid, we envisioned polyhydroxy-alkyne precursor 9 to be available from the addition of terminal alkyne 14 to aldehyde 15 (C15-C16 bond formation). However, if we were to implement a protecting group free synthesis, this aldehyde 15 would entirely exist as the corresponding non-electrophilic lactol 16, leading us to contemplate the natural product pulvilloric acid (4, the dehydration product of 16) as a viable electrophilic coupling partner instead. It is at this point that our synthetic analysis metamorphosed to one that appeared more reminiscent of the biosynthetic pathway proposed in Scheme 1 and contemplated that a metal-catalyzed 5-exo-dig hydroalkoxylation of alkynol 14 should deliver enol 17,²⁰ a material that would combine with the ortho-quinone methide tautomer of pulvilloric acid (4) via a formal cycloaddition to deliver directly the chroman/isochroman/spiroketal tetracycle of berkelic acid (i.e. bypassing proposed precursor 9). Ideally, a one-pot operation would be identified that enables (1) the dehydration of lactol 16 to pulvilloric acid 4, (2) the cycloisomerization of 14 to 17, and (3) the final cycloaddition between 4 and 17 to deliver berkelic acid (2).²³

Scheme 2.

Retrosynthetic analysis and strategy.

Our synthetic studies began with the development of a concise route to a lactol similar to 16 as a precursor to pulvilloric acid (4). As shown in Scheme 3, selective triflation of commercially available methyl 2,4,6-trihydroxybenzoate 18 provided triflate 19, which readily underwent Suzuki-Miyaura cross coupling with known (E)-1-heptenylboronic acid (20)²⁴ to yield styrene derivative 21 in 83% yield from commercial 18.²⁵ Installation of the homobenzylic alcohol was best accomplished via a three-step sequence from 21 including MOM protection of the free phenols, styrene epoxidation with m-CPBA, and benzylic hydrogenolysis providing alcohol 22 in 76% overall yield from 21.²⁶ Methanolysis of the MOM protecting groups and methyl ester delivered alcohol 23 in 97% yield and was followed by a condensation with triethyl orthoformate via a procedure adapted from the synthesis of pulvilloric acid^14c to deliver the ethanol adduct of methyl pulvillorate 24 in excellent yield.

Scheme 3.

Synthesis of lactol 24, a precursor to pulvilloric acid (4). Reagents and conditions: (a) Tf₂O, lutidine, CH₂Cl₂, 0 °C, 16 h (91%); (b) 20, Pd(dppf)Cl₂ (5 mol%), K₂CO₃, THF/H₂O (10:1), reflux, 2.5 h (91%); (c) MOMCl, i-Pr₂NEt, CH₂Cl₂, 0 °C → rt, 18 h; (d) m-CPBA, CH₂Cl₂, rt, 5 h (84%, 2 steps); (e) Pd/CaCO₃, H₂ (1 atm), MeOH, rt, 20 h (90%); (f) 10-CSA, MeOH, rt, 1 d (97%); (g) (EtO)₃CH, TFA, rt, 18 h (99%); (h) Bu₂SnO (1 mol%), TMSE-OH, PhMe, 100 °C, 16 h (86%); (i) MOMCl, i-Pr₂NEt, CH₂Cl₂, 0 °C → rt, 24 h; (j) m-CPBA, CH₂Cl₂, rt, 4 h; (k) Pd/CaCO₃ (cat), H₂ (1 atm), MeOH, rt, 24 h; (l) HCl/MeOH, rt, 16 h (70%, four steps); (m) Alcaligenes sp. lipases (10 weight%), vinyl acetate, MTBE, 4Å MS, rt, 7 d; (n) DEAD, PPh₃, AcOH, PhMe (0.1 M), 0 °C →rt (78%); (o) HCl/MeOH, rt, 16 h (99%).

