*3.2. In Silico Modeling*

Since CGs are well-known to be NKA inhibitors, we strived to determine whether Hyr and deHyr share the same fate. Using molecular docking, which is a method for studying protein-ligand complex interactions, we performed an in silico study of these CGs and NKA. The data were compared with Ob, Dg, and Cy. The structures of all five examined NKA ligands consisted of a steroid skeleton, which is substituted by a lactone and a saccharide moiety at the C-17 and C-3 position, respectively. Hyr and deHyr were docked into the NKA binding site, which is located in the transmembrane domain of the NKA α-subunit between the helices one to six. This binding site is divided into a polar (L-Gln111, L-Glu117, L-Asp121, L-Asn122, and L-Thr797) and nonpolar part (L-Ile315, L-Phe316, L-Gly319, L-Phe783, L-Phe786, and L-Leu793) [45]. In Figures 3 and 4, the far and near views of the CGs docked into NKA are shown in the lowest binding energy mode (Table 1). From Table 1, based on the binding energies, it is clear that the NKA-Ob and NKA-Hyr complexes were the most and the least stable ones, respectively, from the docked ligands.

**Figure 3.** Far view of ouabain (green) docked into Na+/K+-ATPase (purple) with the lowest binding energy. Images were taken using PyMOL 2.3.3 (Schrödinger, LLC, New York, NY, USA).

**Figure 4.** Near view of cardiac glycosides (stick representations) docked into Na+/K+-ATPase (purple, image representations; orange, individual residues, stick representations) with the lowest binding energy mode. The cardiac glycoside binding site in the Na+/K+-ATPase is indicated. (**A**) Ouabain, (**B**) cymarin, (**C**) digitoxin, (**D**) hyrcanoside (in blue) and deglucohyrcanoside (in green). The images were taken using PyMOL 2.3.3 (Schrödinger, LLC, New York, NY, USA).


**Table 1.** Binding energies of cardiac glycosides docked into Na+/K+-ATPase.

\* Ob = ouabain; deHyr = deglucohyrcanoside; Dg = digitoxin; Cy = Cymarin; Hyr = Hyrcanoside.

The NKA-Ob complex, which served as a reference, contained a hydrogen bridge between the conserved β-hydroxyl group at the C-14 position and L-Thr797, and between the hydroxyl group at the C-1 and C-19 positions with L-Gln111 and C-11 and C-5 with L-Asn122 and L-Glu117, respectively. Thus, Ob was docked into NKA as expected, i.e., with its polar surface facing the polar amino acids. Similarly, the nonpolar surface of Ob was oriented towards the nonpolar part of the NKA cavity.

Another CG docked into NKA, the second reference ligand Dg, interacted like Ob with L-Thr797 via the β-hydroxyl group at the C-14. Unlike Ob, Dg contains only one of the aforementioned β-hydroxyl groups at the C-14, but even so, the orientation of its steroid skeleton was identical to that of Ob, underlining the importance of nonpolar interactions (especially with L-Phe783 and L-Leu793) for CG binding to NKA.

Similarly, the other CG ligands were docked to the binding site of NKA with the same steroid skeleton orientation as Ob and Dg; β-Hydroxyl groups at the C-14 position interacting with L-Thr797 were also present in the NKA-deHyr and NKA-Cy complexes. In the case of Cy, interactions of the carbonyl group at the C-19 and β-hydroxyl group at the C-5 with L-Gln111 and L-Glu117, respectively, were also found. The interaction of the conserved β-hydroxyl group of C-14 with L-Thr797 was also present in the NKAdeHyr complex. However, it was not present in the NKA-Hyr complex; on the contrary, this hydroxyl group interacted with L-Glu117, which lies closer to the NKA cavity surface. The absence of Hyr interaction with L-Thr797 of NKA was because Hyr did not penetrate deep enough into the CG binding site of NKA as the other evaluated ligands. This is probably caused by the presence of β-D-glucopyranosyl as a second saccharide unit (with the first being β-D-xylopyranosyl in the case of both Hyr and deHyr, as is illustrated in Figure 1) at the C-3 position of Hyr. Hyr interactions with the key amino acid residues (L-Glu115, L-Glu116, and L-Arg886) of NKA were detected only from the β-D-glucopyranosyl unit, not from the β-D-xylopyranosyl, as it was in the case of deHyr interaction with L-Glu116 and L-Glu117.

