Modular total synthesis and cell-based anticancer activity evaluation of ouabagenin and other cardiotonic steroids with varying degrees of oxygenation.

Abstract

A Cu(II)-catalyzed diastereoselective Michael/aldol cascade approach is used to accomplish concise total syntheses of cardiotonic steroids with varying degrees of oxygenation including cardenolides ouabagenin, sarmentologenin, 19-hydroxysarmentogenin, and 5-epi-panogenin. These syntheses enabled the subsequent SAR studies on 37 synthetic and natural steroids to elucidate the effect of oxygenation, stereochemistry, C3-glycosylation and C17-heterocyclic ring. Based on this parallel evaluation of synthetic and natural steroids and their derivatives, glycosylated steroids cannogenol-L-α-rhamnoside (79a), strophanthidol-L-α-rhamnoside (92), and digitoxigenin-L-α-rhamnoside (97) were identified as the most potent steroids demonstrating broad anticancer activity at 10–100 nM concentrations and selectivity (nontoxic at 3 μM against NIH-3T3, MEF and developing fish embryos). Further analyses indicate that these molecules show a general mode of anticancer activity involving DNA damage upregulation that subsequently induces apoptosis.

Graphical Abstract

1. INTRODUCTION

Cardiotonic steroids are a group of natural products that include cardenolides and bufadienolides.^1,2 Around a thousand different natural cardenolides and bufadienolides have been isolated to date from plants and animals. In addition to variations in the C17 heterocycle and C3 glycosylation, structural variations arise due to the presence of multiple oxygenation sites and changes in the skeletal and substituent stereochemical configurations (Figure 1).²

Figure 1.

Examples of variations in the oxidation state and stereochemistry in cardiotonic steroids

The majority of cardenolides and bufadienolides possess cardiotonic activity, which is attributed to their ability to inhibit Na⁺/K⁺ATPase pumps within cardiac myocytes.^1,3 Recent findings also indicate additional signaling functions, e.g., regulation of renal sodium transport and arterial pressure as well as cell growth, differentiation, apoptosis, fibrosis, modulation of immunity and carbohydrate metabolism, and the control of various central nervous functions and behaviors.³ Cardiotonic steroids have also been investigated as potential therapeutic agents with anti-cancer,⁴ anti-viral,⁵ anti-bacterial,⁶ immunoregulatory,⁷ neural outgrowth differentiative,⁸ anti-inflammatory⁹ and anti-hypertensive¹⁰ properties. Notably, overwhelming evidence indicates cardiotonic steroids as potential anti-cancer therapeutics.^4a,11

However, as with most electrophilic compounds, the selectivity of cardiotonic steroids for transformed cells and also their general utility for cancer therapy is debated. This question has proven difficult to answer because various model systems/comparisons used to assess therapeutic value have been criticized.¹² The mode(s) of action of cardiotonic steroids also remain poorly-understood. Many studies agree that these molecules induce one of the most common cell death pathways, apoptosis^13a,b,c; however, modes other than apoptosis have also been proposed, including: autosis, ^13d autophagy-related, caspase-independent cell death; and anoikis, ^13e a caspase-independent loss of adherence leading to cell death. Modes of action may also be cell-type specific. Reversible quiescence, rather than death, was reported in neuroblastomas treated with ouabain.¹⁴ Despite all compounds sharing a similar core structure, different modes of action have been proposed for different cardiotonic steroids. Even for the most-commonly proposed mode of cardiac steroid-induced activity, apoptosis, numerous mechanisms have been put forward, e.g., inhibition of Na⁺/K⁺-ATPase (inactivation of which is linked to apoptosis, especially in cancer cells^13d); inhibition of HIF1a¹⁵ transcription (an effect that is independent of Na⁺/K⁺-ATPase); and DNA damage.

DNA damage is an appealing mode of action that is used by numerous anticancer drugs. However, there is considerable lack of certainty surrounding this behavior. Several studies describe that cardiac steroids are adjuvants to DNA-damage inducers, including radiation,¹⁶ and small-molecule genotoxins.^16a In mice, the synergy of cardiac steroids with the toxicity of DNA-damage inducers requires an active immune system, consistent with immunogenic cell death¹⁷ (that cannot occur in typical cell culture). But it is difficult to see how such behavior links directly to cardiotonic steroid-specific toxicity. Interestingly, a few studies have reported DNA-damage upregulation independent of additional DNA-damage stimulation.¹⁴ One study reports DNA-damage-inducing effects in K562 cells (a BCR-ABL-positive leukemia line) treated with only digitoxin. However, these effects were transient,¹⁸ indicating that DNA damage was not the principal cause of toxicity. Furthermore, several DNA-damage markers are upregulated during apoptosis,¹⁹ rendering it difficult to assign cause and effect without careful controls. Moreover, few studies have compared outputs between sensitive and resistant lines, thus it is unknown which effects are on pathway. Thus, several outstanding questions remain: (1) Do cardiac steroids target a common pathway? (2) Do these compounds have selectivity for transformed versus non-transformed lines? (3) Is apoptosis the key pathway in cardiac steroid toxicity? (4) Is DNA damage involved in death and does this precede apoptosis or is it a consequence? (5) Are these processes on-target or off-target? Finally, assuming favorable answers to these questions are reached, it would be helpful to establish a mechanistic probe that displays all the selectivity characteristics of cardenolides for downstream studies.

