The origin of all tested cardenolides is indicated in Supplementary Table 1. Compounds were dissolved in dimethyl sulfoxide (DMSO) (Merck, Darmstadt, Germany). Paclitaxel was from Sigma-Aldrich (St. Louis, USA). Etoposide, 3-aminobenzamide (3-ABA), cathepsin L inhibitor, and bafilomycin A1 were from Sigma-Aldrich (Bornem, Belgium). z-VAD-FMK (z-VAD), necrostatin (Nec)-1, and calpain inhibitor PD150606 were from Calbiochem (Leuven, Belgium). Cathepsin B inhibitor was from Cell Signaling Technology (Bioke, Leiden, The Netherlands). Mammalian Target of Rapamycin (mTOR) inhibitor PP242 (Torkinib) was from Sigma-Aldrich.

Human non-small cell lung cancer (NSCLC) A549 cells (ATCC, Manassas, USA) and normal fetal lung fibroblast cells (MRC-5, ECACC, Salisbury, UK) were grown in Dulbecco's Modified Eagle's Medium (DMEM; Gibco® Carlsbad, CA, USA) supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco®). MRC-5 cells were complemented with 2 mM glutamine (Cultilab, Campinas, São Paulo, BR) and 1% non-essential amino acids (Gibco®). NSCLC cell lines H1573, H1975, H1437, and H1299 were from ATCC (LGC Standards, Molsheim, France). HT-29 (human colon adenocarcinoma), SK-N-AS and SH-SY5Y (human neuroblastoma), K562 (chronic myelogenous leukemia), U937 (acute myeloid leukemia), Jurkat (T-cell leukemia), and Raji (Burkitt's Lymphoma) cells were from DSMZ (Braunschweig, Germany); cells were cultured in RPMI medium (Lonza, Verviers, BE) supplemented with 10% (v/v) fetal calf serum (FCS) (Lonza) and 1% (v/v) antibiotic-antimycotic (penicillin, streptomycin, and amphotericin B) (BioWhittaker, Verviers, Belgium). Peripheral blood mononuclear cells (PBMCs) were purified using Ficoll-Hypaque (GE Healthcare, Roosendaal, The Netherlands). PBMCs were isolated by density gradient centrifugation from freshly collected buffy coats obtained from healthy adult human volunteers (Red Cross, Luxembourg, Luxembourg). All healthy volunteers gave informed written consent. After isolation, cells were washed twice in 1X PBS and adjusted to 2 × 10⁶ cells/mL in RPMI 1640 (supplemented with 1% antibiotic–antimycotic and 10% FCS (BioWhittaker, Verviers, Belgium). All cells lines were maintained at 37°C and 5% CO₂ in a humidified atmosphere.

Initial structures of the Na⁺/K⁺-ATPase complexed with CGs were obtained from the Protein Data Bank (PDB; PDB ID: 4HYT, 4RES, 4RET) (Laursen et al., 2013, 2015) and coordinates for GEV were generated using ChemBioDraw Ultra 14.0 (PerkinElmer, Waltham, MA, USA). After removing ligands in the original data, we performed computational docking using the Autodock Vina program (Trott and Olson, 2010) with the protein and the compound as receptor and ligand, respectively. Structural superposition of 4HYT, 4RES, and 4RET coordinates was performed using WinCoot (Emsley et al., 2010). Structural representation of docking results was done with PyMOL (The PyMOL Molecular Graphics System, Version 1.8 Schrödinger, LLC., New York, NY, USA).

Spheroids were generated by the hanging drop method. 3 × 10³ A549 cells were seeded in 25 μL drops of culture media on covers of Petri dishes filled with 15 mL PBS. After incubation for 96 h with GEV (0, 1, 10, 50, 100 nM), spheroids were recovered in a 6-well plate in 4 mL culture media. Images were taken (Nikon) and further analyzed by ImageJ software (http://rsb.info.nih.gov/ij/docs/index.html). 3D picture analysis and quantification were carried out using ReViSP software (https://sourceforge.net/projects/revisp/).

A549 cells (10³ cells/ml) were seeded with 0, 10, 50, or 100 nM GEV and then grown in semi-solid methylcellulose medium (Methocult H4230, StemCell Technologies Inc., Vancouver, Canada) supplemented with 10% FBS. Colonies were detected after 10 days of culture by adding 1 mg/ml of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT) reagent (Sigma) and were scored by ImageJ software (U.S. National Institute of Health, Bethesda, MD, USA).