With a viable racemic route established, we next explored the possibility to intercept racemic homobenzylic alcohol 22 for a late-stage enzymatic resolution. We subjected (±)-22 to a set of 21 different lipases for acetylation of the homobenzylic alcohol with vinyl acetate and identified three enzymes that provided the desired acetate 25 at a synthetically practical rate. Of these, a lyophilized formulation of lipase from Alcaligenes sp. exhibited high selectivity and scalability, allowing for the preparation of (R)-25 and (S)-22 in 95% and 93% ee, respectively.²⁷ As it was ultimately discovered that berkelic acid was of the C9-(R) configuration,^3-5 a Mitsunobu esterification of the enantioenriched homobenzylic alcohol (S)-22 (93% ee) with acetic acid provided additional (R)-25 in 78% yield, for an overall yield of 85% of (R)-25 from (±)-22. Global deprotection of (R)-25 with acidic MeOH gave (R)-23 in quantitative yield. Treatment with triethylorthoformate as described above afforded (R)-24. Expecting potential difficulties with a selective late-stage methyl benzoate saponification based on precedent by Fürstner and Snider,^4,8 we briefly explored alternative benzoate esters at this stage. For example, transesterification of 21 using catalytic Bu₂SnO afforded 2-(trimethylsilyl)ethyl (TMSE) benzoate 26 in good yield (Scheme 5).²⁸ Elaboration via the method described for 23 supplied racemic 27 for enzymatic resolution. However, screening of the same set of 21 lipases as above did not identify a single lipase to affect this transformation. Literature suggests this failure is likely due to the occupation of the lipase active site by the TMSE ester, preventing transesterification.²⁹ At this point, we decided to move on with methyl benzoate (R)-24 and force ourselves to find a solution for its selective removal at a later stage.

Scheme 5.

One-pot Ag-catalyzed synthesis of nor-berkelic acid methyl esters (44/45) from alkynol (±)-33 and lactol (±)-24.

At the time we started our synthetic studies in early 2007, the stereochemistry at the quaternary center (C22) was not assigned. In light of that, and our desire to explore a rather bold approach as outlined in Scheme 2, we decided to perform our exploratory chemistry toward a gem-dimethyl substituted berkelic acid analog (i.e. nor-berkelic acid). The synthesis of the required alkyne coupling partner analogous to 14 (Scheme 2) is depicted in Scheme 4. To this end, formation of the corresponding phosphonium bromide from commercially available α-bromo ketone 29,³⁰ followed by ylide formation and Wittig olefination with known 2-(4-methoxybenzyloxy)acetaldehyde (30)³¹ gave α,β-unsaturated ketone 31 in 58% yield over three steps. Alternatively, a more efficient one-step titanium-mediated dehydrative aldol condensation between commercially available methyl ketone 28 and aldehyde 30 yielded enone 31 in one step (69% yield). Next, a conjugate addition of the enamine derived in situ from propanal and diethylamine to enone 31 afforded aldehyde 32 in acceptable yield but poor diastereoselectivity (75%, dr = 2:1).^32,33 The anti/syn ratio further eroded to unity during the ensuing Seyferth-Gilbert homologation of aldehyde 32 with the Ohira-Bestmann reagent.³⁴ NMR analysis of the chromatographically homogeneous material obtained following oxidative deprotection with DDQ indicated that the product exists as a mixture of open-chain alkynols syn-33 and anti-33 (1:1) and 33-lactol isomers (verified by gHMBC).³⁵ Samples of 33 enriched in each diastereomer (anti-33:syn-33 - 14:1; anti-33:syn-33 - 1:3.8) were obtained by repeated flash chromatography with arbitrary division of material and served to confirm anti-33 as the less polar diastereomer.

Scheme 4.

