Total Synthesis and Target Identification of the Curcusone Diterpenes

Abstract

The curcusone natural products are complex diterpenes featuring a characteristic [6-7-5] tricyclic carbon skeleton similar to the daphnane and tigliane diterpenes. Among them, curcusones A-D demonstrated potent anticancer activity against a broad spectrum of human cancer cell lines. Prior to this study, no total synthesis of the curcusones was achieved and their anticancer mode of action remained unknown. Herein, we report our synthetic and chemoproteomics studies of the curcusone diterpenes that culminate in the first total synthesis of several curcusone natural products and identification of BRCA1-associated ATM activator 1 (BRAT1) as a cellular target. Our efficient synthesis is highly convergent, builds upon cheap and abundant starting materials, features a thermal [3,3]-sigmatropic rearrangement and a novel FeCl₃-promoted cascade reaction to rapidly construct the critical cycloheptadienone core of the curcusones, and led us to complete the first total synthesis of curcusones A and B in only 9 steps, C and D in 10 steps, and dimericursone A in 12 steps. The chemical synthesis of dimericursone A from curcusones C and D provided direct evidence to support the proposed Diels-Alder dimerization and cheletropic elimination biosynthetic pathway. Using an alkyne-tagged probe molecule, BRAT1, an important but previously “undruggable” oncoprotein, was identified as a key cellular target via chemoproteomics. We further demonstrate for the first time that BRAT1 can be inhibited by curcusone D, resulting in impaired DNA damage response, reduced cancer cell migration, potentiated activity of the DNA damaging drug etoposide, and other phenotypes similar to BRAT1 knockdown.

Table of Contents

INTRODUCTION

Natural products have been valuable sources and inspirations of lifesaving drug molecules.¹ Their accumulated evolutionary wisdom together with their structural novelty and diversity make them unparalleled for novel therapeutic development. However, their natural scarcity, structural complexity, and unknown mode of action often hamper their further development in the drug discovery pipeline. Total synthesis has been employed as one way to solve the material supply and chemical probe synthesis for comprehensive biological profiling and target identification.² Meanwhile, many biologically validated disease targets are considered as “undruggable” from a chemical standpoint due to the lack of enzymatic activity and/or small molecule binding sites.³ The BRCA1-associated ATM activator 1 (BRAT1) protein has been validated as an oncogenic protein involved in various cancers but belongs to the “undruggable” category with no known small molecule inhibitors to date. Herein, we report a collaborative effort in the total synthesis and target identification of the curcusone natural products which yielded the first total synthesis of curcusones A and B in 9 steps, C and D in 10 steps, and dimericursone A in 12 steps and revealed BRAT1 as a key cellular target of the curcusones.

The curcusone diterpenes (Figure 1A) were isolated from Jatropha curcas, a widely used ingredient in traditional remedies for a variety of ailments including cancer. Structurally, they share a characteristic [6–7-5] tricyclic carbon skeleton with the daphnane and tigliane diterpenes.⁴ Curcusones A-D (1a-d), isolated by Clardy and co-workers in 1986, were unambiguously identified as two epimeric pairs at the C2 position.⁵ Since then, around thirty curcusone molecules have been isolated including curcusones F-J,⁶ which lack the dienone moiety in the seven-membered ring. Structurally rearranged analogs like spirocurcasone (3) and dimeric analogs such as dimericursone A (2a) and dimericursone B (2b) were discovered recently.^7,8 Among them, curcusones A-D (1a-1d) exhibited low micromolar IC₅₀ values against a broad spectrum of human cancer cell lines.⁶ However, no total syntheses of curcusones A-D (1a-1d) and their dimeric products (2a and 2b) were reported prior to this study and their mode of action remained unknown.

Figure 1.

Curcusone diterpenes and their synthesis.