We propose that these key questions can be tackled by profiling toxicity/selectivity of a number of cardiotonic steroids across a panel of sensitive and insensitive cell lines. The SAR studies are scarce in terms of the structural features explored as they rely either on readily-available natural steroids or their semi-synthetic derivatives. Some aspects of SAR are known, e.g., C3 glycosylation is important for activity.²⁰ However, these conclusions do not always hold over numerous cell lines. A total synthesis-based route can potentially provide control over these parameters and thus enhance the medicinal exploration and optimization of cardiotonic steroids. However, the majority of the synthetic efforts towards the synthesis of cardiotonic steroids have focused on developing partial syntheses starting with androstane/pregnane derivatives.^21–23 Such routes are limited because readily-available androstanes and pregnanes often lack oxygenation at key positions of the steroid skeleton (i.e., at C1, C5, C11, C14, and C19). Some of these challenges were recently addressed by landmark studies on ouabagenin (1) and 19-hydroxysarmentogenin (4) by the Deslongchamps (Scheme 1a),²⁴ Baran (Scheme 1b)²⁵ and Inoue (Scheme 1c) laboratories.²⁶

Scheme 1.

Synthetic approaches to the highly oxygenated cardiotonic steroid, ouabagenin (1)

However, a general and flexible strategy to access cardiotonic steroids with various oxidation patterns is still highly desired. In our search for such a strategy, we focused on developing a convergent approach to steroidal cores 11 and 12 from relatively simple starting building blocks such as 8a or 8b. Rapid assembly of 11 and 12 and flexibility in adjusting the stereochemical configurations and oxidation state at the C5 and C11 positions in the subsequent manipulations are key in enabling the access to various cardenolides with varying degrees of oxygenation and with altered stereochemistry of the substituent- and ring-junction-bearing stereocenters (Scheme 1d).^27,28 These synthetic efforts are enabled by Cu(II)-catalyzed stereoselective Michael/aldol cascade reactions²⁷ allowing rapid assembly of functionalized steroidal skeletons 9 and 10, which are then transposed to fully functionalized precursors 11 and 12 (6–7 steps, gram scale)²⁸ In this article, we apply this method to the concise total synthesis of highly-oxygenated cardenolides, ouabagenin (1) and sarmentologenin (3), and less oxidized steroids 4–7, and numerous other analogs. With these compounds, we begin to address the critical biological questions above.

2. RESULTS AND DISCUSSION

We envisaged that a large subgroup of cardiotonic steroids containing C11 oxygenation could be accessed by a diastereoselective Cu(II)-catalyzed Michael reaction, providing rapid access to fully-functionalized steroidal intermediate with oxygenation at C3, 11, C14, C17 and C19 (Scheme 2A).²⁷ The oxygenated intermediate thus accessed could then divergently be elaborated to various natural products following C14/C19 hydroxyl-controlled transformations.

Scheme 2.

Diastereoselective Michael/aldol cascade reaction-based approach to ouabagenin (1) and related cardenolides

Reagents and conditions. (Scheme 2B): (a) 20 mol% Cu(OTf)₂, rt, 12h, then 30 mol% DBU, THF, reflux, 12h. (Scheme 2C): (a) acrolein, H₂O, 12 h, 97% yield; (b) 10 mol% (S)-2-(diphenyl((trimethylsilyl)oxy)methyl)pyrrolidine (18), (BzO)₂, hydroquinone, THF, H₂O, 1.5 h, 79% yield; (c) 1-(tripheylphosphoranylidene)-2-propanone, toluene, 75% yield; (d) LiHMDS, THF, −78 °C, 1h, then diethylcarbonate, rt, 12h, 73% yield; (e) 1 N HCl, THF: H₂O = 12:1, 6h, 84% yield; (f) PCl₃, CHCl₃, 0 °C to rt, 12h, 78% yield; (g) Cu(OTf)₂ (50 mol%), neat, rt, 12 h then add p-TSA (10 equiv), acetonitrile, 55 ºC, 72 h, 45% yield.

We first undertook model studies to test the requisite diastereo-induction at C9 and C10 imposed by the C11 stereocenter (Scheme 2B).^28a,29 The Cu(OTf)₂-catalyzed Michael reaction of β−ketoester 13 and C11 oxygenated enone 14 proceeded with high levels of stereocontrol, yielding steroid 16 (44 % yield over 2-steps, >20:1 dr). As discussed previously,^{27, 28a} the reaction likely proceeds via an open transition state 15 through a stepwise mechanism. An alternative mechanism involving a hetero-Diels–Alder reaction followed by the opening of the resultant ketal^29b is less likely because ketoester 20 reacted equally well with s-trans dienes (cyclohexenone) under similar conditions (SI).

Our subsequent studies focused on developing the asymmetric variant of this reaction employing a chiral version of enone 14 and β−ketoester 20 containing the pre-installed C3 oxygenation (Scheme 2C). The enantioselective synthesis of enone 14 commenced with commercially-available 1,3-diketone 17 that was subjected to: Michael reaction with acrolein; organocatalytic oxidation³⁰ with benzoyl peroxide/catalyst 18;³¹ and Wittig reaction with commercially-available 1-(tripheylphosphoranylidene)-2-propanone to provide chiral enone 14 in 57% yield (3 steps) and 95% ee on a 20-g scale. Similarly, β-ketoester 20, was obtained via acylation, acid-hydrolysis of ethyl-vinyl ether and chlorination.^28a The key Michael reaction/double aldol cyclization cascade reaction proceeded smoothly and subsequent heating of the reaction mixture in acetonitrile with p-TSA at 70 °C for an additional 72 hours led to clean formation of compound 21 (45% yield, 1g scale).^28a

We pursued an approach relying on converting steroid 21 into key framework 12, via intermediate 10 (Scheme 3). Our calculations of the single point energy values of 21 and 10 (DFT, B3LYP, 6–31+G*, nonpolar solvent) indicated that diastereomer 10 is 2.0 kcal/mol more stable than 21. Thus, we focused on isomerizing 21 into 10 under thermodynamic control.²⁶ Such isomerization could proceed under basic catalysis via retroaldolization of 21 to 22, with intramolecular aldol of 22 giving 10. Extensive optimization of the base, solvent and temperature showed that (Table 1-SI) Li⁺ and K⁺ counter-cations were not optimal (entries 1–3). However, Na⁺-containing bases offered more promising results, with NaHMDS in toluene (entry 5, 6), giving 10 in 48% yield (68% brsm, 5.2 g scale) in >20:1 dr.