For other glycosylated ligands (Ob, Dg, Cy), the saccharide moiety also contributed to their binding to NKA. Ob and Cy interacted with L-Arg880 and L-Glu312 by α-L-rhamnopyranosyl present in Ob and β-D-cymaropyranosyl present in Cy. In the case of Dg, there was an interaction of the first β-D-digitoxopyranose with L-Arg880 and L-Asp884 of NKA and the third β-D-digitoxopyranose with L-Arg886.

To summarize, all ligands (Ob, Dg, deHyr, Hyr, and Cy) that docked into the NKA binding site differed in the type of glycosylation at the C-3 position, as well as in the number of hydroxyl and carbonyl groups present at the steroid skeleton. As is evident from our findings, the number of these groups is important for the strength of the bond with NKA. Ob, which contains the most hydroxyl groups in its structure and is, therefore, able to form the most hydrogen bridges with NKA, was docked with the lowest binding energy (Table 1). The importance of these interactions, especially with L-Gln111 and L-Asn122, is documented by the crystal structure of NKA with Ob in [45], and in several mutagenesis studies in which a significant reduction in CG affinity to NKA was observed due to their substitution.

Dg, deHyr, and Cy have also docked as Ob to the CG binding site in NKA with the same orientation of the steroid skeleton as Ob in the aforementioned NKA crystal structure [45]. The same orientation of the steroid skeleton was also observed in the crystal structure of NKA with digoxin [46] and strebloside, which is a structural analog of Cy [47]. However, deHyr and Cy showed higher binding energies compared to Ob caused by a lower number of interactions with NKA, which is due to the lower number of substituents on the steroid skeleton. For the same reason, the reference ligand Dg also had higher binding energy compared to Ob. As for Hyr, it was also docked with the same steroid skeleton orientation as Ob and Dg, however, its binding energy was higher not only compared to these ligands, but also compared to other glycosylated ligands, including deHyr, which contains only one saccharide unit in its structure. Precisely due to the presence of two saccharide units, Hyr did not dock as deeply into the NKA binding cavity as did the other tested ligands. It has been reported that the number of carbohydrate units affects the strength of the NKA–CG interaction [48], and, for this reason, there was an increase in binding energy for Hyr compared to deHyr. However, this argument does not apply to Dg, which contains trisaccharide in its structure, meaning that the size of the binding energy is determined by a combination of both factors, i.e., the appropriate type of substituents on the steroid skeleton and the degree of glycosylation. Overall, Hyr was

docked with higher binding energy due to both the different number of substituents on the steroid skeleton and the different degrees of glycosylation compared to other glycosylated ligands.

#### *3.3. Anticancer Potential of the Evaluated Steroid Glycosides*

CGs as well-established therapeutics for the treatment of cardiac insufficiencies and arrhythmias have lately been subjected to drug repurposing, since it has been reported that they also exhibit great anticancer potential. Often, however, high systemic toxicity was also described. However, this phenomenon could be circumvented by higher cancer cell selectivity and/or a partial decrease in the overall toxicity. Therefore, as a next step, we evaluated the cytotoxicity and cancer cell selectivity of the CGs, Hyr, and deHyr (compared with Ob, Cy, and Dg), isolated from *C. varia* in a panel of human cancer cell lines. The activity was compared with toxicity results from noncancerous primary human cells, as well as mouse cells, which should be, in general, less sensitive to CGs due to different NKA isoforms.