While the closely related daphnane and tigliane diterpenes have attracted a significant amount of synthetic interest,⁹ the curcusone molecules have surprisingly received little attention despite their therapeutic potential. In 2017, we reported the first total syntheses of the putative structures of curcusones I and J (1i and 1j) in 21 steps (Figure 1B), ultimately leading to the conclusion that the originally proposed structures of both 1i and 1j were incorrect.¹⁰ Our synthesis involves a gold-catalyzed tandem furan formation and furan-allene [4+3] cycloaddition to build the 5,7-fused ring system with an oxa bridge and a Diels-Alder reaction to construct the 6-membered ring. In 2019, Stoltz and co-workers reported their studies toward synthesizing 1a-1d (Figure 1C).¹¹ Their approach features an elegant divinylcyclopropane-cycloheptadiene rearrangement to forge the 7-membered ring and reached advanced intermediate 14 after 12 steps from 8.

Our ongoing interest in natural products that can covalently modify cellular proteins¹² prompted us to continue pursuing the total synthesis and target identification of curcusones A-D (1a-1d) with an electrophilic cycloheptadienone moiety. This unique structural feature could allow them to form a covalent bond with nucleophilic residues of certain cellular proteins.¹³ Previous cytotoxicity studies found that reduction and/or oxidation of the C6-C7 double bond greatly reduced their anticancer activity.⁶ As such, an approach allowing variation of the C6 and C7 substituents would be highly desirable. We envisioned 15 as an advanced intermediate (Figure 1D). α-Halogenation followed by two methylation reactions would lead to 1a and 1b, which could be oxidized to 1c and 1d via α-hydroxylation. A ring closing metathesis (RCM) or an intramolecular aldol condensation was planned to form the 7-membered dienone (16/17→15). Both 16 and 17 could be prepared from vinyl triflate 18 via Pd-catalyzed carbonylative cross couplings. At this stage, disconnection of the C9-C10 bond could break 18 into two simple pieces, but a direct intermolecular C-C bond formation to construct such a bond is challenging. Thus, we opted for a Claisen rearrangement to forge this C-C bond in a stereoselective manner and designed 19 with a masked aldehyde as the Claisen rearrangement precursor. 19 could be assembled from simple building blocks 20a/20b and 21. The latter could be derived from a cheap chiral pool molecule (S)-(−)-perillaldehyde 8.

RESULTS

Total Synthesis.

Our synthesis started with preparing 23 (Figure 2), a known compound synthesized from 8 in three steps – extended silyl enol ether formation, vinylogous Mukaiyama aldol reaction, and NaBH₄ reduction.¹⁴ We combined the first two steps into a one-pot reaction; crude 22 was then subjected to NaBH₄ reduction directly to produce multi-decagram scale of 23 in one batch. We next needed to prepare 24 for the Claisen rearrangement. NaH-promoted addition-elimination between 23 and 20b afforded 24, albeit in low yield (35%). We then used a Mitsunobu reaction between 23 and 20a to prepare 24, but the hydrazine byproduct derived from diethyl azodicarboxylate could not be separated from 24. The recently reported redox-neutral organocatalytic Mitsunobu conditions were also explored but failed to provide 24.¹⁵ Fortunately, the hydrazine byproduct could be tolerated in the Claisen rearrangement. After 24 was heated at 140–150 °C in DMF for 18 h, the Claisen rearrangement did occur, but the rearranged product 25 further cyclized to provide tricyclic compound 26 (CCDC 2033828) as a single diastereomer in 48% yield from 23.

Figure 2.

Total syntheses of curcusones A-D, dimericursone A, and their analogs.

To avoid further cyclization with the acetal group, we sought to use a more stable protecting group for the aldehyde and converted 23 to thioacetal 38 (Figure 3). With a more stable thioacetal, TsOH-catalyzed addition-elimination could be used to prepare the new Claisen rearrangement precursor 39 in 79% yield. Interestingly, upon microwave-heating 39 in DMF at 130 °C, in addition to the desired rearrangement product 40 (24% yield), 29% yield of byproduct 41 was obtained and an equal amount of 39 was recovered. The structure of 41 was unambiguously established by X-ray crystallography analysis (CCDC 2055698). Mechanistically, 41 was produced from 39 via a remarkable tandem Claisen rearrangement and thermal Conia ene reaction. We tried to suppress the formation of 41 by shortening the reaction time or lowering the reaction temperature, but these efforts led to overall decreased yield of 40.

