Synthesis, structure confirmation, identification of in vitro antiproliferative activities and correlation of determined lipophilicity parameters with in silico bioactivity descriptors of two novel classes of fused azaisocytosine-like congeners

2.1 Instrumentations and general materials

Commercially available chemicals and solvents were employed in the synthesis of the title compounds. Chemicals (2-oxobutanoic acid, ethyl 2-oxo-4-phenylbutyrate, triethylamine) were purchased from Fluka (Germany) as the highest grade available and used without further purification. Solvents (n-butanol, N,N-dimethylformamide) were purchased from Merck (Germany). Two polar solvents, such as the distilled water (Merck, Millipore, Germany) and methanol (Roth, Germany), were employed with the goal of removing an unwanted by-product – triethylammonium iodide, which was soluble in these solvents. The used solvents were evaporated from the reaction mixture to a suitable volume under reduced pressure by using a Heidolph rotary evaporator (Germany). Melting points (m.p.) were determined employing a Boetius melting point apparatus (Germany) and are uncorrected.

The purity as well as homogeneity of each analytical sample was tentatively established by employing thin-layer chromatography (TLC). A TLC analysis was routinely carried out on commercial SiO₂ 60 F₂₅₄ Merck aluminium sheets (Germany), possessing a fluorescence indicator. The TLC spots were visualised by their absorption under ultraviolet light at an excitation wavelength of 254 nm. NMR spectra that characterise two novel classes of the original compounds (7–12 and 13–17) were recorded on a Bruker Avance spectrometer (Germany), using the deuterated dimethyl sulfoxide ((CD₃)₂SO, DMSO-d₆, Merck, Germany) as a reference solvent (revealing the “extra peaks” at δ = 2.5 ppm and 39.5 ppm in ¹H NMR and ¹³C NMR spectra, respectively). ¹H NMR and ¹³C NMR spectra were run at 300 MHz and 75 MHz, respectively, and reported in ppm referenced to tetramethylsilane (Me₄Si, TMS, Sigma-Aldrich, Germany) as an external standard (δ = 0 ppm). Coupling constants (J) are reported in units of frequency, e.g., hertz (Hz) with up to one digit after the decimal. Abbreviations for signal multiplicities are indicated as follows: s, singlet; d, doublet; t, triplet; q, quartet and m, multiplet. The resulting combustion microanalyses of C, H, Cl, N were agreeable within ±0.4% with theoretical values. The retention time (t_R) of each molecule was measured on the column with an immobilised artificial membrane (IAM) at 293 K (20 °C); a mobile phase containing 50% acetonitrile (MeCN) in the buffer adjusted to pH 7.4 was used.

3.1 Synthesis of two novel classes of fused azaisocytosine-like congeners (7–12 and 13–17)

The general synthetic strategy (shown in Scheme 1) was employed in preparing two novel classes of fused azaisocytosine-like congeners such as 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (7–12) and 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (13–17). All these original and closely related compounds, comprising the incorporated 6-azaisocytosine scaffold, were generated by the successive heteroannelation according to the [4 + 2] pattern (Rusinov et al., 2008).

Close

Scheme 1 Approaches for the synthesis of two novel classes of compounds based on the 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one template (7–12 and 13–17, respectively), utilising nucleophilic and electrophilic centred substrates.

Export to PPT

The optionally substituted 1-aryl-2-hydrazinylideneimidazolidine hydroiodides (1–6) (Sztanke et al., 2006a,b; Sztanke, 2012) were employed as the excellent nucleophilic centred building blocks in the synthesis of novel fused azaisocytosine-containing congeners. The synthesis of these functionalised precursors was achieved in good yields under mild reaction conditions starting from 1-aryl-2-methylthioimidazolines. The methylsulfanyl functionality in outgoing molecules was displaced by strongly nucleophilic deprotecting reagent – hydrazine hydrate on heating in an inert methanolic medium, using a patent pending methodology, as described earlier (Sztanke, 2012). Some 1-aryl-2-hydrazonoimidazolidine hydroiodides have previously been disclosed as potent antibacterial agents (Sztanke, 2012).

Novel 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (7–12) (Sztanke and Sztanke, 2014a) were obtained in a well-tried efficient and scalable synthetic procedure in good overall yields under mild reaction conditions. It has been established experimentally that the best conditions for effective cyclisation of 7–12 were to treat the 1-aryl-2-hydrazinylideneimidazolidine hydroiodides with 2-oxobutanoic acid in refluxing n-butanol containing a small molar excess of triethylamine. In turn, the synthesis of novel 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (13–17) (Sztanke and Sztanke, 2014b) was achieved from 1-aryl-2-hydrazonoimidazolidine hydroiodides upon condensation with ethyl 2-oxo-4-phenylbutyrate in the presence of triethylamine by employing the analogous synthetic route (in which the suitable 2-oxoester was employed instead of 2-oxoacid) to that used for preparation of 7–12.