Wild-type zebrafish (Danio rerio) were obtained from the Zebrafish International Resource Center (ZIRC, University of Oregon, OR), maintained per SNU guidelines at 28.5°C with 10 h dark/14 h light cycles. Viability and abnormal development were assessed under light microscopy (Carl Zeiss Stereo Microscope DV4, Seoul, Korea). Pictures were taken by fixing zebrafish embryos onto a glass slide with 3% methyl-cellulose (Sigma-Aldrich). For cancer xenograft assays, after mating, fertilized eggs were incubated in Danieau's solution with 0.003% of phenylthiourea (PTU) at 28.5°C for 48 h. Micropipettes for injection and anesthesia were generated from a 1.0 mm glass capillary (World Precision Instruments, FL, USA) by using a micropipette puller (Shutter Instrument, USA). 48 h post fertilization (HPF), zebrafish were anesthetized with 0.02% tricaine (Sigma, MO) and immobilized on an agar plate. A549 cells were treated for 30 h with GEV (0, 10, 50, 100 nM) and stained last 2 h of incubation with 4 μM Cell tracker CM-Dil dye (Invitrogen). 200 A549 cells were injected into the yolk sac (PV820 microinjector, World Precision Instruments, FL, USA). Subsequently, zebrafish were incubated in 96-well plates containing Danieau's solution with 0.003% phenylthiourea (PTU) at 28.5°C for 72 h. Fish were then immobilized in a drop of 3% methylcellulose in Danieau's solution on a glass slide. Pictures were taken by fluorescence microscopy (Leica DE/DM 5000B). Area of fluorescent tumors was quantified by ImageJ software.

A549 and U937 cells (2 × 10⁵ in six-well plates) were treated with GEV and lysed with M-PER® (Mammalian Protein Extraction Reagent; Pierce, Erembodegem, Belgium) lysis buffer. After centrifugation at 10,000 × g for 15 min, supernatants were collected and total proteins (20–40 μg) were separated by size using sodium dodecyl sulfate—polyacrylamide gel electrophoresis (SDS-PAGE), transferred to polyvinylidene difluoride membranes (PVDF) (Amersham Hybond P, GE Healthcare, Buckinghamshire, UK) and blocked with 5% non-fat milk in PBS/Tween 20 for 1 h. Equal loading of samples was controlled using β-actin. Blots were incubated with primary antibodies: cyclin B1 (1:1000, Merck Millipore, Overijs, Belgium), p53 (1:1000, Santa Cruz Biotechnology, Boechout, Belgium), caspase-3 (1/1000, Santa Cruz Biotechnology 31A1067), p62 (1:1000, Cell Signaling Technology), beclin-1 (1:1000, Sigma-Aldrich), LC3II (1:1000, Sigma-Aldrich), and β-actin (1:10000, Sigma-Aldrich). All antibodies were diluted in a PBS-Tween 20 solution containing 5% of bovine serum albumin (BSA) or 5% of non-fat milk. After overnight incubation with primary antibodies, membranes were washed with PBS-Tween, followed by incubation for 1 h at room temperature with the corresponding secondary horseradish peroxidase-conjugated antibodies from Santa Cruz Biotechnology. After washing three times with PBS/Tween 20, specific immunoreactive proteins were visualized by ImageQuant LAS 4000® mini system (GE Healthcare) using a chemiluminescent detection reagent kit (Amersham ECL Prime, GE Healthcare).

Results were expressed as mean ± SD of three independent experiments. Statistical analyses were performed by one-way ANOVA followed by appropriate post-hoc tests. A value of p < 0.05 was considered statistically significant. Graphs were performed with GraphPad Prism software 7 version (GraphPad, La Jolla, CA, USA).

We evaluated the anticancer potential of a panel of 46 natural or hemisynthetic cardenolides on A549 cells (Supplementary Table 1). The most cytotoxic cardenolide was glucoevatromonoside (GEV, Figure 1A) which showed 14-fold greater potency compared to paclitaxel used as a positive control (IC₅₀ = 19.3±4.5 nM vs. 260.5±70.8 nM).

We then further evaluated the anti-cancer potential and selectivity of GEV. We generalized our findings with additional lung cancer cell lines and other types of solid tumor compared to non-adherent cancer models (Table 1). Our results show that GEV was cytotoxic at nanomolar concentrations and presented selectivity for lung cancer cell models (H1299 = H1573 = H1975 = A549 = U937 > SH-SY5Y > JURKAT > K562 > RAJI > SK-N-AS > H1437 >HT-29 cells).

