Agarose, propidium iodide (PI), acridine orange, calf thymus DNA (CT-DNA), tetrazolium salt (MTT), dimethylsulfoxide (DMSO), bovine serum albumin (BSA), tetramethylrhodamine methyl ester (TMRE), 5,5'-dithiobis(2-nitrobenzoic acid) (DTNB) and the protease inhibitor cocktail were purchased from Sigma-Aldrich; Merck KGaA. Antibiotics, Dulbecco's modified Eagle's medium (DMEM), Dulbecco's phosphate-buffered saline (PBS) and fetal bovine serum (FBS) were purchased from Gibco; Thermo Fisher Scientific, Inc. The kit of PI/ribonuclease A (RNAse) were acquired from Immunostep. Phosphatase inhibitor cocktail was purchased from Calbiochem (Merck Biosciences). Rabbit polyclonal antibodies raised against CDK2 (cat. no. sc-163), cyclin A (cat. no. sc-596), cyclin B1 (cat. no. sc-752), PARP1 (cat. no. sc-7150), actin (cat. no. sc-7210) and mouse monoclonal antibodies raised against Bax (cat. no. sc-7480), Bcl-xL (cat. no. sc-8392) and p53 (cat. no. sc-126) were from Santa Cruz Biotechnology. Rabbit polyclonal antibody raised against phospho-Rb (cat. no. 9308) was acquired from Cell Signalling Technology. Polyclonal goat secondary antibodies for anti-rabbit IgG (cat. no. AP132P) and for anti-mouse IgG (cat. no. AP181P), as well as the chemiluminescence detection kit of HRP-coupled antibodies were from Merck Millipore. All other chemicals used were American Chemical Society (ACS) grade reagents.

Human breast carcinoma (MCF-7), cervix adenocarcinoma (HeLa) and normal mouse fibroblast (McCoy) cells were purchased from the Rio de Janeiro cell bank, Brazil. The cells were maintained at constant temperature (37˚C) under a 5% CO₂ atmosphere with 95% air humidity. The culture medium used was DMEM supplemented with 10% FBS, penicillin (100 U/ml) and streptomycin (100 µg/ml).

The cell viability was assessed using the MTT assay (14). Cells were plated at a density of 1x10⁴ cells/well in 96-well plates. After reaching 80% confluence, the cells were treated with different concentrations of harmine (0.1 to 1,000 µM) for 48 and 72 h. The negative control was treated only with standard culture medium containing 0.1% DMSO. After the treatment, cells were washed with PBS and incubated for 2 h with MTT (0.5 mg/ml). The formazan crystals were solubilized by adding DMSO (100 µl/well), and the colored solutions were read at 550 nm. Three independent experiments were conducted and the results are presented as the half maximal inhibition concentration (IC₅₀), calculated using Graph Pad Prism 6 (GraphPad Software Inc.). The selectivity index (SI) of the compounds under study are expressed as previously reported by Koch et al (15), with a minor modification: SI (selectivity index) = IC₅₀ of the compound in a normal cell line/IC₅₀ of the same compound in the cancer cell line.

The type of cell death induced by harmine was evaluated by a staining method, using PI and acridine orange (16). This method allows the differentiation of viable cells (green) from those that are dying through apoptosis (orange) or necrosis (dark red). MCF-7 cells were seeded (2x10⁵/well) in 6-well plates. After confluence was reached, the cells were treated with IC₃₀ and IC₅₀ concentrations of harmine for 72 h. Fresh medium was used as a negative control. After treatment, the cells were stained with PI (100 mg/ml) and acridine orange (100 mg/ml) (5 µl; 1:1 v/v) and then visualized using an Olympus microscope, model BX41 (Olympus Corporation). Results are expressed as the percentages of cells that were viable, apoptotic or necrotic.

Mitochondrial membrane depolarization was determined using the fluorescent probe, TMRE, which accumulates in the mitochondrial matrix in proportion to mitochondrial membrane potential (ΔΨm) (17). MCF-7 cells were seeded at a density of 1x10⁶/well on 6-well plates. After confluence the cells were treated with IC₃₀ and IC₅₀ concentrations of harmine for 48 h. Thereafter, the cells were incubated in the dark with TMRE (25 nM) for 20 min at 37˚C. Cell fluorescence was determined using a BD FACS Canto II flow cytometer (BD Biosciences) and results are presented as the percentages of cells with high TMRE fluorescence.