Synthesis of gem-dimethyl substituted alkynol 33. Reagents and conditions: (a) 29, PPh₃, THF, reflux, 1 h (71%); (b) NaOH, MeOH, rt, 2 h (c) 30, PhMe, reflux, 4 h (81%, 2 steps); (d) 28, TiCl₄, THF, 0 °C, 30 min; Et₃N, −78 °C, 45 min; 30 (1.2 eq), −78 °C → rt, 2.5 h (69%); (e) EtCHO, Et₂NH, dioxane, rt, 21 h (75%, dr = 2:1); (f) Ohira-Bestmann reagent, K₂CO₃, MeOH, 0 °C, 30 min, rt, 17 h (78%, dr = 1:1); (g) DDQ, CH₂Cl₂:H₂O (10:1), rt, 3.5 h (90%, dr = 1:1); (h) preform cuprate with 34, CuBr·DMS, THF, −78 °C, 1 h; then add to 31, THF, −78 °C, 2 h; (i) K₂CO₃, MeOH, rt, 4 h; (j) DDQ, CH₂Cl₂, H₂O, rt, 2.5 h (73%, three steps); (k) 37, Ti(Oi-Pr)4, i-PrMgCl, Et₂O, −45 °C, 2 h; then 31, −45 °C, 15 min → −20 °C, 3 h (54%).

The above enamine-based route was designed with the prospect to eventually control diastereo- and facial selectivity using an enantioselective organocatalytic variant. However, the ease of epimerization of the aldehyde α-stereocenter during the subsequent alkynylation forced us to consider an alternative approach. Conceptually, 1,4-addition of a metalated propargyl/allenyl species to enone 31 should deliver an α-methyl substituted alkyne straightforwardly. Although initial exploration of reagents more commonly used for the propargylation of aldehydes (Marshall chemistry,³⁶ allenylstannanes,³⁷ allenylboronic acids,³⁸ allenylzinc species,³⁹ and allenyl/propargyl lithiates)⁴⁰ proved frustrating (no Michael addition products 35 or 36), traction was gained when moving to cuprate chemistry. After substantial experimentation, we found that addition of 4 equivalents of a cuprate generated from lithiate 34 and CuBr·DMS to enone 31 in THF at −78 °C yielded racemic TMS-alkyne 35 with the desired anti-isomer dominating in a ratio of ∼8-10:1.^41,42 In contemplating an enantioselective route, the synthesis of configurationally defined and stable allenyl/propargyl cuprate species from propargyllithium species 34 appeared daunting.⁴³ Ultimately, we encountered a manuscript disclosed by Sato and coworkers describing the conjugate addition of configurationally stable homochiral allenyltitanium species to alkylidenemalonates.⁴⁴ Using their conditions, we found that treatment of 31 with allenyltitanium reagent 38, derived from 37 (>98% ee),⁴⁵ provided (S,S)-35 in moderate yield and enantiomeric excess (54% yield; ee = 59%)⁴⁶ but excellent diastereoselectivity (anti:syn > 20:1).⁴⁷ Although 31 was completely consumed, a mixture of byproducts including the 1,2-addition product (∼20%) was formed, leading to the diminished yield with respect to copper-mediated conjugate propargylation. Treatment of the crude racemic alkyne (±)-35 with methanolic base to remove the TMS (→ 36) was followed by oxidative PMB-removal to yield the final alkyne (±) anti-33 in 73% yield from enone 31 (dr = 10:1). Enantioenriched (S,S)-35 on the other hand provided enantioenriched (S,S) anti-33 in 92% yield (2 steps; 59% ee; dr >20:1).