To assess for differential toxicity, we used two non-cancer cell models, lung MRC-5 cells and PBMCs from healthy donors. After 48 h of treatment, GEV was 45-fold and 8.5-fold more selective to A549 compared to PBMCs and MRC-5 cells, respectively (Table 1). The effect of GEV on sub-populations of PBMCs was also investigated (Supplementary Figures 1A,B). Results showed that GEV did not affect the lymphocytic population at any concentration (Supplementary Figure 1B). The population of monocytes was affected by GEV after 48h of treatment only at concentrations beyond the range of this investigation (1000 nM) (Supplementary Figure 1A). Considering the differential toxicity and selectivity of GEV toward lung cancer cell models, we selected the lung A549 cells for in vitro and in vivo investigations. Furthermore, we compared results to the acute myeloid leukemia cell line U937, which exhibits a similar susceptibility to GEV as A549 cells.

We previously documented the inhibitory potential of GEV on Na⁺/K⁺-ATPase activity (Bertol et al., 2011) by an in vitro assay (Hu et al., 2009). To validate the interaction mode of GEV and its putative pharmacological target, the Na+/K+ ATPase, we implemented docking simulations using Autodock Vina software (Trott and Olson, 2010) (Figure 1B). In the absence of structural data for the human Na⁺/K⁺-ATPase protein, we instead used the structure of the Na⁺/K⁺-ATPase of Sus scrofa complexed with three different CGs, ouabain, bufalin, and digoxin (Laursen et al., 2013, 2015) as previously published (Zeino et al., 2015). Sequence identities of α1, β1, and γ-subunits of the Na⁺/K⁺-ATPase between human and pig are 98.1, 92.4, and 80.3%, respectively (Sievers et al., 2011). Before predicting the affinity between GEV and Na⁺/K⁺-ATPase, we performed a control docking experiment with the complexed CGs as ligands, after the CGs were removed from the respective complex structures. The predicted affinity energies of GEV in three different complex structures of ouabain, bufalin, and digoxin were comparable to those calculated from the original complex structures; −11.3 kcal/mol, −11.0 kcal/mol, and −11.4 kcal/mol, against −11.1 kcal/mol, −11.1 kcal/mol, and −16.3 kcal/mol, respectively (RMSD distances between docking results and crystal structures of ouabain, bufalin, and digoxin were 0.046 Å, 0.042 Å, and 0.001 Å, respectively) (Figure 1B).

In the docking analyses, GEV was predicted to be located in the cavity of the extracellular ion pathway in the α subunit of the Na⁺/K⁺-ATPase where CGs are bound. Along with three CGs in crystal structures, the five-membered lactone ring of GEV was located in the deeper site of the cavity and the steroid core was stabilized by interacting with the transmembrane helices αM1-6. As shown in Figure 1B, the predicted docking modes of the steroid core are similar to the conformation of CGs in the crystal structure complex, as the α-surface of the steroid cores were facing to the hydrophobic side chains of αM4-6 and the β-surface of the steroid cores were located near the polar side chains of αM1-2. Predicted docking orientation of the disaccharide of GEV was different from that of ouabain as well as from those of bufalin and digoxin. In the case of a docking mode based on the ouabain complex structure, the disaccharide was stretched to the αM1-2 loop. On the other hand, based on the bufalin and digoxin complex structure the disaccharide was in proximity to the αM7-8 loop. When we compared the crystal structure in complex with ouabain and with digoxin, the αM7-8 loop of digoxin complexed structure was located closer to the ligand than that of the ouabain complexed structure. Hence, the docking program calculated that GEV best interacts with the digoxin complexed structure through interaction with the αM7-8 loop. In the case of the bufalin complexed structure, the pore is narrower than the other structures as the αM1-2 loop is located closer to the cavity, and the six-membered lactone moiety of bufalin reaches deeper into the cavity. For these reasons, the GEV disaccharide likely interacts with the αM7-8 loop and the remaining part of the GEV molecule is located deeper in the cavity occupying the area enlarged by the bulky lactone moiety of bufalin. Altogether, based on our docking studies, we hypothesized that GEV achieves its cytotoxicity by inhibiting Na⁺/K⁺-ATPase.

GEV reduced the proliferation of A549 cells in a concentration and time-dependent manner (Figures 2A,B) concomitant with the accumulation of cell death after treatment with 50 and 100 nM GEV (24 h: 21% and 36%; 48 h: 40% and 53%). Observation by fluorescence microscopy of Hoechst-stained nuclei confirmed the progressive accumulation of cells displaying pyknotic nuclei, besides a minor additional fraction with nuclear fragmentation (Figures 2C,D). Analysis by transmission electron microscopy (TEM) further validated severe ultrastructural alterations of cytoplasm and nucleus in GEV-treated A549 cells. After 24 and 48 h, cells committed to death showed a disruption of the cytoplasmic architecture with an accumulation of large vacuoles, resembling intracellular edemas. Meanwhile, nuclei were shrunken but not fragmented, presenting locally condensed chromatin (Figure 2E). Remarkably, no cells with intermediate levels of degeneration were observed, suggesting a rapid evolution after cell death commitment. Real-time videomicroscopy of A549 cells treated with GEV at 50 nM confirmed a peculiar evolution of cellular alterations compared to etoposide (VP16), a canonical apoptosis inducer, used as a control (Figure 2F and Supplementary Videos 1–3). Whereas, VP16 sequentially triggered shrinkage, fusiform morphology, detachment, and fragmentation, GEV induced increased cell volume and granularity as quantified by flow cytometry (Figure 2G). Altogether, these observations suggested that GEV induced a differential, non-apoptotic, cell death pathway.

