Asymmetric Synthesis of Nidulalin A and Nidulaxanthone A: Selective Carbonyl Desaturation Using an Oxoammonium Salt

Abstract

Nidulaxanthone A is a dimeric, dihydroxanthone natural product which was isolated in 2020 from Aspergillus sp. Structurally, the compound features an unprecedented heptacyclic 6/6/6/6/6/6/6 ring system which is unusual for natural xanthone dimers. Biosynthetically, nidulaxanthone A originates from the monomer nidulalin A via stereoselective Diels-Alder dimerization. To expedite the synthesis of nidulalin A and study the proposed dimerization, we developed methodology involving use of allyl triflate for chromone ester activation followed by vinylogous addition to rapidly forge the nidulalin A scaffold in a four-step sequence which also features ketone desaturation using Bobbitt’s oxoammonium salt. An asymmetric synthesis of nidulalin A was achieved using acylative kinetic resolution (AKR) of chiral, racemic 2H-nidulalin A. Dimerization of enantioenriched nidulalin A to nidulaxanthone A was achieved using solvent-free, thermolytic conditions. Computational studies have been conducted to probe both the oxoammonium-mediated desaturation and (4+2) dimerization events.

Graphical Abstract

INTRODUCTION

Dihydroxanthones are rare in nature due to their instability towards aromatization and reduction. Though frequently proposed as key intermediates in tetrahydroxanthone biosynthesis,¹ only a limited number of dihydroxanthones have been isolated as stable natural products. Nidulalin A (1),² a dihydroxanthone, dienone natural product, showed potent inhibition against DNA topoisomerase II (Topo II) (IC₅₀ = 2.2 μM) and cytotoxicity,³ was isolated along with related congeners including 2H-nidulalin A derivatives 2 and 3 (Figure 1A).⁴ Nidulalin B (4) was a co-isolated benzophenone natural product² which underscores the propensity towards aromatization of 1. In 2020, the Zhang group isolated the novel dihydroxanthone-derived homodimer nidulaxanthone A (5) (Figure 1B) from Aspergillus sp.⁵ Structurally, compound 5 features an unprecedented heptacyclic ring system which is highly unusual in comparison to other natural dihydroxanthone dimers such as the 2,2’-linked structures phomalevones A (6) and C (7).⁶ Although (±)-nidulalin A (1) was synthesized by Hosokawa and coworkers in 2009⁷ using a 10-step synthesis, we were interested to develop a concise synthesis of 1 to study methods for chemical dimerization to nidulaxanthone A (5). In this paper, we report our studies to synthesize nidulalin A (1) and the corresponding dimeric congener nidulaxanthone (5) using allyl triflate for chromone ester activation followed by vinylogous addition and carbonyl desaturation using Bobbitt’s oxoammonium salt to rapidly construct the nidulalin A scaffold. We also report computational studies to probe the key desaturation and dimerization events.

Figure 1.

A. Nidulalin A and co-isolated natural products; B. Nidulaxanthone A and other natural dihydroxanthone dimers.

RESULTS AND DISCUSSION

We envisioned that nidulaxanthone A (5) may be derived from nidulalin A (1) by endo-selective dimerization with facial selectivity anti to the sterically demanding ester groups. Nidulalin A (1) may be accessed from the tricyclic scaffold 8 which can be further synthesized from the known substrate 9⁸ and a diene or diene equivalent via Diels-Alder cycloaddition to maximize efficiency (Figure 2). However, in our experiments we did not observe reactivity of butadiene or equivalents (e.g. sul folene) in [4+2] cycloaddition with 9 under a variety of conditions.⁹ With these observations in hand, together with the successful silyloxy benzopyrylium addition chemistry developed by our laboratory,¹⁰ we changed our approach to evaluate an indirect pathway to construct the two-key C-C bonds (Scheme 1A). We envisioned altering our previously employed siloxybenzopyrylium generation protocol to prepare an activated allyloxybenzopyrylium reagent. To implement this approach, we evaluated use of in situ-prepared allyl triflate¹¹ to replace trialkylsilyl triflates for activation of chromone ester 9. We reasoned that if allyl triflate can activate the protected chromone ester, then allyloxy chromenone 12 could be generated directly without intermediacy of siloxy chromenone 14. Although such applications of allyl triflate¹² have not been reported, the strong affinity of this reagent towards heteroatoms drew our attention for application in the current synthesis.