The significance of the elaborated relatively straightforward chemical procedures is that they could be used to synthesise closely related novel classes of compounds (e.g., the original congeners of fused azaisocytosine related in structure to naturally occurring nucleobases) under mild reaction conditions in good overall yields. It has been well-established experimentally that a small molar excess of Et₃N during heterocyclisation process resulted in higher yields of the isolated fused azaisocytosine-like products.

To explain the course of formation of novel 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (7–12) as well as 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (13–17) the most likely two-step mechanism (Scheme 1) was proposed. In the first step the nucleophilic centred 1-aryl-2-hydrazinylidenemidazolidine hydroiodides (1–6) were reacted with the electrophilic centred 2-oxobutanoic acid or ethyl 2-oxo-4-phenylbutyrate to generate, initially, the ring-opened ketimine intermediates, e.g., 2-[(1-arylimidazolidin-2-ylidene)hydrazinylidene]butanoic acids or ethyl 4-phenyl-2-[(1-arylimidazolidin-2-ylidene)hydrazinylidene]butanoates, respectively. This reaction step was carried out in polar solvents such as n-butanol containing a small molar excess of triethylamine and proceeded in both cases with a concomitant loss of water molecule. Simultaneously, a by-product of alkylation such as Et₃NH⁺I⁻ was formed. In the second step, the ring-opened ketimine-type intermediates (because of the presence of protonated endocyclic N-3 nitrogen) in boiling polar solvents (such as n-butanol) as reaction media under basic conditions readily underwent intramolecular heterocyclisation to yield novel fused azaisocytosine-containing congeners of the 7–12 and 13–17 type with a concomitant removal of water or ethanol molecule, respectively.

Results of ¹H NMR and ¹³C NMR spectroscopic investigations gave the definite proof that under well-established reaction conditions the subsequent heterocyclisation of the ring-opened intermediates leads to the stable homogeneous 6:5 bicyclic products (7–12 and 13–17) which are thermodynamically preferred.

According to the best of our knowledge reactions of 1-aryl-2-hydrazonoimidazolidines with 2-oxobutanoic acid or ethyl 2-oxo-4-phenylbutyrate do not appear in the scientific literature, including Chemical Abstracts. Thus, these reactions represent novel synthetic pathways (which were not investigated in the past) yielding two closely related types of unknown fused azaisocytosine-like congeners (7–12 and 13–17).

3.2 Spectroscopic characterisation of two novel classes of fused azaisocytosine-like congeners (7–12 and 13–17)

Structures of the original 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (7–12) were confirmed by employing spectroscopic experiments. This class of small molecules revealed in their ¹H NMR spectra some special spectroscopic features. The signals of the H-7 and H-6 protons (belonging to both endocyclic methylene groups) were registered as a singlet at δ 4.14 ppm (in the case of compounds 7, 10 and 12) and as a multiplet in the region 3.98–4.27 ppm (in the case of derivatives 8, 9 and 11), integrating for four protons and four protons, respectively. A triplet at δ in the region 1.14–1.18 ppm with a coupling constant of J = 7.4 Hz (and integrating for three protons) confirmed the presence of three methyl protons (CH₃CH₂) in the case of all the synthesised molecules of this class (7–12). In turn, a quartet at δ in the region 2.62–2.67 ppm with a coupling constant of J = 7.4 Hz (and integrating for two protons) affirmed the attendance of two methylene protons (CH₃CH₂) in the case of 7–12. Additionally, the resonances at δ values ca. 153.2, 151.5 and 152.1 ppm were assigned to quaternary carbon atoms (e.g., C-4, C-3 and C-8a, respectively) within the 6:5 heterobicyclic template.

Structures of 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (13–17) were determined by using spectroscopic methods. All the derivatives of this class which were synthesised (13–17) revealed their ¹H NMR spectra signals of the H-7 and H-6 protons. These signals were registered as a singlet at δ ca. 4.14 ppm (in the case of compounds 13, 16 and 17), as a triplet at δ 4.13 ppm with a coupling constant of J = 2.7 Hz (in the case of molecule 14) and as a multiplet in the region 3.99–4.28 ppm (in the case of solute 15), integrating for four protons. Additional signals derived from four protons of the exocyclic ethylene moiety joined at the C-3 of the scaffold were registered as a singlet at δ ca. 2.95 ppm, integrating for four protons (in the case of all the compounds of this class). In turn, the carbon resonances that were assigned to the C-4, C-3 and C-8a atoms (within the 6:5 heterobicyclic scaffold) were shifted at δ values ca. 152.3, 151.2 and 151.8 ppm, respectively.