Figure 3.

Investigation of the Claisen rearrangement

We then decided to continue with 26 and explore the hidden cyclopentane-1,3-dione symmetry to synthesize 17. We started with investigating 1,2-addition of lithiated ethyl vinyl ether (27) to 26 theorizing that a global hydrolysis would release the methyl ketone and the aldehyde at once to form 17 for the aldol condensation. This 1,2-addition turned out to be nontrivial. When two equiv. of 27 was used, only less than 10% yield of 28 was obtained. Owing to their oxophilicity, cerium chloride and lanthanum chloride have been used to promote 1,2-additions.¹⁶ Unfortunately, both failed in our case. Eventually, the 1,2-addition was improved by increasing the amount of 27 to 10 equiv.,¹⁷ and 28 was prepared in 57% yield. 28 was then subjected to hydrolysis upon the treatment with TsOH and 17 was obtained in 51% yield from 26. Meanwhile, we were delighted to observe the formation of 15 in the same reaction, albeit in very poor yield (<5%). We were encouraged to achieve a global hydrolysis/aldol condensation cascade to synthesize 15 from 28 in one step and identified FeCl₃¹⁸ in combination with TMSCl as the optimal conditions. When crude 28 was treated with a premixed FeCl₃ (0.2 M in 2-methyltetrahydrofuran) and TMSCl in toluene at room temperature, desired product 15 was produced in 32% yield from 26 together with 21% of 30b. The failure of converting 17 to 15 with FeCl₃ and the isolation of 30b led us to propose that the FeCl₃/TMSCl-promoted cascade process went through intermediates 29 and 30a to form 15 in one step. In intermediate 30a, the α-H_a and the EtO group are anti-periplanar, thus the subsequent 1,2-elimination is facile and occurred under the FeCl₃/TMSCl conditions. In intermediate 30b, α-H_b and the EtO group are both in the pseudo equatorial positions, which was supported by NMR analysis and computational modeling. Therefore, its conversion to 15 is more difficult. In order to maximize the overall yield of 15, we investigated conditions to promote the 1,2-elimination of 30b and identified that it could be converted to 15 in 65% yield by heating with TsOH in toluene at 50 °C, which brings the overall yield of 15 from 26 to 46%.

With the [6–7-5] tricyclic carbon skeleton quickly assembled in only six steps, we next needed to introduce the two methyl groups. Johnson α-iodination converted 15 to iodoenone 31, which was surprisingly unstable. Therefore, after a quick workup, crude 31 was immediately subjected to the next Stille cross coupling with tetramethylstannane to provide 32 in 57% yield over two steps. Finally, α-methylation of enone 32 at C2 position delivered a 1:1 mixture of separable (−)-curcusone A (1a) and (−)-curcusone B (1b) in 63% yield (9 steps total). α-Hydroxylation of 1a with KHDMS and MoOPH gave separable (−)-curcusone C (1c) and (−)-curcusone D (1d) in 63% yield (d.r. 1:1; 84% brsm; 10 steps total). Additionally, in order to obtain analogs for biological activity comparison, we converted (−)-1b to (+)-spirocurcasone (3) and (−)-33 (a synthetic derivative named as pyracurcasone) by following a reported one-step procedure.⁷ The ¹H, ¹³C NMR, and other analytic data of our synthetic samples matched well with the reported ones, which also conclude that the absolute configuration of 1a-1d assigned by Clardy et al. in 1986 is opposite of the actual ones.

We then set out to synthesize dimericursone A (2a) from 1a-1d via a biomimetic dimerization. The proposed biosynthesis of 2a consists of a sequence of oxidative dehydrogenation of 1a/1b or dehydration of 1c/1d to form a reactive cyclopentadienone intermediate followed by Diels-Alder dimerization and cheletropic extrusion of carbon monoxide.⁸ From there, 2a could be converted to 2b via another oxidative dehydrogenation and double bond isomerization. We started with 1c and 1d. After an unfruitful attempt to synthesize 2a by heating them directly at elevated temperatures, a 1:1 mixture of them was first converted to their mesylates (34). After extensive exploration, we identified that using triethylamine as a base in 1,4-dichlorobenzene at 150 °C (−)-2a was produced in 18% yield over two steps, which provides direct evidence to support the proposed biosynthetic pathway. Under our conditions, the formation of 2b was not observed.