Figure 2.

Retrosynthetic analysis for nidulaxanthone A

Scheme 1.

A. Allyl-OTf activation of chromone ester 9 and vinylogous addition of a silyl ketene acetal. B. Proposed mechanism for generation of chromenone 13.

In our experiments, we found the allyl triflate could indeed activate chromone ester 9 and form the desired allyloxy benzopyrylium triflate intermediate 10 which readily underwent vinylogous addition with silyl ketene acetal 11.¹³ The stoichiometry of 11 (1.46 equiv.) was found to be important to maximize conversion to product and minimize production of TMSOTf (vide infra). After quenching the reaction with TBAF and triethylamine,¹⁴ allyloxy chromenone 12 could be isolated via silica gel chromatography (Scheme 1A). Interestingly, the trime-thylsilyl triflate (TMSOTf) byproduct generated was found to be detrimental to the reaction. Specifically, substrate 12 is electron-rich and appeared to exchange with the silyl triflate to afford allyl triflate and siloxy chromenone 14 (Scheme 1B). Accordingly, byproduct 13 was also isolated after quenching reactions with TBAF. To suppress this undesired process, we employed an excess of allyl triflate. With 2.8 equiv. of the crude allyl triflate used for activation of 9, the desired allyloxy chromenone 12 was isolated in 55 % yield (100 mg scale) and 47 % yield (1 g scale). In all reactions, the desilylated chromenone 13 was found to be the only observable byproduct. With intermediate 12 in hand, Claisen rearrangement in refluxing toluene overnight followed by ring-closing-metathesis (RCM) with Grubbs-II catalyst (0.75 mol%) smoothly afforded tricyclic tetrahydroxanthones 15 and 16 in a 1.2:1 ratio (Scheme 2A). Due to the close polarity of the two diastereomers, we carried them both forward without purification. At this point, desaturation and allylic oxidation transforms were required to access nidulalin A (1). We began evaluation of the final stages towards 1 by comparing the order of the two events. In the same pot, crude 15 and 16 were treated with LiHMDS followed by addition of N-tert-butylbenzenesulfinimidoyl chloride¹⁵ (Mukaiyama reagent). Surprisingly, only the trans-diastereomer 15 was capable of enolization and afforded dienone 19 in 60 % yield while cis-diastereomer 16 was recovered. Efforts to epimerize the 16 into 15 using a variety of conditions, including acid, base, or thermolysis resulted in enrichment of stereoisomer 16. DFT computations (r²SCAN-3C/CPCM(CH₂Cl₂)¹⁶) showed that the trans isomer 15 is 1.76 kcal/mol less stable than 16.⁹ Examination of molecular models of both diastereomers indicated poor alignment of the α-keto-methine of 16 with the ketone moiety which may prevent enolization (Figures 3A & B). Next, dienone 19 was demethylated via treatment with magnesium iodide (MgI₂) followed by allylic oxidation using selenium (IV) oxide (SeO₂). Although we were able to produce trace amounts of nidulalin A (1), the major product of the reaction was benzophenone 20 (Scheme 2B). As we realized the necessity of the allylic alcohol to block aromatization, we revised our strategy to an allylic oxidation-desaturation sequence. In an initial attempt, treatment of crude 15 and 16 with SeO₂ led to difficulties in product purification due to multiple products generated from both diastereomers. To avoid this issue, we demethylated 15/16 using MgI₂ in the same pot which afforded the readily separable phenols 17 and 18 in 41 % and 27 % yields respectively (1.5:1 d.r.) from allyloxy chromenone 12 (Scheme 2A). Attempts to epimerize 18 to 17, similar to the case of 16 to 15, failed under a variety of conditions. Again, DFT computations showed that trans isomer 17 was 0.78 kcal/mol less stable than 18 with poor alignment of the α-keto-methine of 18.⁹

Scheme 2.