The cytotoxicities of the CGs after 72 h of incubation were expressed as the halfmaximal inhibitory concentrations (IC50), which are summarized in Table 2. The results showed marked differences in *in vitro* toxicity between the CGs both in potency and selectivity. For all compounds, we detected a concentration-dependent cytotoxicity profile in all evaluated cell lines. From the results, it is obvious that Hyr and deHyr, even though they exhibited lower cytotoxicities to cancerous cell lines than that of well-described CGs Ob and Dg, manifested a good selectivity for cancer cells when compared to noncancerous cells. As for Hyr, the most pronounced selectivity (compared against MRC-5 cells) was observed for human cancer cells from lung (A549), pancreas (MiaPaCa-2), breast (MCF-7), and transformed kidney cells (HEK 293T). Fairly good selectivity was also detected for human leukemic cells, cells from colorectal carcinoma regardless of *p53* deletion, and cells from bladder carcinoma. Concerning deHyr, it exhibited the highest cancer cell selectivity (compared to BJ cells) for A549 cells, which was followed by good selectivity for leukemic, pancreatic, breast, prostate, and colorectal (with and without *p53* deletion) cancer cells. Dg and Cy shared almost identical selectivity (compared to BJ cells) to cancer cells derived from lung and colon carcinoma, as well as to leukemic cells (Ob mainly to A549 cells).

As expected, cytotoxicity of the tested CGs was significantly affected by the type of attached saccharide at the C-3 position of the steroid skeleton. Based on the data gained in this study, we concluded that the derivatives with one attached carbohydrate moiety exhibited the highest cytotoxicity compared to derivatives glycosylated to a higher extent, which is in agreement with what is known for the sugar vs. cytotoxicity relationship for CGs in general [13]. However, interestingly, the least toxic CG, Hyr, contains two saccharide moieties. It exhibited even lower toxicity than Dg, which contains trisaccharide. This contradiction might be explained by the molecular docking into NKA, in which the presence of the third carbohydrate unit stabilized Dg in the NKA cavity to a greater extent than in the case of Hyr.


**Table 2.** Summary of cytotoxic activities (IC50, nM) of the examined cardiac glycosides: ouabain (Ob), cymarin (Cy), digitoxin (Dg), hyrcanoside (Hyr), and deglucohyrcanoside (deHyr) after 72 h of incubation with cancerous and noncancerous human and mouse cells.

<sup>a</sup> Cytotoxic activity was determined by MTS assay following 72 h of incubation. The values represent the mean of IC50 from three independent experiments. The tested cell lines: CCRF-CEM (childhood T-cell acute lymphoblastic leukemia), K562 (chronic myeloid leukemia), A549 (lung adenocarcinoma), HCT116 (colorectal carcinoma), HCT116p53-/- (HCT116 with deleted *p53* gene), MiaPaCa-2 (adenocarcinoma of pancreas), MCF-7 (breast carcinoma), U-2 OS (osteosarcoma), 5637 (bladder carcinoma), PC-3 (prostate carcinoma), and HEK 293T (transformed kidney cells). Noncancerous human cells: MRC-5 (lung fibroblasts) and BJ (fibroblasts from foreskin). L929, mouse transformed fibroblasts. The colors in the first slope represent the types of the cells: orange—human cancerous or transformed cell lines, blue—human primary noncancerous cells, green—mouse cells.