As noted above, our initial approach to berkelic acid entailed the addition of a spicifernin-like alkyne to an electrophilic pulvilloric acidlike synthon, followed by a final cyclopropargylation/cycloisomerization of the thus formed linear alkyne coupled intermediate (Scheme 2, path a). As a prelude to explore this coupling reaction, we attempted to protect the free alcohol of model alkynol 33 (anti:syn = 1:1) as a methyl ether 39 (or acetal 40) with Ag₂O and MeI (Scheme 5A).⁴⁸ Instead, we obtained a crude mixture containing compounds consistent with structure 41 (diastereomers) and some endocyclic enol ether 42 (crude NMR, not isolated, assignment tentative). It was this observation that led us to explore the possibility for a 1-pot metal-mediated cascade entailing: (1) cycloisomerization of an alkynol, (2) dearomatization of a lactol to a quinone methide, and (3) in situ cycloaddition to deliver the berkelic acid scaffold (Scheme 2, path b). After a brief survey of transition metal sources,⁴⁹ we found that AgSbF₆ stroke an ideal balance in effecting both the alkynol cycloisomerization and the quinone methide formation from lactol and acetal precursors, respectively (vide infra). As shown in Scheme 5A, in situ NMR monitoring of a solution of 33 (anti:syn = 1.1:1) and AgSbF₆ (1.4 eq) in d₁₀-Et₂O showed 100% conversion to a mixture of compounds featuring resonances consistent with enol ether 41 within 5 min (no signals indicative of 42). When using sub-stoichiometric AgSbF₆ (0.3 eq) a far more complex mixture was observed after 5 min that simplified over time (≤2.5 h) to a mixture similar to that observed with supra-stoichiometric AgSbF₆ but significant accumulation of endocyclic enol ether 42. Although it has been reported that the carboxylic acid corresponding to 24 will yield the quinone methide pulvilloric acid (4) upon removal of ethanol under ultra high vacuum,^14c we opted to explore the possibility of generating quinone methide 43 in situ via a Lewis acid catalyzed dearomatization.⁵⁰ Gratifyingly, through titration experiments in CD₂Cl₂ monitored by ¹H NMR, we found that the alkynophilic AgSbF₆, effective for the cycloisomerization of alkynol 33, enjoyed sufficient hard Lewis acidity to promote a concentration-dependent equilibrium between (±)-24 and quinone methide 43 (Scheme 5B).

With the above defined parameters in hand, we were more than pleased to observe that stirring an ethereal solution of racemic alkynol 33 (anti:syn = 1.85:1) and racemic lactol 24 in the presence of AgSbF₆ (2.5 eq.) at room temperature cleanly converted to a mixture of five methyl berkelates 44 (∼60% of total) and 45a-d (∼40% of total, 2.5:1.0:1.5:1.2 ratio) comprising >95% of the aromatic containing products (Scheme 5C, entry 1).^51,52 The same results were obtained with an equimolar mixture of diastereomerically enriched (±)-anti-33 (anti:syn = 14:1) and (±)-24 (Scheme 5C, entry 2). Repeated semi-preparative HPLC yielded pure samples of (±)-44 (60% isolated) and the four individual diastereomers 45a-d, the structures of which were unambiguously assigned by X-ray crystallography (44 and 45a-c) and multidimensional NMR studies (45d; Figure 2). Diastereomer 44 corresponds to the revised structure of berkelic acid, whereas diastereomers 45a-d correspond to the equilibrating diastereomers identified by Fürstner when attempting to synthesize the originally assigned structure of berkelic acid.^4,5 In agreement, when individual diastereomer 45a (originally assigned berkelate stereochemistry) was stirred at room temperature with HCl/MeOH, an equilibrium was established yielding a mixture of diastereomers 45a-d in the same 2.5:1.0:1.5:1.2 ratio as observed by NMR of the crude reaction mixture obtained according to Scheme 5C.⁵³ These and other published data,^3-8 thus establish that the C15, C17 and C18 stereocenters can epimerize under the reaction conditions, and therefore, theoretically either anti- or syn-33 could provide access to the correctly configured diastereomer 44. To investigate this idea, a mixture of (±)-24 (1.0 eq), AgSbF₆ (1.5 eq), and (±)-syn-33 (1.1 eq, syn:anti = 3.8:1) in Et₂O was stirred at room temperature, which led to a complex mixture comprised of ∼25% berkelates in addition to previously unobserved, non-berkelate byproducts (crude NMR, Scheme 5C, entry 3). However, pre-mixing an ethereal solution of syn-33 (dr = 3.8:1, 1.3 eq) and AgSbF₆ (2.0 eq) for 30 min prior to the introduction of (±)-24 (1.0 eq) generated a mixture similar (∼6:4, ∼80% of mixture, crude NMR) to that observed previously with alkynol anti-33 (Scheme 5C, entry 4). These results are consistent with a 5-exo-dig hydroalkoxylation in which alkynol anti-33 undergoes cyclization at a rate competitive with berkelate formation, while syn-33 does not. This is likely to the result of a build-up of an unfavorable eclipsing interaction during the cycloisomerization of syn-33 versus anti-33. Pre-incubating syn-33 with AgSbF₆ for 30 min enables a more advanced conversion to in situ formed cyclic enol ether 41 prior to the addition of acetal 24.