A. One-pot sequence to the nidulalin A carbocyclic core. B. Initial evaluation of desaturation-allylic oxidation route and allylic oxidation-desaturation sequence.

Figure 3.

A. DFT model of 15. B. DFT model of 16.

With compounds 17 and 18 in hand, we evaluated allylic oxidation by exposure of 17 to SeO₂ in refluxing dioxane (Scheme 2B). We were excited to find that nidulalin A (1) was generated in 5 % yield. However, we were unable to optimize conditions to improve the yield of 1 in a single transformation. Moreover, nidulalin A (1) was found to be inseparable from allylic alcohol 21 (vide infra). Fortunately, allylic alcohol 2 could be isolated in pure form via column chromatography as the major product along with the Dauben-Michno ketone 22 and xanthone 23 as major byproducts.

After further optimization, we found that SeO₂ oxidation of 17 in refluxing dioxane with inclusion of water afforded ketone 22 and xanthone 23 as major products along with the desired allylic alcohol 2 as a minor product (Figure 4A). Interestingly, oxidation of 17 using SeO₂ buffered with sodium bicarbonate in anhydrous dioxane (100 °C, 24 h) afforded allylic alcohol 2 as the major product in ~20 % conversion. To expedite this sluggish reaction, we found that use of a slight excess of SeO₂ in anhydrous toluene employing microwave conditions (2 h, 130 °C) could significantly shorten the reaction time. Allylic alcohol 2 unambiguously matched reported data for the natural product 2H-trans-nidulalin A (2).⁴ The only major byproduct using microwave conditions was benzophenone 20 which originated from dehydration of allylic alcohol 2 followed by aromatization. Fortunately, using the microwave conditions developed, allylic alcohol 2 was isolated in 47 % yield on a deca-milligram scale (Figure 4B). On the other hand, diastereomer 18 exhibited poor control in allylic oxidation (Figure 4C), and the major product 21 did not match the literature reported natural product 2H-cis-nidulalin A (3).⁴

Figure 4.

Study and optimization of allylic oxidation of 17/18 using SeO₂.

Mechanistically, we believe that the byproducts may be derived from allylic alcohol 2 (Figure 5). Xanthone 23, whose structure was verified by single X-ray analysis,⁹ may originate from desaturation of 2 to nidulalin A 1 followed by oxidation to dienedione 24 and thermal [1,5]-acyl shift¹⁷ via 25. To validate this possibility, we prepared dienedione 24 (Figure 4) by treatment of 1 with excess MnO₂.¹⁸ Xanthone 23 was indeed obtained by thermolysis of crude 24 at 100 °C in dioxane.⁹ Diketone 22 may be derived from a Dauben-Michno oxidative allylic transposition process via intermediate 27 followed by desaturation of 28. Benzophenone 20 may be obtained from dehydration of 2 followed by aromatization of 29. The presence of water appears to enhance overoxidation to byproducts 22 and 23. However, addition of 4Å molecular sieves to the mixture of 17 and SeO₂ under standard conditions completely shut down reactivity and led to recovery of 17.

Figure 5.

Proposed mechanism for generation of 20, 22, and 23.

With allylic alcohol 2 in hand, we envisioned that nidulalin A (1) could be obtained after a final desaturation step. Although numerous desaturation methods employing transition metal catalyst¹⁹ and traditional Saegusa-Ito protocols¹⁹ have been developed, desaturation adjacent to a hydrogen-bonded carbonyl has not been previously reported using such methods. Employing transition metal desaturation methods reported by Su (Cu^II, TEMPO),²⁰ Dong (Pt^II, Zn^II),²¹ Newhouse (Pd^II, Zn^II),²² and Stahl (Pd^II, O₂)²³ on substrate 2 failed to afford 1. Use of a stoichiometric amount of IBX²⁴ or (PhSeO)₂²⁵ to generate 1 from 2 resulted in decomposition. Moreover, refluxing a mixture of DDQ and 2 in toluene only returned starting material. Although we were able to prepare a silyl enol ether derived from 2 using HMDS and TMSI,^26,27 this substrate also showed lack of reactivity towards desaturation reagents including Pd(OAc)₂ and DDQ. In particular, the inability of DDQ to mediate desaturation of 2 or its derived silyl enol ether may be due to the presence of an allylic alcohol and ester on both faces which may prevent formation of the requisite charge transfer complex.^26,27