The DNA content of the cells was measured by flow cytometry using a PI (50 µg/ml) and RNAse (0.2 mg/ml) solution kit. MCF-7 cells were seeded (2x10⁵/well) in 6-well plates. After confluence was reached, the cells were synchronized using nocodazole (30 ng/ml) for 14 h and then treated with harmine (IC₃₀) for 72 h. Subsequently, the cells were fixed overnight with ethanol (70%) at -20˚C, washed with a PBS/albumin (2%) solution, and later incubated with the PI/RNAse solution for 15 min at room temperature. The cells were evaluated using the BD FACS Canto II flow cytometer (BD Biosciences) and categorized into each phase of the cell cycle, according to the content of DNA. Flowing Software 2.5.0 (Perttu Terho; Cell Imaging Core, Turku Center for Biotechnology, University of Turku) was used to process the data.

The potential interaction of harmine with calf-thymus DNA (CT-DNA) was assessed through Spectrophotometric UV-Visible Scanning Titration (18). Increasing concentrations of harmine (0-250 µM) were used for an absorption titration assay with a constant concentration of CT-DNA (150 µM). To obtain the spectra of the samples, a Hitachi U-2910^® spectrophotometer was used; and UV-Visible scanning was performed from 200 to 800 nm. The changes in CT-DNA absorbance, after incubation with harmine, as well as the maximum absorption wavelength shift, were determined.

Potential DNA intercalation was evaluated by fluorescence titration measurements, using the DNA-intercalating agent PI, according to Da Silveira et al (19). CT-DNA (150 µM) was saturated with PI (300 µM) in a phosphate buffer (50 mM) containing 0.1 M NaCl (pH 7.4). Increasing concentrations of harmine (0-250 µM) were then incubated with fixed concentrations of CT-DNA and PI for 10 min at room temperature. Doxorubicin, a standard antitumor intercalating drug, was used as a positive control. The SpectraMax Paradigma^® Multileader was used to measure the variation of the sample fluorescence. Excitation and emission wavelengths of 492 and 620 nm were used, respectively.

The circular dichroism (CD) DNA interaction assay was performed as described by Bertoldo et al (20). The CD spectra were measured in a JASCO-810 spectropolarimeter (JASCO). A fixed concentration of CT-DNA (150 µM), prepared in a phosphate buffer (50 mM) containing 0.1 M NaCl (pH 7.4), was titrated with increasing concentrations of harmine to achieve a molar ratio of 0.04, 0.05 and 0.07 DNA/harmine. All CD spectra of DNA and DNA/harmine were recorded over the range of 220-350 nm and the final data were expressed in millidegrees (mdegs).

The B-DNA template structure sequence d (CGTGAATTCACG) (PDB ID1G3X) was retrieved from Protein Data Bank (21). The ligand INE.mol2 file came from the ZINC database (http://zinc.docking.org/substance/18847046) and was minimized after removal of original co-precipitated ligands and water with UCSF Chimera 1.13(22) using the AMBER99bsc1 force field (23). The ACPYPE tool, based on Python (24), was used as an Antechamber to generate INE topology for molecular simulations (25). Molecular docking simulations were applied to AutoDock MGLTools 1.5.6rc3 to generate a DNA-INE complex input file (.pdbqt) (26). Geometries and binding affinity energies were then calculated through AutoDock Vina 1.2.2 algorithm, using default parameters (27). H-bonds and/or hydrophobic interactions predicted between INE and DNA nucleotides were visualized with PyMOL open source 1.8.7.0’ for Python 3.6(28) and with LigPlot⁺ 2.1(29). However, subsequent molecular dynamic (MD) simulations used only the ligand pose (pdbqt) that had the lowest RMSD and binding affinity. Avogadro (30) was used to convert this file to .pdb format, before MD simulations with GROMACS version 2018.4 (31,32) using an AMBER99bsc1 force field (23). The TIP3P water model was applied in order to solvate and neutralize the system with two Na+ ions inside an octahedron box (33). The runs made for energy minimization and solvent equilibration kept grid space under NVP and NPT ensemble conditions (T=300 K, P=1 bar). All simulations used the same time step (1 fs). Position restricted heavy atoms of DNA accessible to the solvent, without disturbing the complex structure.