GEV treatment did not result in a significant level of caspase-3 cleavage or activity even at 100 nM after 48 h (Figures 3A,B), in contrast to VP16. Moreover, the pan-caspase inhibitor z-VAD did not prevent GEV-induced cell death (Figures 3C,D) or morphological alterations of A549 cells (Figure 3E and Supplementary Videos 1–5).

Considering the capacity of CGs to trigger autophagy and to verify autophagy involvement in GEV-treated cells, we monitored cell morphology by TEM (Figure 4A), together with the assessment of LC3I-II conversion, p62 and beclin-1 expression levels at earlier time points and different concentrations (Figure 4B and Supplementary Figure 2) in presence/absence of bafilomycin A1. All results excluded the induction of parossistic autophagy by GEV. Rather, the robust autophagic flux of A549 cells was abrogated by this treatment (Figure 4B and Supplementary Figure 2).

To assess for specific non-canonical cell death pathways potentially triggered by GEV, we treated A549 cells with poly-(ADP-ribosyl) polymerase (PARP)-1 inhibitor 3-aminobenzamide (3-ABA), the receptor-interacting protein kinase 1 (RIP-1) inhibitor necrostatin-1 as well as the calpain inhibitor PD 150606 which were unable to prevent GEV-induced reduction of A549 viability. Results were confirmed by real-time imaging (Supplementary Figure 3 and Supplementary Videos 6–11).

Of note, annexin-V-FITC/PI staining revealed an accumulation of cells exposing phosphatidylserine (PS) over the time of treatment with GEV, which could be prevented by the pan-caspase inhibitor z-VAD but not by several cathepsin inhibitors (Figures 3E,F and Supplementary Figure 4). This result confirms caspase-dependency involved in PS exposure and points at two possible alternative scenarios where apoptotic and non-apoptotic features of cell death might co-exist in different cell subpopulations or the same cell.

Hematopoietic non-adherent U937 cells were as susceptible to GEV as the lung epithelial A549 cells (Table 1). Figure 5A shows a reduction of the proliferation of U937 cells in a concentration-dependent manner, similarly to A459 cells. This effect was accompanied by the accumulation of around 40% of Trypan blue-positive cells after 48 h of treatment with 100 nM GEV (Figure 5B). GEV-treated U937 cells undergo changes in their nuclear morphology in a concentration-dependent manner leading to apoptotic nuclear condensation and fragmentation (Figures 5C,D). GEV induced cleavage of caspase-3 after 24 and 48 h detectable from 50 nM (Figure 5F) like etoposide. Moreover, z-VAD significantly rescued over 50% of U937 cells from nuclear fragmentation (Figure 5E). Altogether, these results indicate the ability of GEV of triggering a caspase-dependent apoptosis in U937 cells.

GEV induced an accumulation of approximately 30% cells in the G2/M phase after 24 and 48 h in A549 cells, accompanied by a decrease of cells in G1 phase and a small increase of cells in sub-G0/G1 phase concomitant with cyclin B1 accumulation at 10 nM followed by a reduction at 50 and 100 nM as early as 12 h. It also down-regulated p53 at all concentrations tested as earlier as 12 h (Figure 6). In U937 cells, GEV (100 nM) led to the accumulation of 32 and 81% cells in sub-G0/G1 phase after 24 and 48 h, respectively, with a corresponding decrease of cells in G1, compared to untreated controls (Supplementary Figure 5).

We then confirmed the potential of GEV to impair the ability of A549 cells in in vitro and in vivo 3D tumor formation assays in the presence of increasing concentrations of the compound. GEV strongly prevented A549 spheroid formation at 10 nM and completely abrogated their development at 50 nM (Figure 7A). In addition, colony formation was completely abolished by GEV at 50 nM (Figure 7B). To extend our colony formation assays to an in vivo situation, we assessed the capacity of GEV to abrogate A549 tumor formation in a zebrafish xenograft model. Our results confirmed a strong abrogation of tumor formation after injection of 50 nM GEV-pretreated, fluorescently stained A549, validating our in vitro results (Figure 7C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.