To solve this desaturation challenge, use of ammonium N-oxide reagents drew our attention based on recently developed methods employing the Iwabuchi oxidant^28,29 (AZADO-BF₄) 32. Although allylic alcohols tend to be reactive towards such reagents,^29,30 we reasoned that the ester of 2 might prevent alcohol oxidation, as we only recovered starting material 2 after treatment of DDQ. Based on a literature search which identified Bobbitt’s oxoammonium salt 30 as an oxidant for both oxidation of alcohols³¹ and desaturation of ketones,^32–36 we found that 30 cleanly converted allylic alcohol 2 to nidulalin A (1) in 78 % yield in refluxing DCE without alcohol oxidation (Figure 6A). On the other hand, use of the AZADO oxidant 32 resulted in decomposition and TEMPO-BF₄ 31 resulted in a lower yield of 1 along with unidentified byproducts. Based on the excellent chemoselectivity observed when using oxoammonium salt 30, we evaluated different substrates to probe the desaturation mechanism. We found that the presence of the allylic alcohol in substrate 2 was dispensable as substrates 17 and 33 were cleanly converted to desaturated products 20 and 34, respectively (Figure 6B). Interestingly, treatment of substrates 18 and 21, each bearing cis-bicyclic stereochemistry, with the oxidant 30 led to recovery of starting materials.

Figure 6.

A. De novo desaturation using Bobbitt’s oxoammonium salt. B. Substrate scope. C. Formation of enone 36 using Bobbitt’s salt 30 and hydrogenated substrate 35.

Given that Bobbitt’s salt 30 is known to serve as a hydride acceptor,^36,37 together with the premise that an ester n(O)-to-σ*(C-H) interaction should activate the β-hydrogen of ketone from computational analysis of 2 (left inset, Figure 7), we considered that desaturation of successful substrates 2 and 17 may occur by a hydride transfer mechanism which is facilitated by the adjacent ester moiety. Accordingly, we hydrogenated substrate 17 to 35 (Figure 6C) to remove the cyclohexenyl moiety. Treatment of 35 with Bobbitt’s salt 30 (40 °C, MeCN) led to the clean formation of enone 36 in 54 % yield. This control experiment reinforced a desaturation mechanism involving the ester n(O)-to-σ*(C-H) activation rather than the allylic π(C=C)-to-σ*(C-H) activation.³⁶ At present, we cannot rule out a mechanism involving enol formation from substrate 2 followed by desaturation by oxoammonium salt 30 which was also proposed in the literature.^33–34 Although DFT computations (r²SCAN-3C/CPCM (CH₂Cl₂)) showed that the enol desaturation is almost barrierless (TS-SB, 2.3 kcal/mol),⁹ we were unable to identify an energetically reasonable enolization process. The lowest energy transition state (TS-SA, 30.2 kcal/mol)⁹ along these lines utilized the pendant amide of 30 to mediate enolization.

Figure 7.

DFT structures of 2 & 30 and energy diagram for concerted desaturation of 2 using 30 (r²SCAN-3C/CPCM (CH₂Cl₂)).

After extensive computational studies, we found that concerted, asynchronous desaturation of 2 by 30 to afford 1 and 37 had the lowest energy barrier. By way of comparison, the transition state from β-hydride to O of 30 (Figure 7, right, TS-B, 27.1 kcal/mol) is favored by 3.6 kcal/mol relative to β-hydride to N of 30 (Figure 7, left, TS-A, 30.7 kcal/mol), likely due to the fact that oxoammonium 30 resides in a boat conformation with hydrogen bonding to the substrate. Based on DFT calculations (r²SCAN-3C/CPCM (CH₂Cl₂)) Bobbitt’s salt 30 itself (right inset, Figure 7) appears to be in a stabilizing boat conformation with interaction between the amide and oxoammonium which differs from its solid-state structure.³⁸ We believe that the boat conformation and pendant amide of 30 play a crucial role to lower the energy barrier for hydride transfer and provide substrate stabilization. Use of TEMPO-BF₄ 31 (Figure 6A) for desaturation of 2 resulted in lower yields, presumably due to lack of the pendant amide. Moreover, an n(O)-to-σ*(C-H) interaction of 2 (Figure 7) should weaken the BDE of the β C-H bond, further lowering the energy barrier for the hydride transfer process.