MM-PBSA calculations performed after MD simulations showed some conformational fluctuations and residue energy contributions for free binding energy (34). The MM-PBSA method calculated the energies and trajectories related to DNA and the ligand INE complex. This method, using explicit solvent, also enables the calculation of non-bonded potentials (electrostatic and van der Waals) under default parameters. Results were visualized with Visual Molecular Dynamics (VMD) (35) for LINUXAMD64, version 1.9.4a12 (2017). Calculations of trajectories used Verlet through MD neighbor searching and the vdW cutoff-scheme (36). Additionally, the ‘readHBmap.py’ (Python 2) program showed frames with H-bonds. Additionally, H bond occupancy graphics showed frames (1 frame equal to 2 ps) where this interaction contributed to the macromolecule - ligand complex stability (37). Finally, frames selected were visualized (Chimera and LigPlot⁺ 2.1) and key intermolecular interactions between DNA-ligand were evaluated.

The DNA fragmentation was evaluated by the comet assay, according to Singh et al (38). MCF-7 cells were seeded (2.5x10⁴/well) in 24-well plates. After confluence, the cells were treated with IC₃₀ and IC₅₀ concentrations of harmine for 72 h. After resuspending the cells in low-melting point agarose (0.75%), they were placed on slides pre-covered with agarose (1.5%) and incubated for 10 min at -8˚C. The cells were then lysed for 7 days using the lysis solution (2.5 M NaCl; 10 mM Tris; 100 mM EDTA; 1% Triton X-100 and 10% DMSO; pH 10.0) and then submitted to horizontal electrophoresis (25 V and 300 mA) in alkaline buffer (300 mM NaOH; 1 mM EDTA; pH 13.0) for 20 min at 8˚C. Subsequently, the slides were neutralized and fixed with the respective solutions; neutralizing solution (0.4 M Tris HCl; pH 7.5) and fixing solution (15% TCA; 5% ZnSO₄; 5% glycerol). These procedures were intercalated with water washing. After drying at room temperature, the slides were stained (0.1% NH₄NO₃, 0.1% AgNO₃, 0.25% tungstosilicic acid, 0.15% formaldehyde, 5% Na₂CO₃, v/v), washed with acetic acid (0.01%) and visualized on an Olympus microscope, model BX41 (Japan). The comets were classified according to Ross et al (39), in which the score 0 represented the undamaged nuclei and a score of 4 represents maximally damaged nuclei.

MCF-7 cells were treated with the IC₃₀ concentration of harmine for 48 h and then washed with PBS and lysed in RIPA buffer (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1% NP40; 0.25% Na-deoxycholate and 1 mM phenylmethylsulfonyl fluoride), supplemented with protease inhibitor (1%) and phosphatase inhibitor (3%) cocktails. Laemmli buffer (60 mM Tris-HCl, 2% sodium dodecyl sulphate (SDS), 10% glycerol, 5% β mercaptoethanol and 0.01% bromophenol blue, pH 6.8) was used to denature the proteins. Subsequently, 25 µg of denatured protein were submitted to sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by electrotransfer to polyvinylidene fluoride (PVDF) membranes (40). Following blocking and washing, the membranes were incubated with the primary antibodies overnight at - 8˚C (Rabbit polyclonal antibodies raised against phospho-Rb (1:1,000, v/v), CDK2 (1:1,000, v/v), cyclin A (1:1,000, v/v), cyclin B1 (1:1,000, v/v) and PARP1 (1:1,000, v/v); Mouse monoclonal antibodies raised against Bax (1:200, v/v), Bcl-xL (1:200, v/v) and p53 (1:200, v/v). Lastly, the membranes were washed and incubated with secondary antibodies for at least 1 h. The polyclonal goat anti-rabbit IgG antibody (1:5,000, v/v) and polyclonal goat anti-mouse IgG antibody (1:3,000, v/v) are both peroxidase conjugated. A chemiluminescence detection kit for HRP-coupled antibodies was used for immunodetection. Actin (1:1,000, v/v) was used as a loading control. Images acquired (ChemiDoc MP System; Bio-Rad Laboratories, Inc.) were normalized with actin.