Probe Synthesis.

To elucidate the anticancer mechanism and identify potential cellular targets of curcusones, an alkyne-tagged probe molecule 37 was designed for chemoproteomics studies. Since the dienone is likely protein-reactive and is critical for the observed activity, we decided to minimize structural perturbation of this part and used the tertiary alcohol as a handle to link with a terminal alkyne. 37 was synthesized in 59% yield from (−)-1d via a DCC-promoted coupling with 36.

Cytotoxicity and Target Identification.

We evaluated the cytotoxicity of curcusones and their analogs in breast cancer MCF-7 cells using the WST-1 assay (Figure 4A; see Table S1 for structures and EC50 values). Synthetic 1a-1d, natural 1b and 1d, and intermediates 15 and 32 exhibited micromolar EC₅₀ values against MCF-7 cells with 1d being the most potent curcusone. Importantly, the cytotoxicity values for synthetic 1b and 1d were virtually identical to the values of their naturally isolated counterparts. Analog 33 showed slightly better antiproliferation activity indicating the feasibility of finely tuning the cycloheptadienone moiety to improve potency, but 2a was not active even at 100 μM. Likely due to the full confluency of the tested MCF-7 cells, the EC₅₀ values we obtained were about one order of magnitude higher than previously reported (1.6–3.1 μM EC₅₀ values for 1a-1d).⁶ Gratifyingly, 37 retained similar anticancer properties of 1d, thus warranting its use in competitive chemoproteomic studies.

Figure 4.

Identification of BRAT1 as a cellular target of 1d. (A) Viability of MCF-7 cells following 24 h compound treatment with EC₅₀ values in brackets (n = 3). (B) Scatter plot of proteins competed by 1d from probe 37 (10 μM) enrichment in MCF-7 cells with BRAT1 highlighted (n = 3). (C) Western blot (top) and quantification (bottom) of thermal shift assay of overexpressed FLAG-BRAT1 lysate treated with 1d. (D) Western blots of BRAT1 following treatment of live cells with 1d and enrichment by 37 in competitive pulldown assays. (E) Western blot (top) and quantification (bottom) of pulldown of BRAT1 with 37 in HeLa cells and dose-dependent competition with 1d. All error bars are S.D.

We then identified the cellular targets of curcusones by competitive chemoproteomics using probe 37. MCF-7 cells were treated with 1d or DMSO for 4 hours followed by lysis, treatment with 37, CuAAC with biotin azide, enrichment, digestion, and LC-MS/MS analysis using label-free quantification (Figure 4B and Table S2). The best competed target was BRAT1, which acts as a master regulator of the DNA damage response (DDR) and DNA repair by binding to BRCA1 and by activating DDR kinases such as ATM and PRKDC (DNA-PKcs) following DNA damage.^19–21 Knockdown of BRAT1 increased the constitutive level of apoptosis in human osteosarcoma cells¹⁹ and decreased cancer cell proliferation and tumorigenicity in vitro and in mouse tumor xenografts.²¹ BRAT1 is also an unfavorable prognostic marker in kidney and liver cancers.²² Therefore, targeting BRAT1 is a promising strategy for cancer treatment.