Besides the number of saccharide moieties, the distribution of substituents on the steroid skeleton of CGs also significantly affected their overall cytotoxicity. As aforementioned in the docking part of this study, the binding site for CGs is divided in terms of amino acid distribution into a polar and nonpolar part, which means that the presence of polar substituents on the steroid skeleton affects the level of NKA inhibition to a greater extent. Of the monoglycosylated CGs in this study, the highest cytotoxicity was exhibited by Cy, which, like deHyr, contains a carbonyl group at the C-19, but also contains a β-hydroxyl group at the C-5 position. In contrast, Ob contains β-hydroxyl group at the C-1, α-OH at C-10, and C-19 positions, but it lacks the β-OH at the C-5. Thus, it seems that the β-hydroxyl group of the C-5 might contribute to the overall cytotoxicity. However, as evidenced by Levrier et al., the cytotoxicity is also significantly affected by the carbonyl group at the C-19; when a hydroxyl group at the C-19 is substituted for a carbonyl moiety, the cytotoxicity of the resulting derivative increases approximately 150-fold [49]. The carbonyl group is also present in Cy, which was the most cytotoxic from the CGs in this study. Therefore, based on our data, we conclude that the cytotoxicity of the five CGs evaluated was mainly influenced by the carbonyl group at the C-19, the β-hydroxyl group at the C-5, and the decreasing number of carbohydrate units.

The trend of the dose response curves for the evaluated CGs was somewhat similar for all examined human cell lines. Nonetheless, the situation substantially differed for mouse fibroblasts (L929; see Table 2), for which none of the CGs exhibited any signs of cytotoxicity up to the highest tested concentration (10 μM), which is probably caused by increased resistance of the mouse α-subunit of NKA to CGs [50]. Amino acid substitutions Q111R and N122D are present in the murine α-subunit of NKA [51] and, therefore, mouse cells can tolerate up to at least 1000 times higher concentrations of CGs than corresponding human cells [52].

#### *3.4. Steroid Glycoside Toxicity to Mouse Macrophages*

Even though it is known that CGs trigger immunogenic cell death and that they stimulate the immune response, not much is known about their effect on primary macrophages.

For the first time, we bring data on the effect of Ob, Hyr, deHyr, and Cy on these cells: primary macrophages, the innate immune cells. For this task, the cells were isolated directly from mice, and cell viability after 24-h CG treatment was determined by WST-1. The results are summarized in Figure 5. It is obvious that the viability of mouse macrophages was reduced in the presence of all tested compounds in comparison to untreated control cells, but, for some, only marginally. The viability of mouse macrophages treated with Ob decreased only by 10–20% (compared to the control) quite independently on the used concentration (1 nm–100 μM), while cytotoxicity of Cy and Hyr was more pronounced—about 60% of the control. The most cytotoxic to mouse macrophages was deHyr, which reduced their viability to 50% of the control. Surprisingly, a dose-dependent decrease in cell viability was not found for any of the compounds, despite a wide range of tested concentrations: 0.001–100.0 μM.

**Figure 5.** Viability of mouse peritoneal cells. Isolated cells were cultured for 24 h. Individual compounds, i.e., ouabain (Ob), cymarin (Cy), hyrcanoside (Hyr), and deglucohyrcanoside (deHyr), were applied at concentrations of 0.001, 0.01, 0.1, 1.0, 10.0, 50.0, and 100 μM. The effect of DMSO was also analyzed; its concentration corresponded to 50 μM concentration of compounds. WST-1 assay was used for viability evaluation. The results are expressed as the percent of untreated controls (ctrl) ± SEM (standard error of the mean) ofn=8 values from two independent experiments.

To summarize, in our in vitro conditions, the most cytotoxic was deHyr; Cy and Hyr made a position between deHyr and Ob. Cytotoxicity of Ob was previously studied in neuronal-like SH-SY5Y cells after 24 and 48 h, for which Ob reduced their viability by 40 and 10%, respectively [53]. Distinct effects of Ob on the survival of human and rat vascular smooth muscle cells, endothelial cells, and astrocytes were confirmed by Akimova et al. [54]. Unlike human cells, their rodents counterparts perfectly survived in the presence of high concentrations of Ob (3–3000 μM), despite the complete inhibition of the NKA and inversion of the [Na+]I/[K+]i ratio. These dramatic differences in Ob effects on rat and mouse cells are driven by variations of rodent α1 NKA isoform. This difference could also explain the non-existing dose-dependent curve in mouse macrophages.

