Design, synthesis, and biological evaluation of new 6,N2-diaryl-1,3,5-triazine-2,4-diamines as anticancer agents selectively targeting triple negative breast cancer cells

New 6,N2-diaryl-1,3,5-triazine-2,4-diamines were designed using the 3D-QSAR model developed earlier. These compounds were prepared and their antiproliferative activity was evaluated against three breast cancer cell lines (MDA-MB231, SKBR-3 and MCF-7) and non-cancerous MCF-10A epithelial breast cells. The synthesized compounds demonstrated selective antiproliferative activity against triple negative MDA-MB231 breast cancer cells. The most active compound in the series inhibited MDA-MB231 breast cancer cell growth with a GI50 value of 1 nM. None of the tested compounds significantly affected the growth of the normal breast cells. The time-dependent cytotoxic effect, observed when cytotoxicity was assessed at different time intervals after the treatment, and morphological features, observed in the fluorescence microscopy and live cell imaging experiments, suggested apoptosis as the main pathway for the antiproliferative activity of these compounds against MDA-MB231 cells.


Introduction
Despite signicant advancements in cancer therapy, cancer remains one of the diseases having the most negative impact on society. According to the World Health Organization, cancer was the second leading cause of patient lethality in 2018 causing almost 10 million deaths. 1 Moreover, the cancer prevalence and mortality from cancer have been continuously growing worldwide, in both developing and developed countries. It was projected that from 14 million people suffering from cancer in 2012 the number of new cases per year will double by 2030.
Breast cancer had the highest incidence rates among all types of cancer in 2018 (46.3 per 100 000 females). In females, breast cancer is the most frequently diagnosed type of cancer and the prevalent cause of cancer deaths. 1 Breast cancer is a rather heterogeneous form of cancer with cancer cells significantly varying in their properties and thus requiring different therapeutic approaches. 2 On the basis of presence or absence of molecular markers, breast cancer is classied into 4 main subtypes: (1) human epidermal growth factor 2 (ERBB2) positive cancer with cells expressing ERBB2, (2) luminal A breast cancer with cells expressing estrogen or progesterone receptors but not ERBB2, (3) luminal B breast cancer with cells expressing hormone receptors and ERBB2 negative cells, and (4) triple negative breast cancer with cells lacking molecular markers used for this classication.
The current therapeutic options and agents under development for the treatment of different types of breast cancer vary signicantly. The cancer cells overexpressing hormone receptors can be targeted by anti-estrogenic medicines, like tamoxifen, by aromatase inhibitors, like letrozole, or other medicines for endocrine therapy. To improve therapeutic outcome of the endocrine therapy, other agents with different mechanisms have been investigated: pan-class I phosphatidylinositol 3kinase (PI3K) inhibitors (e.g. alpelisib and buparlisib), 3,4 mammalian target of rapamycin (mTOR) inhibitors (e.g. everolimus), 5,6 and cyclin-dependent kinase CDK4 and CDK6 inhibitors (e.g. palbociclib and ribociclib). [7][8][9] For the treatment of ERBB2-positive breast cancer, PI3K and mTOR inhibitors are used together with ERBB2-targeted antibodies. Due to the absence of any targeted therapy for triple negative breast cancer, the general chemotherapy remains the main option available for the treatment of this most aggressive and mortal subtype of breast cancer. Typical medicines used against triple negative breast cancer include platinum drugs, taxanes, and anthracycline. 10 A group of promising emerging medicines, poly(ADPribose) polymerase inhibitors (e.g. olaparib and talazoparib), have been identied as a more specic therapy for a subgroup of triple negative breast cancer with cells having a mutation of BRCA1/BRCA2 genes. 11 New effective and selective anticancer agents are urgently needed for the safer and more effective treatment of triple negative breast cancer. The search for new potent compounds targeting breast cancer broadly covers various types of chemical structures. [12][13][14] 1,3,5-Triazine ring has been effectively used as a skeleton for the construction of new anticancer agents. 15 Recently, we identied 6,N 2 -diaryl-1,3,5-triazine-2,4-diamines selectively targeting triple negative MDA-MB231 breast cancer cells. 16 We also developed a 3D-QSAR model for the prediction of antiproliferative activity of this type of compounds against MDA-MB231 breast cancer cells. Herein, we are testing predictive power of this model for the design of new anticancer agents with the 6,N 2 -diaryl-1,3,5-triazine-2,4-diamine scaffold and continue our efforts on the development of highly potent and selective anticancer agents.