The registered chemical shift values of the C-6 and C-7 atoms in the ¹³C NMR spectra of two closely related types of compounds (7–12 and 13–17) gave a valuable information for deducing an unequal character of both endocyclic methylene groups (in the imidazolidine ring). This would be expected additionally on the basis of our previous structural investigations on fused azaisocytosine-like congeners (Sztanke et al., 2011). The δ values were shifted in two regions: 40.0–41.3 ppm for the C-6 and 44.7–47.1 ppm for the C-7. The assignment of the C-6 as well as C-7 atoms was done after analysing DEPT-135 spectra of all the analytical samples in which the negative signals of these secondary methylene carbons were markedly seen.

Moreover, it was found that the ortho-chloro substitution at the phenyl moiety in both novel classes of fused azaisocytosine-containing congeners (see the ¹³C NMR spectra of compounds 9 and 15) markedly shifts the δ values of the C-6 and C-7 signals to highest field.

3.3 Anticancer activities of novel fused azaisocytosine-like congeners (7–12 and 13–17)

All novel fused azaisocytosine-containing congeners (7–12 and 13–17) at a concentration of 50 μg mL⁻¹ were highly effective at inhibiting proliferation in four recruited neoplastic cells (A549, HeLa, T47D and TOV112D) after 24-, 48- and 72-h incubation periods (Tables 1 and 2). It is worth noting that all the screened molecules revealed higher antiproliferative effects in A549, HeLa and T47D cells than the known folate antimetabolite – pemetrexed at a concentration of 100 μg mL⁻¹ (Sztanke et al., 2019). The investigated compounds may inhibit proliferation of cancer cells due to their competition for nucleotides required for DNA biosynthesis (Rusinov et al., 2008).

Table 1 Cytotoxicity expressed as the percentage growth inhibition of non-tumoural and tumoural cells by 50 μg mL⁻¹ concentration of the examined small molecular weight compounds.

Compound	R	Incubation time (h)	Growth inhibition (%) in cell lines
			Vero	A549	HeLa	T47D	TOV112D
7	H	24	50 ± 4.2	80 ± 7.5	45 ± 3.2	60 ± 5.5	85 ± 7.9
		48	60 ± 5.5	90 ± 8.2	80 ± 7.1	85 ± 7.9	100 ± 8.9
		72	90 ± 8.6	100 ± 9.1	90 ± 6.8	90 ± 8.2	100 ± 10.1
8	4-CH3	24	50 ± 3.5	50 ± 6.5	50 ± 3.5	70 ± 6.5	75 ± 6.9
		48	60 ± 5.1	60 ± 5.5	80 ± 6.5	70 ± 5.8	85 ± 8.1
		72	75 ± 6.6	70 ± 6.1	95 ± 7.7	75 ± 6.7	100 ± 9.3
9	2-Cl	24	25 ± 1.7	25 ± 1.5	50 ± 4.6	80 ± 8.2	80 ± 7.9
		48	60 ± 5.5	60 ± 5.5	90 ± 7.9	85 ± 7.9	90 ± 8.7
		72	80 ± 7.2	80 ± 7.2	100 ± 6.5	90 ± 8.5	100 ± 9.6
10	3-Cl	24	10 ± 0.9	75 ± 6.9	60 ± 4.7	80 ± 7.2	80 ± 7.5
		48	50 ± 3.7	80 ± 7.2	80 ± 5.9	85 ± 7.5	90 ± 8.8
		72	65 ± 5.9	90 ± 8.4	90 ± 7.9	95 ± 8.6	100 ± 9.7
11	4-Cl	24	50 ± 3.7	50 ± 4.1	75 ± 5.8	75 ± 7.1	80 ± 7.9
		48	60 ± 5.2	60 ± 5.2	85 ± 7.5	90 ± 8.5	90 ± 7.8
		72	80 ± 7.3	90 ± 8.1	95 ± 6.7	100 ± 9.6	95 ± 9.2
12	3,4-Cl2	24	5 ± 0.6	50 ± 3.9	60 ± 4.5	80 ± 7.1	65 ± 5.9
		48	45 ± 2.6	60 ± 4.9	75 ± 6.5	85 ± 7.9	90 ± 8.2
		72	65 ± 5.8	90 ± 7.8	100 ± 7.2	95 ± 8.2	95 ± 9.1
Pemetrexed (Sztanke et al., 2019)		24	5 ± 0.2	10 ± 0.9	5 ± 0.7	10 ± 0.8	nd
		48	10 ± 1.1	30 ± 2.4	10 ± 0.9	20 ± 1.2	nd
		72	20 ± 1.9	50 ± 2.8	20 ± 1.5	30 ± 2.2	nd

Non-tumoural cell line: Vero – (ECACC 88020401) – African Green Monkey kidney cells.