In order to achieve the asymmetric synthesis of 1, acylative kinetic resolution³⁹ (AKR) using (2S, 3R)-Hyper-BTM catalyst 38^40,41 was next evaluated (Scheme 3). While direct AKR on 1 gave poor enantioselectivity (70:30 er), AKR on 2 using isobutyric anhydride showed optimal enantioselectivity (50 % conversion, 42 % yield, >99:1 er) to afford (−)-40 and (+)-2. Unfortunately, (−)-ester 34 gave a 1:3 ratio of (−)-2 and (−)-3 after mild saponification conditions. Fortunately, we found that (−)-34 and (−)-40 were viable desaturation substrates using oxoammonium salt 30. Finally, (−)-1 was saponified using K₂CO₃/MeOH⁷ (Figure 8). Interestingly, desaturation of (+)-2 using 30 at 85 °C in DCE underwent partial racemization while the same reaction at 40 °C in MeCN completely retained enantiopurity. Desaturation of allylic esters (−)-34 or (−)-40 did not lead to racemization at either 40 °C or 85 °C.

Scheme 3.

Initial screening of conditions for AKR using Hyper-BTM catalyst 38.

Figure 8.

Asymmetric synthesis of 1 via AKR.

Although the detailed mechanism is unknown, we believe that at high temperatures, excess Bobbitt’s salt 30 may form an adduct (+)-43 with (+)-1 which may be followed by thermal retro-6π−6π-electrocyclization via triene 44 which partially racemizes substrate 1 (Figure 9A). A proposed, stabilized alcohol-Bobbitt’s salt adduct has been reported by the Rutjes group.⁴² A DFT model (Figure 9B) showed that adduct (−)-43 is stabilized by intramolecular hydrogen bonding. We also conducted a ¹H NMR experiment by mixing 1 and 30 in CD₃CN which clearly showed the disappearance of the allylic hydroxyl signal of 1 after treatment with 30.⁹

Figure 9.

A. Proposed mechanism for racemization of 1 with 30. B. DFT model of (−)-43 ((r²SCAN-3C/CPCM (CH₂Cl₂)).

In order to probe the selectivity for AKR of the enantiomers of substrate 2, we also conducted a computational study (r²SCAN-3C/CPCM (CH₂Cl₂)) for transition states of isobutyrylated hyper-BTM catalyst 45 with (−)-2 and (+)-2, respectively. A transition state model (TS-C) of 45 with (−)-2 clearly showed an n-to-cation interaction⁴³ of the chromenone oxygen to 45 (distance 3.15 Å, Figure 10A) while the transition state model (TS-D) of 45 with (+)-2 has poor alignment due to steric hindrance. Only a π-to-cation interaction of the alkene of (+)-2 to 45 is allowed in the latter transition state (distance 3.17 Å, Figure 10B). As a result, the energy barrier for acylation and formation of (−)-40 is 14 kcal/mol lower than that of (+)-40.

Figure 10.