QSAR-guided design of compounds
We previously reported synthesis of 6,N 2 -substituted 1,3,5triazine-2,4-diamines (126 compounds) and their cytotoxic activity against breast cancer cell lines (MDA-MB231, SKBR-3 and MCF-7) and non-cancerous epithelial breast cells (MCF-10A). 16 Some of the prepared compounds demonstrated selective activity against triple negative breast cancer cells (MDA-MB231). Twenty-ve most active compounds were further evaluated and their GI 50 values were estimated and used for the development of a 3D-QSAR model suitable for the design of new potent anticancer agents. 16 The model is based on the activity of compounds with different substituents in phenyl rings A and B (Fig. 1).
The developed 3D-QSAR model indicated that bulky electron donating groups at the phenyl in the position 6 of the triazine, i.e. ring A would improve antiproliferative activity of compounds against MDA-MB231 cells. Based on this model, we designed a group of compounds bearing suitable functional groups at the phenyl rings A and B, with an expectation of higher activity against triple negative breast cancer, and applied the model to predict pGI 50 values for these compounds (Table  1).
To test the earlier developed model, two main groups of compounds were selected for the synthesis. Number of substituents in each of the phenyl rings A and B for the rst group of compounds (1-9) was limited to one functional group. The second group included 3,4,5-trimethoxyphenyl substituted compounds 10-21 to test effect of multiple substituents in ring A on the activity. Previously, we noticed that compounds with the R 1 group in meta-position of ring A retained activity with a greater variety of substituents at another phenyl ring. Contrary, activity of compounds with para-position of R 1 was very sensitive to the type and position of R 2 , disappearing when R 2 was located in the para-position of ring B. Selecting compounds 10-21 with the preferred methoxy groups located in positions equivalent to the para-and both meta-positions of ring A, we intended to test which activity pattern they will follow. The predicted pGI 50 values obtained from the 3D-QSAR model justied synthesis of the compounds.