We next characterized the physical interaction between 1d and BRAT1. We overexpressed FLAG-BRAT1 in HEK-293T cells and performed a thermal shift assay by treating lysates with 1d, heating as indicated, and probing the remaining soluble FLAG-BRAT1 by Western blotting (Figure 4C). We observed thermal destabilization of 1d-treated BRAT1 indicating a direct interaction. To validate endogenous BRAT1 as a target of 1d in live cells, we employed a competitive pulldown experiment. MCF-7 cells were treated with 1d or DMSO for 4 hours before lysis, treatment with probe 37, CuAAC with biotin azide, streptavidin enrichment, elution, and Western blot visualization. Indeed, native BRAT1 was enriched by 37 and was competed by 1d (Figure 4D). Additionally, 1d competed the enrichment of BRAT1 from cervical cancer HeLa and triple negative breast cancer MDA-MB-231 cells, thus validating native BRAT1 as a cellular target of 1d across these cell lines. In situ treatment of 1d in live HeLa cells competed BRAT1 enrichment by 37 at low micromolar concentrations (EC₅₀ = 2.7 μM; Figure 4E). To evaluate the binding reversibility, we performed a competitive BRAT1 pulldown assay with the irreversible cysteine-reactive probe iodoacetamide (IAA). HeLa lysate was treated with 37 (10 μM) for 1 h with either pre- or posttreatment with IAA (30 mM) for 1 h followed by enrichment and Western blot analysis as described above (Figure S1A). Indeed, both pre- and posttreatment with IAA competed BRAT1 enrichment by 37, indicating that the curcusone probe forms a covalent bond with a cysteine in BRAT1 and this bond is reversible. We further treated live HeLa cells with different concentrations of 1d for 4 h to evaluate BRAT1 target occupancy via our competitive pulldown assay, which revealed that 1d reaches 50% fractional occupancy (f occ. (50)) at 2.7 μM concentration (Figure 4E). Collectively, these results demonstrate that 1d is the first small-molecule binder of BRAT1.

BRAT1 Modulation.

To determine whether 1d inhibits BRAT1 in cells, we generated stable BRAT1 KD HeLa cells via shRNA retroviral transduction (Figure S1B). We then compared the protein expression profiles of BRAT1 KD cells versus 1d-treated cells (3 μM, 24 h) by global proteomics analysis (Figure 5A–C and Table S3). Among 3347 quantified proteins in compound-treated cells, we found only 36 up- and 42 down-regulated proteins. Importantly, 31 of the 78 dysregulated proteins were also dysregulated in BRAT1 KD cells, thus indicating that 1d functionally inhibits BRAT1 in cells. Notably, several well-known cancer migration and progression drivers were downregulated (Figure 5B), including TRIM47 which mediates cancer migration,²³ the bona fide oncoprotein and potential biomarker WBP2,²⁴ and frequently highly amplified oncogene FNDC3B.²⁵ None of these proteins have previously been functionally linked to BRAT1. We then investigated the effect of 1d treatment and BRAT1 KD on cancer cell migration in WT and BRAT1 KD HeLa cells, as well as WT MCF-7 and MDA-MB-231 cells (Figure 5D–G). As expected, BRAT1 knockdown greatly diminished migration of HeLa cells, and treatment with 1d at 1 μM concentration also reduced migration of all cell lines by ∼4-fold.

Figure 5.

Evaluation of BRAT1 inactivation by 1d. (A) Global proteomic analysis of HeLa cells comparing 1d treatment to BRAT1 knockdown versus DMSO or non-targeting (WT) controls, respectively. Color dots indicate proteins which are downregulated in both conditions (n = 3). (B) Table showing common downregulated proteins involved in cell migration and, (C) in DNA repair. (D) Crystal violet images and, (E) quantification of cells in a transwell migration assay following 1d (1 μM; 24 h) treatment (n = 3). (F) Crystal violet images and, (G) quantification of WT and BRAT1 KD cells (24 h) in a transwell migration assay (n = 3). (H) Fluorescence microscopy images and, (I) quantification of WT or BRAT1 KD cells treated with DMSO, 1d (3 μM), etoposide (30 μM), or both compounds showing γH2AX staining (n = 3). (J) Images of cells (top) and quantification (bottom) after 24 h treatment with DMSO, 1d (3 μM), etoposide (50 μM), or both compounds or, (K) following BRAT1 knockdown and subsequent treatment with etoposide. All error bars are S.D. All scale bars indicate 30 μm. ***P<0.001, ****P<0.0001, Welch’s two-sided t-test.