Figure 2.

Representations of X-ray crystal structures of 44 and 45a-c and NMR structure of 45d (diagnostic nOe correlations are indicated with blue arrows).

In an attempt to render the reaction catalytic, we were surprised to find that stirring alkynol (±)-33 (anti:syn = 2.7:1; 1.2 eq.) and acetal (±)-24 in the presence of a substoichiometric amount of AgSbF₆ (0.4 eq.) generated an inseparable mixture of 3 fused tetracyclic ring products (60% yield)⁵⁴ from which the major endo-isomer 46 was crystalized by slow evaporation from a EtOAc/hexanes/dichloromethane solution. X-ray crystallographic analysis unambiguously determined the structure to be the one shown in Scheme 6. Of the two conceivable pathways, interception of the quinone methide 43 derived from acetal 24 with either a silver enolate 47 derived from β-keto ester 33 or the endocyclic enol ether 48 derived from a silver-mediated dehydration of lactol-33, we prefer the silver enolate pathway (via 47). Indeed, under otherwise identical conditions, suprastoichiometric AgSbF₆ provided spirocyclic structures (44/45) via 33-derived exocyclic enol ether 41 (Scheme 5C), but the hydroxymethylene required for this cycloisomerization (33 → 41) would already be sequestered if the reaction proceeds through endocyclic enol 48.⁵⁵ Moreover, given the bidentate nature of β-keto esters it is not unreasonable to envision that 33 would sequester the first equivalent of Ag(I) as the silver enolate 47, leading to fused products (e.g. 46) in reaction with quinone methide 43. In agreement with our in situ NMR-studies (Scheme 5A) further cycloisomerization of the γ-hydroxy alkynol in 47 to an exocyclic enol ether is slow with substoichiometric AgSbF₆ possible due to a decreased alkynophilicity of bidentate-sequestered Ag (cf. 47). Longer incubation times or suprastoichiometric AgSbF₆ (Scheme 5A) are thus required to complete the cycloisomerization to exocyclic enol ether 41 en route to the spirocyclic berkelate products 44/45 (Scheme 5C).⁵⁶

Scheme 6.

Reaction of alkynol (±)-33 and lactol (±)-24 with catalytic AgSbF₆ produces fused instead of spirocyclic products.

As noted in the introduction, attempts by Fürstner and coworkers to selectively hydrolyze the phenolic methyl ester of berkelic acid methyl ester were unsuccessful. They, and later Snider and coworkers, solved this issue by resorting to orthogonal ester protecting groups (phenolic benzyl and allyl ester, respectively).^3-8 We decided to reexamine the issue of selective demethylation and after some experimentation found that heating a solution of nor-berkelic acid methyl ester 44 with excess (Bu₃Sn)₂O in toluene provided a mixture of the desired nor-berkelic acid 49 (42% isolated) and a decarboxylated byproduct 50 (34% isolated) in a ratio of ∼1.5:1 (based on crude NMR, Scheme 7).⁵⁷ Byproduct 50 arises from double methyl ester cleavage and spontaneous decarboxylation at the C28 terminus.

Scheme 7.

Synthesis of nor-berkelic acid 49 via chemoselective methyl ester cleavage.