The antitumor activity of harmine was evaluated using male Balb/c mice (20±2 g, n=12) kept under controlled conditions (12 h light-dark cycle, 22±2˚C, 60% air humidity), receiving water and food ad libitum. Previous tests were done to select the maximal safe dose of harmine with the optimal dilution and the doses of 10 mg/kg/day and 20 mg/kg/day were chosen to continue the experiments. The induction of Ehrlich ascitic carcinoma (EAC) was carried out according to a method previously reported by Kviecinski et al (41). Before tumour induction, all mice were weighed (g) and their abdominal circumferences (cm) were measured. The mice were then inoculated intraperitoneally with EAC cells (5x10⁶ cells/200 µl). After 24 h, the animals were divided into 4 groups (n=12). The negative control group was treated with saline containing 1% DMSO. The positive control group received doxorubicin (0.6 mg/kg/day) as previously reported (42). Test groups received harmine at 10 or 20 mg/kg/day, respectively. The treatments were administered intraperitoneally for 9 consecutive days. Twenty-four hours later, all mice were weighed and their abdominal circumferences measured again. Using the formula reported by Felipe et al (43), the inhibition (%) of tumour growth was calculated as follow: [(variation in waist circumference of the treated group x 100)/variation in waist circumference of the control group] - 100. The variation in body weight of the treated animals was calculated as reported by Kviecinski et al (41): Final weight (after 9 days of treatment) - Initial weight (day zero, day of tumour inoculation). Then, the animals (n=12) were kept alive and assisted on a daily basis to evaluate survival (lifespan), according to Kaplan and Meier (44). Lastly, the treatment was repeated and the animals were euthanised after 9 days of treatment. The ascitic fluid was collected in graduated falcon conical tubes and centrifuged at 5,000 x g for 5 min to measure packed tumour cells volume (41). Viability of the tumour cells was assessed through the trypan blue assay (45).

All in vitro assays were performed in technical triplicates, whereas the in vivo experiments were performed in biological replications (n=12; α=0.05; Power (1 - β)=0.90; critical t=1.81; df (degree of freedom)=10; Effect size |ρ|=0.67). The results were presented as means ± standard deviations or as percentages. Data were analysed using analysis of variance one-way ANOVA followed by the Bonferroni post-hoc test or log-rank test, when necessary. Comparisons were performed using the software GraphPad Prism 6 (GraphPad Software Inc.). P<0.05 was considered to indicate a statistically significant difference.

The cytotoxic effect of β-carboline alkaloid harmine was evaluated against two tumour cell lines (MCF-7 and HeLa) and a normal cell line (McCoy). The values obtained for the IC₅₀ and SI for harmine in 48 and 72 h are shown in Fig. 1A. Harmine significantly decreased tumour cell viability in a time and dose-dependent manner (Fig. 1B and C). Additionally, harmine showed selectivity for all tumour cell lines tested (Fig. 1A) and it is worth noting that this alkaloid presented the highest selectivity for MCF-7 cells (SI=4.44).

Results indicated that harmine was able to induce significant cell death (Fig. 2A) and the numbers of apoptotic cells in the presence of this alkaloid (27.88 and 46.47 µM) were significant. However, only the concentration of 46.47 µM harmine caused an increase in necrotic cells (8.67±3.47%), when compared to the negative control (1.81±0.70%). Also, harmine (46.47 µM) decreased the number of viable cells by approximately 32.9% while increasing the number of apoptotic (82.3%) and necrotic (79.1%) cells, compared to the negative control.

The modulators effects of harmine on pro- and anti-apoptotic proteins were investigated by exposing MCF-7 cells to harmine (27.88 µM) for 48 h. Fig. 2C and D show that harmine increased the expression of Bax protein, while reducing the expression of Bcl-xL protein. Accordingly, the Bax/Bcl-xL ratio was significantly increased when MCF-7 cells were treated with harmine, suggesting cell death through apoptosis via the mitochondrial pathway. However, harmine induced apoptotic cell death independently of p53, since MCF-7 cells show underexpression of this pro-apoptotic protein (Fig. 2C and E).