Synthesis
Microwave irradiation has been widely used to facilitate synthesis of 1,3,5-triazines. 17 Sometimes, microwave irradiation  also changes outcome of reactions. The one-pot reaction of cyanoguanidine, benzaldehydes, and anilines in ethanol in the presence hydrochloric acid under conventional heating, followed by the treatment with aqueous sodium hydroxide (excess) was reported to produce 6,N 2 -diaryl-5,6-dihydro-1,3,5-triazine-2,4-diamines. 18 However, a similar reaction under focused microwave irradiation resulted in the formation their fully aromatic analogues. 19 This microwave-assisted methodology we applied for synthesis of new 6,N 2 -diaryl-1,3,5-triazine-2,4diamines , which were designed using the 3D-QSAR model as describe above. The reactions were performed in a one-pot manner with the three-component condensation of cyanoguanidine, benzaldehydes, and anilines at the rst stage and the rearrangement accompanied with dehydrogenative aromatization at the second one (Scheme 1). The structure of the resulting 6,N 2diaryl-1,3,5-triazine-2,4-diamines  was conrmed by the NMR spectroscopic data and X-ray crystallographic study on one representative product, compound 20. The three diagnostic signals of the aromatic triazine ring quaternary carbon atoms appear in 13 C NMR spectra of 1-21 in the region 164.4-170.2 ppm. In the 1 H NMR spectra, the downeld shi of signals for protons in the ortho-positions of the phenyl ring directly attached to the 1,3,5-triazine ring should be attributed to the anisotropic effect of the coplanar triazine p-electron system.
X-ray crystallography of 20 ( Fig. 2) showed that to a rst approximation the molecule is planar and has the shape of the letter U as both appended aromatic rings are orientated to the same side of the molecule. Within the triazine ring, the near equivalence of the C-N bond lengths is indicative of substantial delocalisation of p-electron density over the ring. Details of crystallographic analysis are available in ESI. †  one point concentration (10 mM) for preliminary assessment of their antiproliferative potential ( Table 2). Percentage cell viability was calculated 72 h aer treatment with compounds. In general, triple negative breast cancer cells (MDA-MB231) were more responsive than hormone positive breast cancer cells (SKBR-3 and MCF-7) to the treatment with the compounds. These results are similar to the trend observed earlier for their structural analogues. 16 Since all compounds 1-21 demonstrated signicant antiproliferative activity against MDA-MB231 cells at the screening concentration, they were further tested at concentration ranging from 0.00002 mM to 20 mM to estimate their 50% growth inhibitory concentrations (GI 50 ) against breast cancer cells (Table 3). Nilotinib and methotrexate were used as positive controls. For compounds active at the screening concentration against SKBR-3 and MCF-7 cells, concentration-dependent response was also evaluated and the corresponding GI 50 values were estimated.
Compounds 1-21 were also tested against MCF-10A normal breast cells to evaluate their selectivity towards cancer cells. None of the compounds showed signicant inhibition of the normal breast cell growth at the compound concentration of 20 mM.
The prepared 6,N 2 -diaryl-1,3,5-triazine-2,4-diamines 1-21 possessed specic cytotoxicity against triple negative MDA-MB231 breast cancer cells with GI 50 values ranging widely. However, the most intriguing results were obtained for compounds 10-21 with the 3,4,5-trimethoxyphenyl moiety as the ring A. This substitution was exceptionally benecial for the anticancer activity, particularly in a combination with the parasubstitution at the phenyl ring B. Changing location of the substituents to ortho-or meta-position in the phenylamino moiety dramatically decreased potency of compounds. The GI 50 values for these subgroups have a 2-3 order difference. For example, relocation of the methoxy group from the ortho-to para-position of the ring B resulted in a 200-fold increase in the antiproliferative activity (13 vs. 19). Even greater improvement in the activity was achieved when methyl or chloro substituents changed their location at the ring B from meta-to para-position leading to compounds 1300-2000-fold more potent than their regioisomers (14 vs. 16, 15 vs. 18). At the same time, it appeared that for the trimethoxyphenyl-substituted series (10-21) an increase in size of the R 2 group in para-position from the most potent compound with a methyl group (18) decreased the activity. Nevertheless, most of the triazines combining trimethoxyphenyl as the ring A and para-substituted phenylamino moieties as the ring B possessed activity comparable or higher than that of reference drugs methotrexate and nilotinib. These compounds also demonstrated good antiproliferative activity against SKBR-3 cells. The most active 6,N 2 -diaryl-1,3,5-triazine-2,4-diamine identied in the series was compound 18, which was 10-fold more active than methotrexate and 40-fold more potent than nilotinib against MDA-MB231 breast cancer cells. This compound (18) and its analogue 16, with the chloro substituent instead of the para-methyl group in the ring B, were selected for further experiments to better understand processes underlying antiproliferative effects of these compounds.
To assess predictive power of the earlier developed 3D-QSAR model, we compared experimental and predicted pGI 50 values, calculated using the 3D-QSAR model ( Table 4). The residual error values for the rst series of compounds (1-9) were rather acceptable viz. without extreme differences between the experimental and predicted values. However, a large discrepancy between the predicted and experimental values was observed for many trimethoxyphenyl-substituted compounds. These compounds, especially those with the R 2 group in para-position of the ring B (16)(17)(18)(19)(20)(21), appeared to be much more potent than it was predicted by the model. These ndings indicated a limitation of the earlier prepared 3D-QSAR model, 16 which seemed to be valid for compounds with monosubstituted phenyl rings and should be used with a caution for more complex structures.