With a convergent streamlined synthetic strategy established, we set out to tackle the total synthesis of berkelic acid containing the quaternary C22-stereocenter. To that end, we decided to exploit the asymmetric enamine alkylation method developed by Koga and coworkers to install the C22-quaternary stereocenter.⁵⁸ Accordingly, valine-derived enamine 52 was prepared by Lewis acid-catalyzed condensation of commercially available 2-ethyl-3-oxobutanoate (51) and (L)-t-butyl valinate in 82% yield (Scheme 8). Sequential lithiation of a toluene solution of 52 with LDA in the presence of lithium-coordinating THF (2.6 eq), followed by alkylation with methyl iodide provided α-quaternary substituted imine derivative 53 (88%, dr > 15:1). Hydrolysis of this material gave crude β-keto ester 54, which was converted to bromine-containing cyclic hydrazone 55 (40% yield) for single crystal X-ray analysis, thus confirming the (S)-absolute configuration in agreement with the Koga model for stereoinduction.⁵⁸ Continuing with the synthesis, titanium-mediated dehydrative aldol condensation of β-keto ester 54 with aldehyde 30 yielded enone 56 in 42% overall yield from enamine 52 (three steps, no intermediate purifications).⁵⁹ Conjugate propargylation of 56 using the homochiral allenyl titanate (R)-38 as described in Scheme 4 afforded an inseparable mixture of 57a,b in a ratio of 2:1 (anti:syn > 20:1) and 52% yield.⁶⁰ Sequential TMS-removal and oxidative PMB-deprotection delivered alkynols 58a,b in 91% yield (2:1 ratio, inseparable, in equilibrium with corresponding hemiketals). Stirring an ethereal mixture of alkynols 58a,b (2 eq) and enantiopure lactol (R)-24 (1 eq) in the presence of AgSF₆ (3 eq) provided a mixture of methyl berkelates 59 (from 58a) and 60a-d (from 58b). Unlike the corresponding nor-berkelates 44 / 45a-d, individual diastereomers could not be obtained from this inseparable mixture. Crude NMR-analysis indicated a ratio of berkelic acid methyl ester (59) versus four other diastereomers 60a-d in a ratio of 3:2, similar to that observed for nor-berkelates 44 / 45a-d.⁶¹ Treatment of the crude mixture of 59 / 60a-d with (Bu₃Sn)₂O in toluene provided pure (–)-berkelic acid (2) in 47% isolated yield over the two steps from lactol (R)-24. The reaction was interrupted at partial conversion to prevent saponification and subsequent decarboxylation at the C28 terminus as was observed for nor-berkelic acid 49 (Scheme 7). The recycled material was subjected to another round of (Bu₃Sn)₂O treatment to yield additional (–)-berkelic acid (2) for a total isolated yield of 59% from lactol (R)-24 based on 1 recycle. Given that at the outset of our program, the absolute configuration at the quaternary C22-stereocenter remained in doubt, we also prepared the corresponding C22-epi-berkelic acid 61 from the enantiomeric alkynols ent-58a,b (53%, 70% after 1 recycle from (R)-24).⁶² Only the C22-S diastereomer 2 displayed spectral data fully congruent with natural (–)-berkelic acid (2),¹ thus unambiguously establishing the complete stereochemical configuration of the natural product. The spectral data and optical rotation of this synthetic (–)-berkelic acid (2) is in full agreement with the data reported for natural and synthetic (–)-berkelic acid from the Snider, Fürstner, and Fañanàs groups.^1,3,5,11 The spectral data for C22-epi-berkelic acid (61) matched those reported for the same material synthesized by Snider and• Fürstner.^3,5

Scheme 8.

Synthesis of (–)-berkelic acid (2). Reagents and conditions: (a) L-t-Bu-valinate (0.99 eq), BF₃·OEt₂ (5 mol%), PhH, 80 °C, 14 h; (b) LDA, PhMe, −78 °C, 1 h; then THF (2.6 eq), 1 h; then MeI, 16 h (88%); (c) HCl/H₂O, THF, rt, 1 h; (d) TiCl₄, Et₃N, 4Å MS, THF, −78 °C, 1.5 h; aldehyde 30 (1.2 eq) −78 °C → rt, 3 h (48% over two steps); (e) (4-bromo-2-nitrophenyl)hydrazine hydrochloride, EtOH, reflux, 2 d (40%); (f) (R)-38 (1.2 eq), THF, −40 °C → −20 °C, 3 h (52%); (g) K₂CO₃, MeOH, rt, 4 h; (h) DDQ, CH₂Cl₂, H₂O, rt, 2.5 h (91%, two steps); (i) 58a,b (dr 2:1, 2.0 eq.), (R)-24 (1.0 eq), AgSbF₆ (3.0 eq), Et₂O, rt, 4 h; (j) (Bu₃Sn)₂O, PhMe, 115 °C, 8 h (47% over two steps; 59% after 1 recycle of recovered 59/60).