Cancer cell lines: A549 (ECACC 86012804) – human Caucasian lung carcinoma cells; HeLa (ECACC 93021013) – human Negroid cervix epithelioid carcinoma cells; T47D (ECACC 85102201) – human breast carcinoma cells; TOV112D (ATCC CRL-11731) – human ovarian primary malignant adenocarcinoma cells.

nd – not determined since has not been active in the preliminary screening.

A concentration of 50 μg mL⁻¹ corresponds to following micromolar concentration values: 0.206 mM (7), 0.195 mM (8), 0.181 mM (9–11), 0.161 mM (12).

Pemetrexed was used as a positive control at a concentration of 100 μg mL⁻¹, which corresponds to 0.234 mM.

Table 2 Cytotoxicity expressed as the percentage growth inhibition of non-tumoural and tumoural cells by 50 μg mL⁻¹ concentration of the examined small molecular weight compounds.

Compound	R	Incubation time (h)	Growth inhibition (%) in cell lines
			Vero	A549	HeLa	T47D	TOV112D
13	H	24	50 ± 5.2	30 ± 3.5	90 ± 9.4	90 ± 8.9	75 ± 6.8
		48	60 ± 5.8	75 ± 8.1	100 ± 10.1	100 ± 9.7	100 ± 9.7
		72	100 ± 11.3	100 ± 10.2	100 ± 9.9	100 ± 10.1	100 ± 8.9
14	4-CH3	24	50 ± 4.9	50 ± 5.3	90 ± 8.9	90 ± 9.3	80 ± 6.6
		48	60 ± 6.3	80 ± 7.9	100 ± 10.2	100 ± 9.9	100 ± 8.8
		72	95 ± 10.0	85 ± 9.1	100 ± 9.8	100 ± 10.1	100 ± 9.5
15	2-Cl	24	10 ± 1.9	10 ± 0.9	75 ± 8.1	80 ± 7.9	50 ± 4.8
		48	60 ± 6.1	50 ± 4.7	80 ± 7.9	90 ± 9.2	75 ± 8.1
		72	80 ± 7.8	60 ± 6.1	90 ± 10.1	100 ± 10.2	90 ± 8.9
16	4-Cl	24	20 ± 1.9	80 ± 8.5	65 ± 7.2	80 ± 7.5	90 ± 9.1
		48	70 ± 7.3	90 ± 8.9	80 ± 7.9	100 ± 9.8	100 ± 8.9
		72	90 ± 9.8	100 ± 9.7	90 ± 9.1	100 ± 10.2	100 ± 10.1
17	3,4-Cl2	24	90 ± 9.1	90 ± 7.7	90 ± 9.1	90 ± 8.5	90 ± 8.5
		48	100 ± 8.5	100 ± 9.2	100 ± 8.6	100 ± 10.1	100 ± 10.1
		72	100 ± 9.5	100 ± 9.5	100 ± 9.8	100 ± 9.7	100 ± 8.9
Pemetrexed (Sztanke et al., 2019)		24	5 ± 0.2	10 ± 0.9	5 ± 0.7	10 ± 0.9	nd
		48	10 ± 1.1	30 ± 2.4	10 ± 0.9	20 ± 1.2	nd
		72	20 ± 1.9	50 ± 2.8	20 ± 1.5	30 ± 2.2	nd

Non-tumoural cell line: Vero – (ECACC 88020401) – African Green Monkey kidney cells.

Cancer cell lines: A549 (ECACC 86012804) – human Caucasian lung carcinoma cells; HeLa (ECACC 93021013) – human Negroid cervix epithelioid carcinoma cells; T47D (ECACC 85102201) – human breast carcinoma cells; TOV112D (ATCC CRL-11731) – human ovarian primary malignant adenocarcinoma cells.

nd – not determined since has not been active in the preliminary screening.

An effective concentration of 50 μg mL⁻¹ corresponds to following micromolar concentration values: 0.157 mM (13), 0.150 mM (14), 0.142 mM (15, 16), 0.129 mM (17).

Pemetrexed was employed as a positive control at a concentration of 100 μg mL⁻¹, which corresponds to 0.234 mM.