A. Transition state TS-C from 45 and (−)-2. B. Transition state TS-D from 45 and (+)-2. (r²SCAN-3C/CPCM (CH₂Cl₂)).

Finally, we evaluated the chemical dimerization of nidulalin A 1 under a variety of conditions (Scheme 4). We began our screening using (±)-1 as substrate. Photoirradiation (blue or white LED) or water-freezing induced high pressure treatment⁴⁴ of 1 in either in solution or neat resulted in recovery of starting material (Scheme 4, entry 5). After extensive experimentation, we found that thermolysis of a neat sample at 120 °C (melting point of 1) afforded approximately 5% of rac-nidulaxanthone A 5 with significant decomposition observed to xanthone 46 and a mixture of aldehydes 47 (Scheme 4, entry 6). Thermolysis of 1 in solution using solvents such as toluene and water resulted in either no reaction or production of trace dimer 5 along with substantial decomposition (Scheme 4, entry 1). Treatment 1 with Lewis acids (e.g. Sc(OTf)₃ and Cu(OTf)₂) resulted in either no reaction or production of nidulalin B 4 via aromatization (Scheme 4, entry 2). Use of basic conditions (e.g. Et₃N or tBuOK, 23 °C) resulted in formation of xanthone 46 and nidulalin B 4 (Scheme 4, entry 4). The intolerance of 1 towards basic conditions to produce 46 has been reported.⁴⁵ We believe that 46 is generated from a decarboxylative dehydration process; a mixture of aldehydes 47 may be generated from retro 6π-electrocyclization which is similar to a thermal racemization process. The formation of 1:1 mixture of aldehyde 47 is likely due to thermal cis-to-trans alkene isomerization. In both cases, rearomatization is the driving force towards production of such undesired byproducts.

Scheme 4.

Chemical dimerization studies of 1.

Given our unsuccessful attempts to dimerize rac-1 in reasonable yield, dimerization of enantioenriched-1 which was derived from AKR experiments was next evaluated. In the event, 1.5 h thermolysis of neat (+)-1 (80:20 er) at 100 °C under argon for 3 cycles (dissolved in CH₂Cl₂ and concentrated in vacuo) resulted in the production of (+)-5 in 26 % yield (46 % brsm, 99:1 er), together with aldehyde 47 (8.6 %) and recovered 1 in 40 % yield (Scheme 4, entry 7). Increasing the reaction time only resulted in decomposition to 47 and unidentified byproducts. Use of enantioenriched monomers (−)-1 or enantiopure (+)-1 resulted in a very similar yield of dimer 5 (Scheme 4, entries 8 & 9). In all Diels-Alder dimerization attempts with 1, nidulaxanthone 5 was the only dimeric product that was isolated.

To understand how the dimerization of nidulalin A (1) occurs, the transformation was modeled with wB97XD/6–31G* density functional theory (DFT).⁴⁶ We sought to understand both the dimerization regiochemistry and the relative energetics of homochiral vs. heterochiral monomer combinations. As represented in Figure 11, the observed dimerization results from homochiral [2+4] cycloaddition of the C2-C3-double bond of the dienone group to positions C3 and C9a in a second molecule of nidulalin A (1). For this Diels-Alder dimerization, standard methods such as frontier MO theory or the electron transfer model we reported earlier⁴⁷ were not expected to provide clarity as there is no inherent polarity difference between reactants. It is logical to expect preferred bonding between termini of the dienone π bonds; the usual endo rule should also provide guidance. When we modeled the different regio- and stereochemically distinct reaction modes, the lowest energy transition state (TS-E) fits these expectations and leads to niduaxanthone A. This transition state TS-E features face selectivity anti to the ester groups on each monomer. Moreover, the computed transition state (Figure 11) was found to be bis-pericyclic, with nearly perfect C₂ symmetry.

Figure 11.

Transition state TS-E for bis-pericyclic dimerization.

Following the initial report by Caramella and co-workers in 2002,⁴⁸ a growing number of pericyclic processes, most often [2+4] dimerizations, but also [6+4] cycloadditions, have been predicted to be bis-pericyclic.^49–56 In a bis-pericyclic transition state, the role of diene and dienophile become ambiguous and the structure is a hybrid of [2+4] and [4+2] reaction modes. Bispericyclic [2+4] reactions may be coupled to a structurally related Cope rearrangement in the same region of reaction space.⁵⁷ Figure 11 shows a 2D representation and a 3D model for the computed bis-pericyclic transition state for dimerization of nidulalin A 1. The structure has near-C₂ symmetry, resisting multiple efforts to locate a perfect C₂ geometry. The 3–3 bond is short (2.06 Å) and the 2–9a and 9a-2 bonds are longer at 2.97 and 2.98 Å, respectively. This structure closely resembles other dimeric bis-pericyclic transition states.^49–56 Other competitive dimerization modes were considered for 1 and found to be of higher energy; results are summarized in Supporting Information.⁹