Scheme 3.

Epimerization and reductive transposition studies

We next proposed that the global reduction of 10 would lead to pentaol 23, which upon exposure to aqueous acid, would ionize to provide a delocalized carbocation. This carbocation would be trapped by water at C3, and the subsequent collapse of the unsaturated C3-chlorohemiacetal would form 12.

The global reduction of 10 turned out to be challenging. The use of LiAlH₄ yielded multiple products including those arising from dechlorination. LiBH₄ and DIBALH were slow to effect C19 ester reduction. However, DIBALH at 60 ºC in THF provided polyol 23, which upon heating with aqueous formic acid in THF at 85 ºC provided 12 (2.0 g scale). The absolute and relative configuration of 12 was confirmed by X-ray crystallographic analysis.

Subsequent studies focused on converting 12 into 19-hydroxysarmentogenin (4) and related isomeric steroids [trewianin aglycone (6), panogenin (7), and 5-epi panogenin (29)] (Scheme 4). Approaches to 4, 6, 7 and 29 share the following features: (1) diastereoselective reduction of diene within 12 to form either β−C5 configuration present in 4 and 29 or α−C5 configuration present in 6 and 7; (2) functionalization of the C11 position to invert its configuration (4 and 6) or to retain it (7 and 29); (3) installation of the C17 butenolide moiety based on the Stille coupling/hydrogenation sequence established by the Inoue and Baran groups.^25,26

Scheme 4.

Synthesis of the isomeric cardenolides 19-hydroxysarmentogenin (4), trewianin aglycone (6), panogenin (7) and 5-epi panogenin (29)

Reagents and conditions. Scheme 4A (5-epi panogenin): (a) H₂, 10% Pd/C (25% w/w loading), KOH (1% w/v), quinoline (1% v/v), MeOH, 83% yield, >20:1 dr; (b) 2,2-dimethoxypropane, (+)-CSA, DMF, 91% yield; (c) DMP, CH₂Cl₂, 2 h, 87% yield; (d) K-Selectride^®, THF, −78 to −30 °C, 45 min; >20:1 dr; (e) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 79% yield; (f) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (27), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C; (g) TMSCl, Imidazole, DMF, rt, 24 h, 66% yield, 2 steps; (h) H₂, Pd/C, EtOAc, 1 h; (i) 0.3 M HCl in MeOH, 1 h 68% yield, 2 steps; Scheme 4B (19-hydroxysarmentogenin): (j) TIPSCl, ImH, DMF, rt, 6 h, 82% yield; (k) DMP, Py, CH₂Cl₂, 2 h, 89% yield; (l) K-Selectride^®, THF, –78 to –30 °C, 1.5 h, 77% yield, >20:1 dr; (m) TBSOTf, Et₃N, CH₂Cl₂, –78 to –30 °C, 1.5 h, 95% yield; (n) Li, NH₃, THF, –78 °C, 30 min; (o) TBAF, THF, –78 °C, 5 min, 73% yield (2 steps); (p) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 90% yield; (q) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (27), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C, 1 h, 92% yield; (r) TMSOTf, 2,6-lutidine, CH₂Cl₂, –78 °C to rt, 2 h, then SiO₂ (dry), 10 h, 64% yield; (s) H₂, Pd/C, EtOAc, 30 min, 2.7:1 dr (β-C17: α C17); (t) HF in CH₃CN/H₂O/CH₂Cl₂, rt, 3 days, 42% yield, 2 steps; Scheme 4C (panogenin): (u) 2,2-dimethoxypropane, (+)-CSA, DMF, 69% yield; (v) DMP, Py, CH₂Cl₂, 1.5 h, 90% yield; (w) H₂, Pd/C (25% w/w), EtOAc/MeOH/CH₂Cl₂ (3:1:1), 20 min, 2.5:1 dr; (x) LiAl(OtBu)₃H, THF, –78 to –40 ºC; 2.5 h, 59% yield, >20:1 dr (2 steps); (y) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 74% yield; (z) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (27), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C, 1 h, 66% yield; (aa) TMSOTf, 2,6-lutidine, CH₂Cl₂, –78 °C to rt, 2 h, then SiO₂ (dry), 10 h, 49% yield; (bb) H₂, Pd/C, EtOAc, 30 min; (cc) 0.3 M HCl in MeOH, 1 h 55% yield, 2 steps; Scheme 4D (trewianin aglycone): (dd) TIPSCl, ImH, DMF, rt, 6 h, 81% yield; (ee) H₂, Pd/C, MeOH, Py, 4 h; 83% yield, >20:1 dr; (ff) DMP, Py, CH₂Cl₂, 2 h, 91% yield; (gg) LiAlH(OtBu)₃, THF, –78 to –40 °C, 4 h, 96% yield, >20:1 dr; (hh) TBSOTf, Et₃N, CH₂Cl₂, –78 to –30 °C, 1.5 h, 87% yield; (ii) Li, NH₃, THF, –78 °C, 30 min; (jj) TBAF, THF, –78 °C, 5 min, 91% yield (2 steps); (kk) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 83% yield; (ll) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (27), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C, 1 h, 93% yield; (mm) TMSOTf, 2,6-lutidine, CH₂Cl₂, –78 °C to rt, 2 h, then SiO₂ (dry), 10 h, 79% yield; (nn) H₂, Pd/C, EtOAc, 30 min; 54% yield of β-C17 and 22% of α-C17; (oo) HF in CH₃CN/H₂O/CH₂Cl₂, 3 days, 91% yield.