Our global proteomics experiment also revealed several commonly downregulated key DNA repair proteins (Figure 5C) such as (i) POLD1 which synthesizes DNA during repair,²⁶ (ii) USP47 which facilitates base-excision repair,²⁷ (iii) FANCI which mediates the repair of DNA double strand breaks and interstrand crosslinks,²⁸ and (iv) BRCC3 which stabilizes the accumulation of BRCA1 at DNA breaks.²⁹ These proteins have not been previously linked to BRAT1 either. Most notably, 1d treatment (24 hours) significantly downregulated the actual physical target, BRAT1 (ratio 0.18), as confirmed by Western blotting (Figure S1C). To evaluate whether this effect is due to proteasomal degradation or altered gene expression, we added the proteasome inhibitor MG132 for the final 4 hours of 1d treatment, which recovered BRAT1 protein levels. In contrast, when we measured BRAT1 mRNA levels by RT-qPCR following 1d treatment for 24 hours (Figure S1D), we found no difference. Collectively, these findings demonstrate the importance of BRAT1 as a master regulator of the DDR and that 1d inhibition of BRAT1 in cells induces proteasomal degradation over time.

We then investigated whether 1d would potentiate the DNA damaging effect of the clinical drug and topoisomerase inhibitor etoposide via BRAT1 inhibition. WT or BRAT1 KD HeLa cells were treated with DMSO, etoposide, 1d, or etoposide and 1d combined. Subsequent DNA damage was then measured by fluorescence microscopy using γH2AX staining (Figure 5H, I and Figure S1E). Treatment with 1d (3 μM) or KD of BRAT1 alone did not increase γH2AX signal. However, co-treatment of 1d with etoposide led to a 2-fold increase. Similarly, etoposide treatment significantly increased γH2AX signal in BRAT1 KD cells, recapitulating the 1d/etoposide co-treatment results. Importantly, 1d treatment did not increase γH2AX signal in etoposide-treated BRAT1 KD cells, confirming that the 1d-etoposide synergism is linked to BRAT1 inactivation. Furthermore, co-treatment of 1d with etoposide also increased cytotoxicity in HeLa, MCF-7, and MDA-MB-231 cells (Figure 5J). Likewise, there was increased cell death in BRAT1 KD HeLa cells following etoposide treatment relative to WT cells (Figure 5K). Altogether, these results demonstrate that targeting BRAT1 with 1d is a promising anticancer strategy for chemosensitization to DNA damaging drugs.

In summary, we completed the first asymmetric total synthesis and target identification of the curcusone natural products. Our convergent synthesis builds upon a cheap and abundant chiral pool molecule (8) and features a thermal [3,3]-sigmatropic rearrangement and an FeCl₃-promoted global hydrolysis/aldol condensation cascade to rapidly construct the critical cycloheptadienone core. This efficient synthetic route yielded curcusones A and B (1a and 1b) in only 9 steps, curcusones C and D (1c and 1d) in 10 steps, and dimericursone A (2a) in 12 steps from (S)-(−)-8. The successful synthesis of 2a from 1c/1d experimentally supports the proposed Diels-Alder dimerization and cheletropic extrusion biosynthesis. By performing chemoproteomics with the alkyne probe 37, we identified the previously “undruggable” oncogenic protein BRAT1 as a key cellular target of 1d. Furthermore, 1d inhibits BRAT1 in cancer cells, thereby reducing cancer cell migration, increasing susceptibility to DNA damage, and inducing chemosensitization to the approved drug etoposide. To our knowledge, 1d is the first known small-molecule inhibitor of BRAT1, a master regulator of the DDR and DNA repair. Many promising clinical trials are underway targeting DDR proteins such as PARP, ATR, ATM, CHK, and DNA-PK as monotherapies or in combination with other treatments.³⁰ Olaparib, a PARP inhibitor, was approved by FDA in 2014 as a monotherapy to treat germline BRCA1/2-mutant ovarian cancer.^30a Our concise and convergent total synthesis has opened the gate for curcusone analog synthesis and structure-activity optimizations, which may thus yield novel BRTAT1 inhibitors as potential lead medicines for monotherapies or combination therapies.