As our calculations support the favorability of a homochiral dimerization, we next explored whether a heterochiral [2+4] reaction might be competitive. Dimerization of a pair of enantiomers by a bis-pericyclic process would require a sterically congested structure close to C_s symmetry. Among the different reaction modes,⁹ energetics of the lowest energy heterochiral transition state TS-F are shown in Figure 12. This is a conventional [2+4] cycloaddition process with the same regiochemistry as TS-E but with exo-stereochemistry. This results in a substantially higher barrier, with the expected formation of heterochiral dimer 48.

Figure 12.

Energetics of homochiral and heterochiral dimerization modes.

The imaginary vibrational mode for bis-pericyclic TS-E is atypical for a Diels-Alder cycloaddition, with animation (see the Supporting Information file dimerization.zip) showing primarily 3–3’ bond formation. Consistent with this observation, calculation of the forward reaction coordinate (IRC)⁹ from TS-E did not point directly to nidulaxanthone A 5, but instead proceeded to a geometry very close to the expected Cope rearrangement. Transition state optimization at this point easily gave TS-G (Figure 12). These results support the existence of a valley-ridge inflection point, wherein the cycloaddition coordinate intersects an orthogonal lower-energy and fully degenerate Cope rearrangement. A similar connection between bis-pericyclic and Cope transition states has been noted for dimerization of 1,3-cyclopentadiene.⁴⁸

Our computational studies thus led to several important conclusions. First, the dimerization of nidulalin A 1 to nidulaxanthone A 5 should proceed through a concerted bis-pericyclic transition state (TS-E) with near C₂ symmetry. The bis-pericyclic character likely provides a small energetic advantage. Synthesis of 5 thus joins a growing list of bis-pericyclic Diels-Alder dimerizations.^49–57 Coupling to a lower energy Cope rearrangement transition state (TS-G) is also supported by our results. Second, our calculations support the experimental results that 5 is the only dimer product from homochiral nidulalin A 1. This outcome is due to better fit of the reaction partners in the transition state which requires the CO₂CH₃ groups on respective monomers to be anti.

In the thermolysis of 1, we found that (±)-1 partially melted above 115 °C, while (+)- or (−)-1 was partially melted above 95 °C. Moreover, monomer 1 was found to be unstable above 100 °C and readily decomposed to unsaturated aldehydes 47. X-ray crystal structure analysis of (±)-1 showed a favored centrosymmetric packing⁹ versus a noncentrosymmetric lattice in the published single enantiomer crystal structure.⁷ We believe that breaking the centrosymmetric lattice of (±)-1 requiring higher temperature than the that of enantiopure 1, together with an unfavorable heterochiral [2+4] cycloaddition process to 48 provides an explanation for the fact that only dimerization of enantioenriched 1 affords nidulaxanthone 5 in reasonable yield.

CONCLUSION

In conclusion, we have developed a four-step, de novo synthesis of (±)-nidulalin A from a chromone ester substrate. Key steps in the process include allyl triflate activation of a chromone ester substrate and desaturation of a γ,δ-unsaturated ketone using Bobbitt’s oxoammonium salt. We have also probed the mechanism of the desaturation process computationally which elucidated a concerted, asynchronous hydride transfer process with Bobbitt’s salt in a boat conformation. The asymmetric synthesis of nidulalin A was achieved by acylative kinetic resolution (AKR) of chiral, racemic 2H-nidulalin A using a Hyper-BTM catalyst. We also achieved dimerization of chiral, non-racemic nidulalin A to nidulaxanthone A as the only dimeric product produced under thermolytic conditions. A computational study revealed a C₂-symmetric, bis-pericyclic transition state for dimerization which agreed with thermolytic dimerization experiments. Further studies on the chemistry of nidulalin A as well as biological profiling of natural products and targeted derivatives are currently in progress and will be reported in due course.