The direct hydrogenation of 12 with Pd/C as the catalyst could be directed by the unprotected C19 hydroxyl group to provide β−C5 stereocenter under basic conditions (Scheme 4A/4B). This event was followed by the selective C11/C19 acetonide protection leading to 24 (Scheme 4A) or by the C19 TIPS protection to provide intermediate 30 (Scheme 4B). The acetonide moiety in 24 was used to protect both the C19 as well as the β-C11 hydroxyls (Scheme 4A). As a result, the subsequent steps that included the oxidation of 24 with DMP (87% yield), and the elaboration of the oxidation product 25 to vinyl iodide 26 (79% yield, 2 steps) did not affect the stereochemistry of the C11 stereocenter. The subsequent Stille reaction with commercially available stannane 27 was carried out under the previously published conditions and provided precursor 28. Similar to the observations of Inoue and Baran, we could not hydrogenate Δ^16–17−olefin in compound 28 directly to get a β−configured C17-butenolide. To circumvent this problem, 28 was reacted with TMSCl and imidazole in DMF forming the C14-protected product in 78% yield. The subsequent reduction proceeded with significantly higher selectivity than in the subsequent synthesis of 4 or 6, providing a 25: 1.0: 2.8 mixture of α : β : Δ^17–20−olefin-containing product of deconjugation. This mixture was treated with HCl in methanol to provide 5-epi-panogenin (29) in 68% yield over 2 steps.

The elaboration of the intermediate 30 into 19-hydroxysarmentogenin (4) followed similar logic, but required additional steps required for the installation of the α-C11 stereocenter (Scheme 4B). This was carried out by subjecting 30 to oxidation with DMP and then performing subsequent site-selective reduction of the C3-ketone moiety with K-Selectride^® (69% yield, >20:1 dr, 2 steps). The C17 ketone moiety of the resultant product 31 was selectively masked as a TBS-silyl enol ether (95% yield), and this product was subjected to the β-selective C11 ketone reduction with Li/NH₃ followed by the cleavage of the silyl enol ether moiety with TBAF (73% yield, 2 steps). The resultant product was then converted to vinyl iodide 32 with N₂H₄/ I₂ (90% yield), which was elaborated to 19-hydroxysarmentogenin (4) using the previously established strategy (4 steps, 25% yield).

The synthesis of the trans-A/B ring junction containing steroids panogenin (7) and trewianin aglycone (6) from 12 involved similar steps; however, required installation of the α-C5 stereocenter (Schemes 4C and 4D). The trans-A/B ring configuration of panogenin (7) was established by protecting the C11/C19 diol moiety of 12 as an acetonide, and oxidizing the C17 position with DMP to obtain intermediate 34 (62% yield, 2 steps) (Scheme 4C). This intermediate was hydrogenated over Pd/C to provide the product containing the trans-A/B ring junction (2.5:1 dr at C5), which was then reduced with LiAlH(Ot-Bu)₃ providing 35 (59% yield, 2 steps). This product was converted to vinyl iodide 36 (74% yield), which was subjected to 4 step butenolide installation sequence based on previously established for 5-epi-panogenin (29) route to provide 7 in 18% yield.

Finally, the synthesis of trewianin aglycone (6) commenced with the protection of the C19 group of 12 with as a TIPS ether (Scheme 4D). This bulky protecting group was essential for the highly selective hydrogenation with Pd/C to provide intermediate 38 with trans-A/B ring junction (>20:1 dr). The subsequent elaboration of 38 into 6 resembled the sequence developed for 19-hydroxysarmentogenin (4) and proceeded in 10 steps and 21% overall yield.

The synthesis of sarementologenin (3) and ouabagenin (1) required developing an “oxidase phase” strategy for the selective reduction of the Δ^6–7−olefin of 12 and strategies for the installation of the oxygenation at the C1 and C5 positions.³² The reduction of the Δ⁶−olefin of 12 required extensive optimization (Scheme 5 and Table 4-SI) that identified LiDBB reduction followed by NaHCO₃ work up as the only viable option leading to 42. The enone within 42 was epoxidized stereoselectively under basic conditions (68% yield, 4:1 dr) with β−epoxide 43 as the major product. Reduction of 43 presented another challenge because the reaction product, 44, readily eliminated water, re-forming enone 42. Numerous reported conditions for the selective reduction of 43 led to decomposition. The reduction promoted by PhSeSePh (10 mol%) and N-Ac-L-Cys-OH under basic conditions³³ was uniquely effective at accessing 44.

Scheme 5.

Application to the synthesis of sarmentologenin (3)

Reagents and conditions: (a) LiDBB, THF, −78 ºC, 15 min, then NaHCO₃, rt, 30 min, 91%; (b) H₂O₂, NaOH, MeOH, 0 ºC to rt, 2 h, 68%; (c) TIPSOTf, 2,6-lutidine, THF:CH₂Cl₂ = 1:1, 5 min, 88%; (d) PhSeSePh (20 mol%), N-Ac-L-Cysteine, NaOH, DMF, rt, 15 min, 91%.