It is generally accepted that CGs, such as clinically used digoxin, are cytotoxic. This is a rational reason to employ such compounds in anticancer therapy instead of the regulation of cardiac function. It seems that some cancer cell lines (especially the ones derived from lung and colon carcinoma) are more sensitive to cardenolides consisting of a lactone ring with five carbons than noncancerous cells. Recent data also document that another lesser known cardenolide derivative, nerigoside, was more cytotoxic in two colorectal cancer cell lines HT29 and SW620 when compared to normal human epithelial cell line NCM460 (determined by a similar in vitro viability assay, as in our study) [55]. Our pilot results showed that the cytotoxic effect of the tested CGs on mouse macrophages was not fully devastating, but the results significantly differed from mouse fibroblasts, for which no toxicity was detected up to 10 μM concentration, while for mouse macrophages, only Ob did not exhibit significant cytotoxicity (up to 100 μM concentration), which corresponds to the known fact that mouse cells are generally insensitive or less sensitive to CGs than human cells based on the expression of different NKA isoforms. Contrary to L929, in mouse macrophages, Cy reached IC50 already at the lowest tested concentration of 1 nM, deHyr ca. at 10 nM, and Hyr at ca. 100 nM concentration. The reason for this difference between the two types of mouse cells remains elusive.

#### *3.5. Cell Cycle and Cell Death Analysis*

Next, we wanted to find out whether Hyr and deHyr can arrest the cell cycle of cancer cells, as has been previously reported for Dg and Ob, which were evaluated in cancer cells of various origin. For a more detailed description of the biological activity of the studied derivatives, we performed cell cycle analysis of the most sensitive CCRF-CEM cells after 24 h of CG treatment (Table 3).

**Table 3.** Effect of cytotoxic compounds on cell cycle, apoptosis, and DNA/RNA synthesis in CCRF-CEM lymphoblasts (% of positive cells). Flow cytometry analysis was used to quantify cell cycle distribution and the percentage of apoptotic cells. The sum of the percentages for G0/G1, S, and G2/M is equal to 100%. <sup>a</sup> Phospho-Histone3 (Ser10); <sup>b</sup> BrDU, 5-bromo-2-deoxyuridine; <sup>c</sup> BrU, 5-bromouridine.


\* Ob = ouabain; deHyr = deglucohyrcanoside; Dg = digitoxin; Cy = Cymarin; Hyr = Hyrcanoside.

After 24-h incubation with 1× IC50 of CGs, CCRF-CEM cells were still viable. The percentage of the sub G0/1 population was only slightly increased compared to the untreated control, while treatment with 5× IC50 caused a typical increase in the cell number present in the sub G0/G1 phase, together with augmented DNA fragmentation. All five evaluated CGs induced an increase in the G2/M phase population. The decrease in cells in the S and G0/G1 populations was caused by the proportional accumulation of cells in the G2/M phase. We did not observe positivity for pH3Ser10, and negativity indicated G2 arrest. Such a finding is consistent with a recently reported study showing ATR-CHK2-CDC25Cmediated G2/M cell cycle arrest after Dg treatment [56].

BrDU is incorporated into newly synthesized DNA, and BrDU pulse labeling is therefore commonly used as a proliferation marker. Low BrDU incorporation into the DNA of treated cells with all compounds at 5× IC50 reflected inhibition of DNA synthesis, indicating irreversible apoptotic changes. The percentage of BrU positive cells incorporating 5-bromouridine is proportional to the transcriptional activity of CCRF-CEM cells.

These values significantly decreased upon treatment with 5× IC50 of CGs, but not with 1× IC50.