Time-dependent cytotoxicity.
To further evaluate cytotoxicity of the prepared compound against cancer cells, timedependent cell viability experiments were carried out with the selected most active compounds 16 and 18 using MDA-MB231 breast cancer cell line. The MDA-MB231 cell viability was assessed aer the exposure of the cells to compounds 16 or 18 for 12, 24, 48, and 72 h at concentrations ranging from 0.2 nM to 125 nM. The GI 50 values were estimated when treatment with the highest concentration (125 nM) of tested compounds resulted in more than 80% of cell growth inhibition (Table 5).
Compound 18 possessed higher antiproliferative activity than 16 against MDA-MB231 cells for all duration of observations. For both compounds, the cytotoxic effect developed gradually and no signicant inhibition of the cell growth was    These results suggest that the antiproliferative effect of compounds 16 and 18 develop gradually and without an immediate toxic effect on the cells. A negligible cytotoxicity 12 h aer the treatment suggests that the compounds are less likely to cause cell necrosis and probably induce apoptosis. To further test this assumption, we performed uorescent microscopy experiments assessing effects of compounds 16 and 18 on the morphology of MDA-MB231 breast cancer cell.

Prediction of ADME properties
In the design of biologically active agents, optimization of lead compounds and selection of drug candidates, in silico evaluation of absorption, distribution, metabolism and elimination (ADME) of compounds has become a common practice. 21  QikProp (version 4.3) module of the Schrödinger soware was used to predict the molecular properties inuencing critical pharmacokinetic parameters of compounds 1-21 (Table 6). Parameters like octanol/water partition coefficient (QP log P, o/ w) and aqueous solubility (QP log S) are important for the prediction of drug absorption, transport and distribution in the body. These parameters calculated for 1-12 have values similar to those, which are typical for commonly used drugs. Steric and molecular surface descriptors i.e., total solvent accessible area (SASA) and its hydrophobic (FOSA) and hydrophilic (FISA) components were also calculated and found to be within the 95% range of values for known drugs. Lipinski's rule of ve has been oen used as a rst lter for the prediction the drug-like properties of compounds. 22 None of the prepared compounds violate Lipinski's rule of ve. The complete absorption and absence of effects on CNS were predicted for compounds 1-21.
Overall, all evaluated compounds were predicted to possess ADME properties favorable for potential agents targeting breast cancer cells. More detailed ADME prole for the compounds predicted by QikProp module is available in ESI. †

Conclusions
We synthesized a library of novel 6,N 2 -diaryl-1,3,5-triazine-2,4diamines designed using the 3D-QSAR data from the previous report. 16 Their antiproliferative activity was evaluated against three breast cancer cell lines and it was found that triple negative breast cancer cells (MDA-MB231) were signicantly more sensitive to the treatment with the prepared compounds. Some 6,N 2 -diaryl-1,3,5-triazine-2,4-diamines demonstrated good antiproliferative activity against SKBR-3 cells, but MCF-7 cells were generally resistant to the treatments with these compounds. Some of the synthesized compounds demonstrated even greater activity against MDA-MB231 cells than it was predicted by the 3D-QSAR model. The 3D-QSAR model limitation might originate from multiple targets responsible for the activity of 6,N 2 -diaryl-1,3,5-triazine-2,4-diamines and hence requires further investigations. The discrepancy between the predicted values and the experimental data was particularly evident for  Paper suggested that the tested compounds induced apoptosis in MDA-MB231 cells. All compounds, including 18, were predicted to have ADME proles favorable for potential antiproliferative agents targeting breast cancer.

General
Melting points (uncorrected) were determined using a Stuart™ SMP40 automatic melting point apparatus. 1 H and 13 C NMR spectra were recorded on a Bruker Fourier NMR spectrometer (300 MHz) using DMSO-d 6 as a solvent and TMS as an internal reference. Microwave-assisted reactions were carried out in the closed vessel focused single mode using a Discover SP microwave synthesizer (CEM, USA) monitoring reaction temperature by the equipped IR sensor.