Isolation of 44 was challenging due to its hydrophilicity. This issue was circumvented by selective TIPS protection of the C19-OH of 43 to provide 45 (88% yield). Epoxide 45 underwent facile ring opening under the optimized conditions for the epoxide opening [PhSeSePh (20 mol%), N-Ac-L-Cys, NaOH] to provide sarmentologenin core 46 in 91% yield.

The sarmentologenin core 46 was then elaborated to sarementologenin (3) in 10 steps (Scheme 6). As before, the reduction of the C3 ketone moiety of 46 was accomplished with K-Selectride^®. A stable boronic acid ester 47a (36% yield) was obtained along with 47b (62% yield); 47a was hydrolyzed to 47b with hydrogen peroxide/sodium acetate (92% yield). The key precursor 47b was then subjected to the standard sequence (Scheme 6) that involved selective reduction from the β−face of the C11 ketone to provide 48 (77% yield, 3 steps), formation of the vinyl iodide 49 from 48 (89% yield) and stereoselective installation of the β−C17 butenolide moiety followed by an acid-mediated global deprotection to provide sarementologenin 3 (21% yield, 4 steps). The absolute and relative configuration of 3 was confirmed by X-ray crystallographic analysis.

Scheme 6.

Completion of the synthesis of sarementologenin (3)

Reagents and conditions: (a) DMP, Py, CH₂Cl₂, rt, 2 h, 91% yield; (b) K-Selectride^®, THF, −78 °C, 45 min, 62% yield for 47b, 36% yield for 47a; (c) H₂O₂, NaOAc, THF : H₂O = 3:1, rt, 16 h, 92% yield; (d) TBSOTf, Et₃N, CH₂Cl₂, −78 to −20 °C, 1 h, 88% yield; (e) Li, NH₃, THF, −78 °C, 15 min; (f) TBAF, THF, −78 °C, 5 min, 87% yield (2 steps); (g) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 96% yield; (h) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (27), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C, 1 h; (i) TMSOTf, 2,6-lutidine, CH₂Cl₂, −78 °C to rt, 2 h, then SiO₂ (dry), 10 h, 52% yield (2 steps); (j) H₂, Pd/C, EtOAc, 30 min; 44% yield of β-C17 and 24% of α-C17; (k) HCl in MeOH, MeOH, rt, 30 min, 83% yield.

Our studies next focused on adapting the route to sarmentologenin (3) to the synthesis of ouabagenin (1). This required installation of an additional hydroxyl group at C1. We anticipated simultaneous installation of both hydroxyl groups from dienone 51 (Scheme 7) via epoxidation of both alkene positions and subsequent opening the resultant oxirane moieties using conditions developed in the synthesis of sarmentologenin (3). However, our attempts to obtain 51 or its protected variants, compounds 54 and 55, were not successful due to facile aromatization of the steroidal A ring. The stability and reactivity of 54 and 55 could be improved by acetylation of the C19 hydroxyl group,³⁴ and the oxidation of 56 and 57 with DDQ provided dienones 58 and 59, respectively. However, all attempts to accomplish selective epoxidation of 58 or 59 failed, so a stepwise installation of the C5 and C1 oxygenation through monoepoxide intermediate 43 was targeted next (Scheme 8).

Scheme 7.

Initial studies towards the synthesis of ouabagenin (1)

Scheme 8.

Installation of oxygenation in ouabagenin (1)

Reagents and conditions: (a) H₂O₂, NaOH, 0ºC to rt, 2 h, 68% yield; (b) Ac₂O, Py,rt, 2 h, 88% yield; (c) SeO₂, chlorobenzene, 90 ºC, 12 h, 36% yield (variable); (d) DDQ, dioxane, 90 ºC, 12 h, 10–50% yield(variable); (e) NaOH,MeOH, rt, 3 h, then H₂O₂, 0 ºC, 30 min, 35% yield; (f) SeO₂, chlorobenzene, 90 ºC, 12h, <30% yield (variable); (g) DDQ, dioxane, 90 ºC, 12 h, <50% yield (variable); (h) Ac₂O, Py, DMAP, rt, 16 h, 89% yield; (i) K-Selectride^®, THF, −30 °C, 30 min, dr = 6:1; (j) m-CPBA, CH₂Cl₂, rt, 2 h, 70 % yield (2 steps); (k) Pd(TFA)₂, NaOAc, O₂, DMSO, 50 ºC, 36 h, 90% yield.

Our studies commenced with the direct installation of the Δ¹−olefin in epoxide 43 using DDQ or selenium(IV) oxide, which led to enone 60 in low yields (30–50% yield) accompanied with unidentified and inseparable impurities (Scheme 8). An alternative route based on the dehydrogenation of bis-acetate 62 leading to 63 was investigated. The presence of the C19 acetate accelerated this oxidation relative to 43, and the transformation of 62 to 63 was accomplished with selenium(IV) oxide or palladium(II) trifluoroacetate;³⁵ however, the yields for this transformation were low (~10–50%) and irreproducible due to concomitant cleavage and migration of acetates. The subsequent epoxidation of 63 to 61 was accomplished in 35% yield. Owing to the low efficiency and irreproducibility of several steps, this route was discarded. In search of a more efficient route, substrate 42 was acetylated (Ac₂O, DMAP) to provide triacetate 64 (89% yield), which was then selectively reduced from the α-face with K-Selectride^® and subjected to C3-directed epoxidation with m-CPBA to provide epoxide 65 in 70% yield (6:1 dr for the epoxidation step) over 2 steps. The C3 alcohol moiety was oxidized with DMP to provide the corresponding C3 ketone, which was then subjected to various dehydrogenation conditions to establish the Δ¹−unsaturation. Unfortunately, oxidation with DDQ or SeO₂ did not efficiently form product 62 due to either epoxide opening or overoxidation at the C2 position leading to side-products (vide supra). However, directly oxidizing 65 with Pd(TFA)₂ under an atmosphere of oxygen furnished enone 66 in 89% yield.