Our results showed that harmine induced a dose-dependent depletion of ΔΨm in the MCF-7 cells (Fig. 2B). At concentrations of 27.88 and 46.47 µM harmine showed a ΔΨm of 79.65±1.85 and 72.26±2.41%, respectively, whereas the negative control was 85.45±3.08%. Thus, harmine at the concentration of 46.47 µM was capable of decreasing the ΔΨm of MCF-7 cells by 15.4%, when compared to the negative control.

Fig. 3A and B show the changes in the MCF-7 cell cycle after treatment with harmine. Cells from the negative control had 65.50±0.96, 13.41±1.03 and 21.01±1.59% of cells in G₁, S and G₂/M phase, respectively. The treated cells presented 67.70±0.85, 3.53±0.58 and 26.12±0.38% of cells in G₁, S and G₂/M phase, respectively. Harmine (27.88 µM) promoted a significant decrease (73.7%) in the number of cells in the S phase and a significant increase (approximately 19.6%) in the number of cells in the G₂/M phase, when compared with the negative control. However, MCF-7 cells treated with harmine was not statistically different in relation to the number of cells in G₁ phase, when compared to untreated cells, suggesting a G₂/M arrest of cells. In addition, after the treatment there was an increase of cells in the sub-G₁ area (2.67±0.11%), compared to the negative control (0.08±0.08%), which is indicative of cell death through apoptosis, confirming the previous result (Fig. 2A).

Treatment of MCF-7 cells with harmine (27.88 µM) for 48 h totally inhibited the phosphorylation of pRb (Fig. 3C and D). Additionally, harmine caused a significant decrease in the expression of cell cycle proteins, CDK2, cyclin A and cyclin B1 (Fig. 3C and D), which may explain the significant reduction of cells present in the S phase and cell cycle arrest at the G₂/M phase (Fig. 3A and B).

Firstly, the interaction between harmine and CT-DNA was analysed by spectrophotometric UV-visible scanning titration, causing both hypochromic and bathochromic effects (Fig. 4A); thereby suggesting that β-carboline alkaloid binds to DNA by intercalation.

In order to find further evidence of the intercalative interaction of harmine with CT-DNA, spectrophotometric titrations were carried out using the DNA-intercalating agent, PI. When the concentration of harmine was increased (Fig. 4B), PI fluorescence decreased, suggesting that harmine succeed in intercalating between the nucleobase pairs of DNA and to displace PI bound to DNA, consequently causing a reduction in the fluorescence intensity of the sample.

CD was also employed to gain insight into the DNA conformational alterations that occur after harmine was bound to DNA. A typical CD spectrum of CT-DNA in its B form shows a positive band with a maximum at 275 nm, due to base stacking, and a negative band with a minimum at 248 nm, due to right-handed helicity (46). Therefore, alterations in the B-DNA secondary structure lead to a change in the CD spectrum (47). Our results showed that harmine was able to induce changes in the CD spectrum by increasing the intensity of the negative and the positive bands of CT-DNA (Fig. 4C). Accordingly, the CD spectrum results from the DNA-harmine complex were consistent with those of the intercalation model.

Molecular docking and dynamic studies of the binding model of harmine with DNA were used to gather further structural details of the DNA-harmine complex. The pose that had the lowest binding energy, predicted by AutoDock Vina (-7.6 kcal/mol; RMSD 0.00000) after submission to MD simulations (1 ns), showed harmine to be intercalated between the hydrophobic portions of nucleotide residues dt620 and dt619 in the strand B of DNA (Fig. 5A). The frame correspondent to 10 ps showed one H-bond between the harmine donor atom N14, which shares the H atom (H15) with the acceptor atom (O4') of residue dt620 (2.007 Å) (Fig. 5A), although this H-bond presents only 2.8% occupancy. Fig. 5B (bottom) also has other H-bonds, which act as acceptors to nucleotides dt619 (0.4%) and da606 (0.2%). The nucleotide da606, however, formed one H-bond with harmine (0.2%) and, for this reason, acted as an H-bond donor. Additionally, variations in the RMSD of harmine (Fig. 5B, insert) indicated an initial stability of the complex that was briefly maintained by H-bonding. Probably fluctuations of RMSD (~5Å) come from the DNA-harmine hydrophobic interactions (residues da605, da606, dt619 and dt620) initially predicted by AutoDock Vina (Fig. S1A), which had the highest hydrophobic contribution value (11.78219) among other predicted poses of harmine. The same hydrophobic interactions remain after MD simulations (Fig. 5C). Additionally, the H-bonds predicted by GROMACS showed a predominance of pairs with distances of up to 3.5Å (Fig. S2A), confirming the H-bond distance progression variations of close to 3.5Å (Fig. S2B). Indeed, the H-bond angle distribution from the ideal of 150˚ to a less favourable angle (50˚) (Fig. S2C) confirmed the minor role of H-bonds in the maintenance of harmine inside the DNA hydrophobic cavity.