4.2.
General method for the synthesis of 6,N 2 -diaryl-1,3,5triazine-2,4-diamines (1-21) The microwave irradiation parameters optimized earlier 19 for the synthesis of 6,N 2 -diaryl-1,3,5-triazine-2,4-diamines were applied for the preparation of 1-21. To a solution of cyanoguanidine (0.21 g, 2.5 mmol), a substituted benzaldehyde (2.5 mmol), and an aniline (2.5 mmol) in EtOH (2 mL) in a 10 mL seamless pressure vial, conc. HCl (0.21 mL, 2.5 mmol) was added. The reaction mixture was heated at 140 C for 50 min by irradiation in the Discover SP (CEM) microwave reactor operating at maximal microwave power up to 150 W. Then, an aq. solution of NaOH (5 N, 1 mL) was added to the reaction mixture and heating was continued for another 15 min at 140 C. Aer cooling, the precipitated product was ltered, washed with water and recrystallized from suitable solvents (EtOH, aq. EtOH, or MeCN) specied below. Yields of products 1-21 are reported as overall isolated yields for the one-pot procedure. Intensity data for a colourless crystal of 20 (0.05 Â 0.09 Â 0.13 mm) were measured at 100 K on an XtaLAB Synergy Dual Atlas diffractometer equipped with a CCD area detector and graphitemonochromated Cu Ka radiation (l ¼ 1.54184Å) so that q max ¼ 67.1 . Data reduction and empirical absorption corrections, based on a multi-scan technique, were applied. 23 The structure was solved by direct methods 24 and rened on F 2 with anisotropic displacement parameters and C-bound H atoms in the riding model approximation. 25 The nitrogen-bound H atoms were rened with a distance restraint N- The nal renement on 292 parameters yielded R ¼ 0.059 (3044 data with I $ 2s(I)) and wR 2 ¼ 0.176 (all 3357 data). The maximum and minimum residual electron density peaks of 1.73 and 0.62 eÅ À3 , respectively, were located 1.85 and 0.69Å from the C19 and F1 atoms, respectively, that is, in chemically non-sensible positions. The molecular structure diagram was generated at the 70% probability level by ORTEP for Windows, 26 and the packing diagrams were generated with DIAMOND. 27 Additional analysis was conducted with PLATON. 28 Crystal data for C 19 (2) , g ¼ 108.517(2) , V ¼ 941.57(4)Å 3 , Z ¼ 2, D x ¼ 1.543 g cm À3 , F(000) ¼ 452 and m ¼ 1.125 mm À1 .

In vitro cytotoxicity assay
The synthesized compounds were tested against three breast tumor cell lines (MDA-MB231, SKBR-3 and MCF-7) and epithelial breast cell line (MCF-10A) by the MTT colorimetric assay. 29,30 All cells were obtained from the American Type Culture Collection. The cancerous cell lines were grown in Dulbecco's modied eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% pen-strep antibiotic. The MCF-10A human epithelial breast cell line was grown in the complete mammary epithelial growth medium containing horse serum 5%, epithelial growth factor 20 ng mL À1 , hydrocortisone 0.5 mg mL À1 , cholera toxin 100 ng mL À1 , insulin 10 mg mL À1 , and pen-step antibiotic. 31 For the cytotoxic assay, 20 to 75 Â 10 3 cells per mL (based on the doubling time for each cell line) were seeded in 96-well plates and the plates were incubated overnight in a humidied air atmosphere at 37 C in 5% CO 2 incubator.
The cells were then treated with compounds at different concentrations. Aer 72 h of incubation, the MTT (0.5 mg mL À1 ) was added to wells, followed by 4 h of incubation. The culture medium was then removed and DMSO (100 mL per well) was added and the absorbance values were measured at 570 nm using the multi-well Tecan NanoQuant, Innite m200 Pro plate reader. Growth inhibitory values (GI 50 ) were calculated using GraphPad Prism 7 (GraphPad Soware, San Diego, USA) by nonlinear regression analysis. Three independent experiments were carried out and the data was expressed in mean AE standard deviation (SD). The concentration-response curves used for the GI 50 calculation are available in ESI. †