The installation of C1/C5 oxygenation en route to ouabagenin (1) was pursued next (Scheme 9). Subjecting 66 to LiOH then hydrogen peroxide resulted in the removal of the acetate protecting groups and installation of the C1/C2 epoxide (74% yield, single diastereomer). The C19 hydroxyl of the resultant bis-epoxide 61 was selectively protected as a TIPS ether giving compound 67 in 84% yield. This protection was important as it increased the hydrophobicity of 67, a feature that was essential for the subsequent steps. The bis-epoxide moiety of 67 was oxidized with Dess-Martin reagent to provide triketone 68 in 98% yield, which was then subjected to the optimized conditions for the epoxide opening (PhSeSePh (20 mol%.), N-Ac-L-Cys, NaOH) that were developed during the studies on sarmentologenin (Scheme 6). Similar to previous reports from the Baran group,²¹ we observed instability of 69 to water elimination and A-ring aromatization under basic or acidic conditions. Careful control of the N-Ac-L-Cys to NaOH stoichiometry was required to avoid these side reactions and obtain the desired product 69 in 87% yield. The C3 ketone moiety of 69 was then transformed to the C3- and C17-silylated enol ether 70 in 86% yield. Further elaboration of 70 proceeded through the standard sequence established for the other steroids and ultimately provided 75 (60% yield), which upon deprotection with HCl/MeOH yielded ouabagenin (1) in 92% yield. Our synthetic ouabagenin (1) was in all respects identical to the sample obtained by deglycosylation of ouabain (2), and its NMR spectra match previously published spectroscopic data.^24,25,26b,36 It should be noted that our total synthesis features only 22 linear steps and 1.0% overall yield.

Scheme 9.

Completion of the synthesis of ouabagenin (1)

Reagents and conditions: (a) LiOH, MeOH, rt, 2 h, then H₂O₂, 0 ºC, 15 min, 74% yield; (b) TIPSOTf, 2,6-lutidine, THF, 5 min, 84% yield; (c) DMP, CH₂Cl₂, rt, 2 h, 98%; (d) PhSeSePh (20 mol%), N-Ac-L-Cysteine, NaOH, DMF, rt, 15 min, 87%; (e) K-Selectride^®, THF, −78 °C, 30 min then TBSOTf, Et₃N, CH₂Cl₂, −20 °C, 30 min, 86% yield; (f) Li, NH₃, t-BuOH (10 equiv), THF, −78 °C, 15 min; then TBAF, THF, −20 °C, 8 h, 86% yield; (g) N₂H₄·H₂O, Et₃N, EtOH, 50 °C, 6 h then I₂, Et₃N, THF, rt, 1 h, 96% yield; (h) 4-(tributylstannyl)-2,5-dihydrofuran-2-one (25), Pd(PPh₃)₄, CuCl, LiCl, DMSO, 50 °C, 1 h, 84% yield; (i) TMSCl, Imidazole, DMF, rt, 72 h, 90% yield; (j) H₂, Pd/C, EtOAc, 30 min; 60%; (k) HCl in MeOH, MeOH, rt, 30 min, 92% yield.

We also targeted steroids lacking C11 oxygenation³⁷ (Scheme 10). This strategy was based on enantioselective Michael reaction of enone 8a and β−ketoester 19a, and it routinely led to the formation of the Michael adduct 76 in good yield, diastereoselectivity and enantioselectivity (92%, >20:1 dr, 92% ee) on 7.2 g scale. 76 was then elaborated to enone 11 via the sequence previously developed for its C11-hydroxy congener 12. Substantial quantities of 11 were obtained and converted to cannogenol (5) in 9 steps and 20% overall yield. Since cannogenol 3-O-α−L-rhamnoside (79a), is a natural derivative of cannogenol (5) with anticancer activity in the nM range,³⁷ the synthesis of 79a as well as other cannogenol glycosides 79b and 79c was pursued.^28b To mask the more reactive C19 hydroxyl group, 5 was subjected to selective protection as a methoxyacetate to provide 77 (99% yield). The C3 alcohol of 77 was then glycosylated with trichloroacetimidate 78a to provide the corresponding α-L-rhamnoside as the only diastereomer. The following deprotection of the sugar benzoates and C19 methoxyacetate leading to cannogenol α−L-rhamnoside (79a) required careful optimization due to the lability of the C17 butenolide under basic conditions. 50% ammonia in methanol achieved efficient cleavage of the ester moieties without destroying the butenolide (56% yield). This protocol was also employed to generate glycosides 79b and 79c in 22% and 65% yield from protected cannogenol derivative 77. In addition to cannogenol (5) glycosides, the α−L-rhamnosides of steroids digitoxigenin,²⁰ bufalin³⁸ and strophanthidol³⁷ were synthesized using similar protocols (SI).

Scheme 10.

Synthesis of cannogenol glycosides 79a-c

Reagents and conditions: (a) (MeOCH₂CO)₂O, DIPEA, CH₂Cl₂, 0 ºC, 12 h; (b) 78a-c, TfOH (30 mol%), CH₂Cl₂, 0 ºC, 5 h; (c) NH₃ (50% in MeOH), rt, 24 h, 56% from 77 for 79a; 22% from 77 for 79b; 65% from 77 for 79c.