As noted in the introduction, berkelic acid (2) appears to be comprised of two other natural products, namely spicifernin (3) and the quinone methide pulvilloric acid (4), with the caveat that the C18 and C19 stereocenters (berkelic acid numbering, see Fig. 1) of spicifernin (3) are of opposite configuration to those found in berkelic acid (2) and a lower oxidation state at C26. Snider had proposed the possibility of a biosynthetic transformation of spicifernin (3) to bis-epi-deoxyspicifernin (6, Scheme 1), followed by incorporation into berkelic acid (2). We offered the possibility of incorporating spicifernin (3) first into bis-epi-oxoberkelic acid (8), followed by epimerization (C18 and C19) and reduction at C26 (Scheme 1). We, and others had already demonstrated that C15, C17, and C18 are all epimerizable within the context of the berkelic acid tetracyclic scaffold.^5,10 The higher oxidation state at C26 (oxo) is thus expected to enable the required additional epimerization at C19 to complete the berkelic acid stereochemical constellation. To test this hypothesis, we embarked on the synthesis of C26-oxoberkelate analogs with C22-gem dimethyl-substituted quaternary carbon (63, Scheme 9) using the silver mediated cycloaddition strategy employed for berkelic acid synthesis (Schemes 5C and 8).

Scheme 9.

Synthesis of oxoberkelates 63a-f. Reagents and conditions: (a) DMP, CH₂Cl₂, rt, 1 h; (b) NaH₂PO₄·H₂O, NaHClO₂·H₂O, 2-methyl-2-butene, t-BuOH, H₂O, 0 °C → rt, 1 h (90%, two steps); (c) NaAuCl₄·2H₂O (5 mol%), MeCN, H₂O, rt, 2.5 h (67%); (d) (+)-24 (1.0 eq), (+)-62 (1.2 eq), AgSbF₆ (1.4 eq), Et₂O, rt, 3 h (50% 63a-e); (e) HCl, Et₂O, CH₂Cl₂, rt, 24 h (67%); (f) AgSbF₆ (1 eq), CD₂Cl₂, rt, 3 h (40%, 63f). The structure of 63b could not be determined unambiguously.

The requisite spicifernin-like γ-lactone 62 was prepared from nor-berkelic acid precursor 33 (racemic) in 60% overall yield by double oxidation (Dess-Martin and Lindgren)^63,64 and gold-catalyzed cycloisomerization.⁶⁵ Treatment of racemic lactone 62 (1.2 eq) with racemic lactol 24 (1 eq) in the presence of AgSbF₆ (1.4 eq) at room temperature in ether for 3 hours produced a mixture containing at least six oxoberkelate diastereomers 63a-f in 61% yield after flash chromatography. Normal phase semi-preparative HPLC purification with two different eluents yielded pure diastereomers 63a (4% isolated), 63b (4% isolated), 63c (13% isolated), 63d (9% isolated), and 63e (2% isolated).⁶⁶ Diastereomer 63e has the same relative configuration as berkelic acid.⁶⁷ Of all the other diastereomers, only 63c was obtained in quantities sufficient enough for equilibration studies. Unfortunately, attempts to equilibrate this diastereomer to 63e (with berkelic acid configuration) have thus far been unsuccessful. For example, treatment of 63c with Bronsted acid (HCl, CH₂Cl₂) – conditions that were effective for the equilibration of berkelate 45a into the 4 berkelate diastereomers 45a-d (Scheme 4) – resulted in isochroman ring-opening to yield elimination product 64, whereas treatment with AgSbF₆ yielded a 1:1 equilibrium mixture of 63c (configuration similar to 45d) and 63f (configuration similar to 45c) from which 63f could be isolated in 40% yield. No equilibration via epimerization at C17, C18, or C19 could be detected. While these initial results do not disprove our biosynthetic hypothesis involving oxoberkelic acid intermediates en route to berkelic acid, our inability to identify suitable epimerization conditions appears to bolster the proposal by Snider that epimerization at C18 and C19 occurs before incorporation of spicifernin into berkelic acid.