Previously, the ability to arrest the cell cycle of cancer cells in the G2/M phase has been already described for Ob, oleandrin, digoxin, Dg, and its synthetic analog monoD [26,57–59]. In this study, the same effect was observed for all tested CGs, although in some, there was an increase in the G2/M phase already at 1× IC50, and in others only at 5× IC50. For substances in which the G2/M phase increased at 1× IC50, on the contrary, it decreased at 5× IC50, which means that each substance had its concentration optimum, above which G2/M decreased again. The same trend was observed by Elbaz et al. [26]. Lower incorporation of BrDU and BrU at 5× IC50 concentrations indicated that DNA and RNA synthesis decreased at these concentrations, although in some cases the number of S-phase cells may be higher, even though nucleic acid synthesis is no longer present and thus, the incorporation of BrDU and BrU decreases.

#### **4. Conclusions**

In this work, we describe a procedure for the isolation of two CGs, Hyr and de-Hyr, by aqueous EtOH extraction from seeds, with an overall yield of 3.3% and 0.0002%, respectively. This was an improvement by several times of magnitude than what has been reported so far. Both CGs, Hyr and deHyr, were assessed for their anticancer activity, which was compared to other well-known CGs Ob, Dg, and Cy. From the results, it is clear that Ob and deHyr outperformed the other CGs, which correlates with the docking study into NKA. The highest anticancer (based on the cancer cell selectivity) potential of all CGs was found against cells derived from lung and colorectal carcinoma. Moreover, all evaluated CGs arrested the cell cycle of CRF-CEM in the G2/M phase. Thus, even though further elaboration in deciphering detailed mechanisms of Hyr and deHyr and other CG anticancer activity is needed, the first results already indicate that these CGs could have a potential for a therapeutic application in cancer treatment. From the number of publications on CGs as anticancer drugs, it is obvious that they have come into the foreground as candidates of anticancer therapy with new mechanisms of actions than the standardly used chemotherapeutics, such as antimitotics [60] or cisplatin, and that their induction of the immune system response brings another added value in the treatment. We assume that our results can contribute to further development of drugs based on NKA interaction, and maybe additional molecular targets with selective cytotoxicity for cancer cells.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2304-8 158/10/1/136/s1; Figure S1: Other known components of *C. varia*; Figure S2: Structures related to naming mismatch; Figure S3: Structures of so-called "securigenin glycosides"; Figure S4: The 1H and 13C chemical shifts of deHyr in CD3OD and CD3SOCD3; Figure S5: The 1H and 13C chemical shifts of Hyr in CD3OD and CD3SOCD3; Figures S6 and S7: MS spectrum of Hyr and deHyr, respectively; Figures S8 and S9: HPLC chromatograms of Hyr and deHyr, respectively; Table S1: 1H and 13C NMR characteristics of Hyr (securigenin s3), deHyr and related securigenins s4 and s5.

**Author Contributions:** S.R., T.Z., M.J., P.B.D., P.D., B.D., M.H., T.R., L.O., and E.K. conceived and designed the experiments; T.Z., S.R., B.D., M.J., J.B., E.K., P.C., L.K., S.G. and P.D. performed the experiments; T.Z., M.J., P.B.D., S.R., E.K., and B.D. analyzed the data; S.R., M.J., J.B., and P.D. wrote the article. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by an internal grant from the budget for the implementation of the activities of the Institutional Plan of the UCT Prague in 2020, the grant of Specific university research A1\_FPBT\_2020\_001, A2\_FPBT\_2020\_015, grant No. A1\_FPBT\_2020\_004, the grant of the Czech Ministry of Education, Youth and Sports (CZ-OPENSCREEN—LM2018130 and EATRIS-CZ— LM2018133), OPPC CZ.2.16/3.1.00/24503, NPU I LO1601, the internal grant of Palacký University (IGA\_LF\_2020\_019) and Czech Science Foundation 14-04329S.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki. All protocols were approved by the institutional ethics committee (No. MSMT15894/2013-310).

**Informed Consent Statement:** Not applicable for studies not involving humans.

**Conflicts of Interest:** The authors declare no conflict of interest.