Poised to address the biological questions, we assayed a collection of 37 compounds (Table 11-SI) comprising those generated above and several commercially-available cardenolides (ouabain (1), anhydro-ouabagenin, digoxin, digitoxin, digitoxigenin, and strophanthidol). Our studies commenced with a smaller subgroup of synthetic aglycones 5, 6, 12, 100–105 (Table 11-SI) using different cell lines: HEK293T (an adenovirus-transformed cell line, containing the SV40 T-antigen); HeLa (a human cervical cancer line containing human papilloma virus 18); Cos7 (an African green monkey line expressing the SV40 T-antigen); Raji (a human Burkitt’s lymphoma line); and NIH-3T3 (a mouse fibroblast line)]. This panel represents both adherent lines and a non-adherent line (Raji), and transformed lines, as well as a non-transformed line (NIH-3T3). Proliferation was measured over 2 days via AlamarBlue®. The assay was validated by comparing the normalized AlamarBlue signal averages of the DMSO-treated samples to those for cells treated with 2 µg/ml puromycin. A global ź factor of 0.66 was observed between these two conditions. Compounds 5, 6, 12, 100–105 were then screened at 2 µM and 0.6 µM. Only cannogenol (5) was active, showing activity against every line except NIH-3T3 at 2 µM, and affecting most lines at 0.6 µM. These data recapitulate previous findings that primate cell lines are sensitive to cardiotonic steroids, but rodent lines are unusually robust.³⁹ Furthermore, non-adherent lines were as sensitive as adherent lines, meaning anoikis is likely not an on-target death mechanism.

The EC₅₀ of 5 in HEK293T was ~1 µM, consistent with the screening data.

We next took HEK293T (Figure 3A) and HeLa (Figure 3B) cells as representative sensitive lines and NIH-3T3 (Figure 2c-SI) as an insensitive line and assayed the 37 compounds for growth inhibition at 1 and 3 µM concentrations.

Figure 3.

Evaluation of anticancer activity of 37 synthetic and natural cardenolides against (A) HEK293T and (B) HeLa (1 µM, 48 hr) (C) EC₅₀ values for the 5 most potent glycosylated steroids against HEK293T, HeLa, U87, HEPG2, MDA-MB-231 and NIH-3T3. Box shows median with standard deviation; whiskers show 95% confidence intervals. (D) Structure of compounds 79a, 92, and 97. (E) Toxicity studies in developing larval zebrafish at 3 µM of indicated compounds (DMSO N=38; 79a N=35; mitomycin C N=36; brefeldin A N =38; actinomycin D N=33). (F) Representative data evaluating teratogenicity. Fish A: coronary edema (a bulge in cardiac tissue); B and C: normal. Fishes D and E: spinal deformities (and coronary edema).

These studies enabled us to extract some important structure activity relationships (SAR). The significant SAR information is depicted in Figure 2 and are discussed in detail in supporting information (SI-2k). From these data, we also hoped to identify compounds highly toxic to HeLa and HEK293T but that did not affect NIH-3T3. If these characteristics were observed, we would preliminarily conclude that the compounds are on target. We evaluated EC₅₀ values for the most potent compounds (none of which were toxic to NIH-3T3), most of them emerged to be glycosylated steroids (ouabagenin (1), ouabain (2), cannogenol α−L-rhamnoside (79a), 79c, digitoxigenin (80), digitoxin (81), digoxin (82), and 92–98), in HEK293T cells. These compounds showed EC₅₀s below 1 µM, consistent with the screens in Figures 3A and 3B.

Figure 2.

Summary of the key features from the SAR studies

The 5 most potent compounds (EC₅₀s below 100 nM in HEK293T cells) 81, 82, 79a, 97 and 92 (Figure S3) were assayed against HeLa and several other cell lines (HepG2, a human hepatocellular carcinoma line; MDA-MB-231, a human invasive triple negative breast adenocarcinoma line; and U87, a human primary glioblastoma cell line), the results of which are depicted in Figure 3C. In HEK293T, 79a was marginally more potent than 97 and 92, and those 3 compounds were all approximately 2–3 fold more potent than 81 and digoxin 82. This trend was retained in HeLa cells, but 79a was marginally less toxic than 92 and 97. In U87, HEPG2 and MDA-MB-231 92 and 97 were significantly (up to 3–4 fold) more potent than 79a. Furthermore, in HEPG2 the difference between 81 and 79a was marginal although in the other lines 79a was around 2-fold more potent than 81. Considering that all these compounds were non-toxic to NIH-3T3 cells at 3 µM, these data indicate that the toxicities were occurring via a conserved mechanism.