Time-dependent cytotoxicity
MDA-MB231 breast cancer cells were grown in DMEM supplemented with 10% FBS and 1% pen-strep antibiotic. The cells were seeded in 96-well plates (20 Â 10 3 cells per mL) and the plates were incubated overnight in a humidied air atmosphere at 37 C in 5% CO 2 incubator. The cells were then treated with compounds 16 or 18 in concentrations ranging from 0.2 nM to 125 nM for 12, 24, 48, and 72 h. Aer specied time for each experiment, MTT (0.5 mg mL À1 ) was added to wells, followed by 4 h of incubation. The culture medium was then removed and DMSO (100 mL per well) was added and the absorbance values were measured at 570 nm using the multi-well Tecan Nano-Quant, Innite m200 Pro plate reader. GI 50 values were calculated using GraphPad Prism 7 (GraphPad Soware, San Diego, USA) by nonlinear regression analysis. All the time-dependent experiments at different times were carried out using the same passage of MDA-MB231 cells. Three independent experiments were carried out and the data were expressed in mean AE standard deviation (SD).

Acridine orange/propidium iodide staining
Fluorescence microscopy was used to visualize the apoptosis in cancer cells with AO/PI staining. 20 MDA-MB231 cells were grown in 6-well plate (5 Â 10 5 cells per well). The cells were treated with respective compounds at concentrations equal to estimated GI 20 and GI 50 values and incubated for 24 h. One of the six wells was treated with 1% DMSO and served as a negative control, another well treated with methotrexate served as a positive control. Aer 24 h of incubation, the wells were washed with phosphate buffered saline (PBS) three times and 100 mL of AO (100 mg mL À1 in PBS) and 25 mL of PI (100 mg mL À1 in PBS) in 1 mL of media were added to each well. The plate was then observed under a motorized inverted uorescent microscope (Eclipse Ti2-E, Nikon). Three independent experiments at each concentration were carried out.

Live cell imaging
MDA-MB231 cells (1 Â 10 6 cells) were seeded in a Petri dish and incubated overnight in a humidied air atmosphere at 37 C in 5% CO 2 incubator. Then, the Petri dish was washed with PBS three times and 300 mL of AO (100 mg mL À1 in PBS) and 75 mL of PI (100 mg mL À1 in PBS) in 3 mL of media were added to the Petri dish. The cells were treated with 10 mM of compound 18 and the dish was then transferred to motorized inverted uorescent microscope (Eclipse Ti2-E, Nikon). The pictures were taken every 10 minutes for 4 h and intercalated into video (see ESI †).

QSAR model testing
The previously reported 16 3D-QSAR model for 6,N 2 -1,3,5triazine-2,4-diamines against MDA-MB231 breast carcinoma was applied. The structures of the proposed compounds were drawn using ChemDraw 15.0 and imported to Discovery Studio v18 (ref. 32) for the activity prediction. The structures were prepared for the 3D-QSAR modeling and aligned to minimum energy using the 'align small molecules' protocol, which is based on 50% steric and 50% electrostatic elds for alignment of molecules. The predicted pGI 50 values were then obtained by the 'calculate molecular properties' protocol in Discover Studio using the previously prepared 3D-QSAR model.

ADME properties prediction
QikProp module of the Schrodinger 33 was used to predict the absorption, distribution, metabolism and excretion (ADME) properties. The QikProp module predicts the descriptors which are pharmaceutically signicant to identify the relevant properties of the organic molecule in relation to the 95% of the marketed drugs. The molecules were drawn and prepared (energy minimized and aligned) in Maestro program (v10.1) of Schrodinger soware suit. QikProp (v4.3) was run with default options in normal processing mode.

Conflicts of interest
There are no conicts to declare.