During the course of our synthetic campaign toward (–)-berkelic acid (2) described herein, we prepared a number of structural analogs – notably nor-berkelic acid (49), a decarboxylated nor-berkelic acid analog 50, nor-berkelic acid methyl ester (44) and four diastereomers 45a-d, C22-epi-berkelic acid 61, and six diastereomeric oxoberkelic acid methyl ester analogs 63a-f. In the original isolation paper, berkelic acid (2) was shown to be a selective inhibitor of the human ovarian cancer cell line OVCAR-3 with a GI₅₀ of 91 nM.¹ Later, Snider reported that synthetic (–)-berkelic acid (2) and C22-epi-berkelic acid (61) failed to register any antiproliferative activity when tested at a10 mM single concentration in the NCI human disease-oriented 60-cell line panel, which includes the OVCAR-3 cell line.³ Recently Rodríguez et al. reported similarly that their synthetic (–)-berkelic acid (2) and corresponding methyl ester 59 were devoid of activity against the OVCAR-3 cell line.⁶⁸ We therefore decided to test our synthetic (–)-berkelic acid (2), C22-epi-berkelic acid (61), nor-berkelic acid (49), decarboxylated nor-berkelic acid (50), nor-berkelic acid methyl ester (44) and diastereomer 45a, and oxoberkelic acid methyl ester diastereomer 63c against the HeLa (cervical cancer), A549 (adenocarcinomic human alveolar basal epithelial cells), and OVCAR-3 cancer cell lines. Unfortunately, none of these compounds demonstrated any activity when measured up to 10μM concentrations. These results in combination with previous literature observations indicate that the original sample of natural (–)-berkelic acid (2) was either contaminated with a minor potent metabolite, or that there was a methodological problem with the assay.⁶⁹

3. Conclusion

In conclusion, we have developed a scalable, highly convergent synthesis of berkelic acid and analogs based on a key silver-mediated cascade involving the cycloisomerization of an alkynol to an exocyclic enol ether, quinone methide formation from an aromatic isochroman lactol, and subsequent cycloaddition to provide berkelic acid methyl ester in a one-pot operation. This highly efficient approach, inspired by the notion that berkelic acid constitutionally appeared as a combination of two other natural products spicifernin and pulvilloric acid, enabled the total synthesis of (–)-berkelic acid (2) in 10 steps overall longest linear sequence (LLS, 17 total steps) and 10% overall yield. This compares favorable to the total synthesis described by Fürstner (19 steps LLS, 26 total steps, 5% overall yield),⁵ Snider (13 steps LLS, 20 total steps, 2% overall yield),³ and Fañanás (10 steps LLS, 18 total steps, 12% overall yield).¹¹ Furthermore, we observed that substoichiometric amounts of AgSbF₆ during the key-step led to fused tetracyclic products instead of the tetracyclic spiroketal/chroman/isochroman ring system present in berkelic acid. In addition to berkelic acid and its C22-epimer, we also synthesized several analogs including nor-berkelic acid, a decarboxylated nor-berkelic acid analog, nor-berkelic acid methyl ester and four diastereomers, and six diastereomeric oxoberkelic acid analogs. The latter were prepared to explore an alternative to Snider's biosynthetic proposal to reconcile the opposite C18/C19 configuration of spicifernin to that found in berkelic acid.^3,18 Our studies with nor-berkelic acid analogs (quaternary gem-dimethyl at C22, versus quaternary Me, Et) enabled us to obtain X-ray quality crystal structures of naturally configured nor-berkelic acid and several diasteromeric methyl esters for the first time. These unambiguously confirm the observations by Fürstner⁴ and Snider³ that the originally assigned configuration of berkelic acid¹ is not a thermodynamically stable entity and readily equilibrates into a more or less statistical mixture of four diastereomers. Finally, we tested berkelic acid, C22-epi-berkelic acid and several analogs for activity against several human cancer cell lines, and found that in agreement with previous observations, synthetic berkelic acid and the various analogs were inactive. This indicates that the natural sample of berkelic acid contained a minor potent contaminant, or a methodological problem with the bioassay.