Compound 79a was next screened against several other lines [Cos7, Raji, HFF1 (a human fibroblast line), RAW (a virally-transformed mouse macrophage line), MEFs (a mouse embryonic fibroblast line) and HF-II-E (a hepatoma line from rat)] (Figure S4). The data observed were bimodal: either cells were sensitive to the compound at low-to-mid nanomolar concentrations or they were resistant to treatment (EC₅₀>1 µM). Similar outcomes were observed for other compounds (e.g. 98, 97, 79b, and 99 were ineffective against RAW; 92, and 97 were ineffective against HF-II-E; 98, 81, 2 and 1 were effective against HFF1 (Figure S4)) against the same lines. Thus, the toxicity was similar for numerous adherent cells versus the single non-adherent line, further arguing against anoikis as a prevalent mode of toxicity. Furthermore, the most active compounds were essentially non-toxic to non-cancerous lines, (NIH-3T3 and MEFS), but killed a wide range of different cancer cells, from humans and other species. This observation could be considered as evidence that these compounds are selectively toxic to transformed cells. However, these compounds were also non-toxic to transformed rodent lines and they were further unable to prevent human-RAS^G12V transformation of NIH-3T3 cells (Figure S5). Thus, although NIH-3T3 and other rodent cells are good model resistant lines, we provide very clear evidence— which is also consistent with previous reports^38,40—that rodent lines are generically resistant to cardiac steroids. We assayed the effects of cardiac steroids on developing zebrafish relative to other cytotoxic compounds (Figure 2D). Zebrafish were chosen because numerous fish, including zebrafish, have been shown to be responsive to the cardiac steroid ouabain, and based on previous data and our own phylogenetic analysis, we expected Na,K-ATPase from zebrafish to be sensitive to cardiotonic steroids (Figure S6). Furthermore, hits from small-molecule screening in zebrafish have begun to enter clinical trials.⁴¹

We performed two assays, a teratogenicity assay, and assessment of the cardiac rate. Both assays involved dosing embryos at the 2–8 cell stage. 3 µM 79a [EC₅₀(HEK293T) = 0.015 µM] exhibited little toxicity or teratogenicity, although there was a slight increase in spinal curvature (Figure S7A). Mitomycin C [EC₅₀(HEK293T)=4 µM], brefeldin A [EC₅₀(HEK293T)=0.2 µM] and actinomycin D [EC₅₀(HEK293T)=0.002 µM] administered at 3 µM, all elicited significant toxicity and teratogenicity, although most of these compounds were less toxic to cancer cells than 79a (Figure S7B). Heart rate was also unaffected by 79a (Figure S7C). These results indicate that non-transformed cells tolerate 79a much better than generically toxic compounds. This result underscores previous hypotheses concerning the suitability of these compounds as selective anticancer agents^13a and is strongly consistent with the common use of these compounds in humans. Although more complex assays are required to define utility for anticancer treatment, these data are encouraging.

Interestingly, we found that cultured MDA-MB-231 cells treated with alkyne-functionalized 79b (500 nM) followed by permeabilization, fixation, Click coupling to Cy5-azide, and extensive washing, exhibited intracellular fluorescence. DMSO-treated cells were less fluorescent. Since these assay conditions should only preserve irreversible adducts,⁴² these data imply the presence of (an) intracellular protein targets to which these compounds covalently bind (Figure S8). We are mindful that the principal biological targets of cardiotonic steroids may not arise from covalent adduction. However, since drug outputs are a combination of both on- and off-target binding events, it is important to identify as many targets as possible to optimize beneficial and deleterious binding interactions, especially for pleiotropic molecules like cardenolides. Probes such as 79b that display a toxicity profile similar to the canonical cardiotonic steroids, are thus useful for further mechanistic studies.

We finally compared effects of compounds 79a, 92 and 97—a set of compounds with differential oxidation of the steroid ring—against (MDA-MD-231; EC₅₀ 0.111, 0.038, and 0.042 µM respectively) and an inherently-insensitive line (NIH-3T3, EC₅₀s>3 µM). We investigated how these compounds affect DNA-damage response (DDR)³⁶ (γ-H2AX upregulation) and apoptosis (caspase-3 activation, an early step in apoptosis) (Figure S9). We first confirmed the validity of our immunofluorescence assays using mitomycin C, which showed DDR-upregulation in MDA-MB-231 and NIH-3T3, but Caspase-3-uregulation only in NIH-3T3, Using identical assay conditions, 24 h treatment with 79a (0.111 µM), 92 (0.038 µM) or 97 (0.042 µM) did not upregulate γ-H2AX or caspase-3 activation in NIH-3T3, consistent with the resistance of this line to cardiotonic steroids. In MDA-MB-231 cells, 24-h treatment with the same dose of these compounds upregulated both γ-H2AX and caspase-3 activation, consistent with apoptosis, possibly occurring through DDR. For all 3 compounds in MDA-MB-231 cells, there was an increase in γ-H2AX puncta, further indicating DDR as a common effect. We next compared effects in HeLa cells treated with 79a (0.015 µM) and 92 (0.012 µM) for 24 h (Figure S10).

Little upregulation in caspase-3 was observed, whereas γ-H2AX was upregulated significantly with an increase in γ-H2AX foci. Staurosporine on the other hand upregulated both γ-H2AX and caspase-3. A caspase inhibitor did not affect γ-H2AX upregulation promoted by 79a and 92. Thus, we conclude that the mode of action for these two compounds and likely 97 follows the following order: DDR upregulation, which in turn upregulates apoptosis. Although more work needs to be done to validate the hypothesis, we propose that this mode of action (apoptosis that accompanies DDR) is common to numerous cardenolides.

3. CONCLUSION

In summary, we have developed a convergent and concise total synthesis of ouabagenin (1) and several other cardiotonic steroids with lower oxidation states including 11-deoxy steroids such as cannogenol (5) and its glycosylated analogs (79a-b). These syntheses enabled the subsequent SAR studies to elucidate the effect of oxygenation, stereochemistry, C3-glycosylation and C17-heterocyclic ring on the anticancer activity of the synthetic and natural cardiotonic steroids. Based on parallel evaluation of synthetic and natural steroids and their derivatives, natural products cannogenol-3-O-α−L-rhamnoside (79a) and strophanthidol-3-O-α−L-rhamnoside (92) were identified as two of the most potent analogs, demonstrating broad anticancer activity in 10–100 nM concentrations and selectivity (nontoxic at 3μM against NIH-3T3, MEF and developing fish embryos).