The MM-PBSA method predicted some energies in the full DNA-harmine complex. The non-ligated potential corresponding to van der Waals energy was -121.93±0.81 kJ/mol, while the electrostatic energy was 14.55±0.28 kJ/mol. The value of complex binding energy (-106.42±0.82 kJ/mol) corresponded to the polar solvation energy (14.54±0.28 kJ/mol) and non-polar solvation energy related to the Solvent Accessible Surface Area (SASA) (-10.55±0.06 kJ/mol) plus molecular mechanic energy (Fig. S2D). Additionally, MM-PBSA calculated the van der Waals energy that represented the main contribution to the molecular energy of the complex at vacuum (Fig. S2E). The interior of the harmine complex had the lowest energy value, which contributed to its total energy (Fig. 5D). Harmine also interacted with the water molecules associated with the solvation shell (Fig. 5C). Finally, explicit solvent modelling (TIP3P) from GROMACS simulations allowed MM-PBSA to calculate the DNA-harmine complex SASA Free Energy of Solvation (ΔGsolv) (Fig. S2F). The ΔGsolv was maintained at approximately -44 kJ/mol throughout the assay. This confirmed that the energy required to move harmine from the aqueous solvent to the non-polar environment (DNA) was very low, which favoured the DNA-harmine interactions.

After treating MCF-7 cells with 27.88 or 46.47 µM, harmine caused DNA damage indexes of 140.61±28.66 and 184.35±23.66%, respectively, while the negative control index was 72.26±27.72% (Fig. 4D). In addition, 27.88 and 46.47 µM harmine increased DNA damage in MCF-7 cells by 48.6 and 60.8% respectively, when compared to the untreated cells.

Furthermore, harmine (27.88 µM, 48 h) significantly decreased the expression of PARP1 (Fig. 4E and F), suggesting a possible reduction of the PARP1-dependent DNA repair mechanism.

The treatment of Balb/c mice with harmine was able to significantly inhibit tumour growth, in a dose-dependent manner, compared to the negative control (Fig. 6A). At the concentration of 10 mg/kg/day, harmine inhibited tumour growth by 16.77±3.88%, whereas the concentration of 20 mg/kg/day had a significantly greater effect, inhibiting tumour growth by 31.10±4.19%. Doxorubicin (0.6 mg/kg/day) used as a positive control was able to inhibit tumour growth by 53.20±5.02%.

Harmine also reduced significantly the variation of body weight and the volume of ascitic fluid, in a dose-dependent manner, when compared to the negative control. However, it is noted that only harmine at concentration of 20 mg/kg/day was able to decrease significantly the volume of packed tumour cells (2.13±0.65), compared to the negative control (3.60±0.49) (Fig. 6B). Furthermore, the proportion of non-viable/viable tumour cells increased when the treatment was done with 20 mg/kg/day of harmine (Fig. 6C). Doxorubicin (0.6 mg/kg/day) also reduced significantly the evaluated parameters.

The effect of harmine on mouse lifespan was evaluated using the method of Kaplan and Meier (1958) (Fig. 6D). Animals receiving 10 mg/kg/day of harmine demonstrated a small prolongation of lifespan (42.0%), when compared to the negative control group (36.7%). In contrast, animals receiving 20 mg/kg/day of harmine presented a significantly longer lifespan (68.3%), when compared with the negative control. No mortality occurred in animals treated with doxorubicin (0.6 mg/kg/day).