Amongst 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones the parent compound (7), bearing an unsubstituted and electron-rich phenyl ring, revealed the highest cytotoxic potencies against A549 and TOV112D cells.

Placing an electron-donating methyl group para to the phenyl ring (structure 8) led to a slight increase in cytotoxicity against HeLa (after 24 and 72 h of incubation) and T47D (after 24 h) cells, while reduced the antiproliferative potency towards the majority of neoplastic cells: A549 (after 24–72 h), T47D (after 48 and 72 h) and TOV112D (after 24 and 48 h). In adding, this substitution pattern resulted in a decrease in cytotoxicity against non-tumoural cells after 72-h incubation period, when compared to 7.

The replacement of the methyl group that is placed at the para position with a chloro substituent led to the bioisostere 11, revealing a significant cytotoxicity against tumour cells. The para-chloro substituent located on the phenyl ring proved to be better than the para-methyl group for the antiproliferative potency against A549 (after 72 h of incubation), HeLa (after 24 and 48 h) and T47D (after 24, 48 and 72 h) cells. However, the para-methyl substitution was responsible for a slight decrease in cytotoxicity towards non-tumoural cells after 72-h incubation period.

Introducing an electron-withdrawing chloro group ortho to the phenyl ring, interestingly resulted in still antiproliferative active compound (9), showing an improved activity against T47D (after 24 h of incubation) and HeLa (after 24, 48 and 72 h) cells and the decreased cytotoxicity against A549 cells (after 24, 48 and 72 h). This modification was responsible for the improved selectivity for tumoural (T47D, TOV112D and HeLa) over non-tumoural cells after 24-h incubation period, when compared to 7.

Moving the chloro substituent from the ortho to the meta position of the phenyl ring produced the evidently potent structure (10) against T47D, TOV112D and HeLa cells and led to an increase in cytotoxicity for A549 cells, in comparison with 9. Fortunately, this substitution pattern was responsible for an increase in selectivity towards tumoural over non-tumoural cells, comparatively to 9, demonstrating a decrease in cytotoxicity towards Vero cells.

Introducing the second electron-withdrawing chloro group to the 3-Cl derivative (10) resulted in the 3,4-dichloro-substituted structure (12), revealing the retained potency against HeLa cells, but the decreased cytotoxicity against A549 (after 24 and 48 h of incubation) and TOV112D (after 24 and 72 h) cells. This structural modification was responsible for a twofold decrease in cytotoxicity towards non-tumoural cells after 24-h incubation period and a slight cytotoxicity decrease after 48 h of incubation (when compared to 10). It should be noted that the compound 12 proved to be the least toxic for non-tumoural cells amongst all the compounds of this class.

Amongst 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (13–17) the highest cytotoxic potencies against all the recruited tumour cell lines were determined for the structure 17, bearing two electron-withdrawing chlorine atoms meta and para to the phenyl ring. This substitution pattern was identified as being beneficial for the increased cytotoxicity.

Two newly synthesised compounds: 13 (bearing an unsubstituted and electron-rich phenyl ring) and 14 (having an electron-donating CH₃ group at the para position of the phenyl ring) proved to be evidently effective against HeLa and T47D cells. In addition, these congeneric molecules evoked remarkable antiproliferative effects in TOV112D cells and demonstrated incubation time-dependent cytotoxic effects in A549 cells. In turn, the designed structure 16 (bearing an electron-withdrawing chloro group para to the phenyl ring) displayed the strongest antiproliferative activities against TOV112D, T47D and A549 cells. However, the modified compound 15 (with an electron-withdrawing chloro group placed at ortho position of the phenyl ring) proved to be the most cytotoxic against T47D and HeLa cells.

Placing an electron-donating methyl group para to the phenyl ring (structure 14) did not alter the strong antiproliferative potency against HeLa and T47D cells but evoked a slight increase in cytotoxicity against TOV112D cells after 24 h of incubation, when compared to the parent compound (13). Additionally, this modification resulted in an increase in cytotoxicity against A549 cells after 24- and 48-h incubation periods, and simultaneously in a decrease in cytotoxicity against A549 cells after 72 h of incubation, comparatively to 13.

The replacement of the methyl group that is placed in the para position of the phenyl ring (structure 14) with a chloro substituent was tolerated since this modification allows to obtain the structural bioisostere 16, revealing the remarkable cytotoxicity against cancer cells. However, the para-chloro substitution on the phenyl ring was found to be better than the para-methyl group for the antiproliferative potency against A549 cells and fortunately evoked 2.5-fold decrease in cytotoxicity towards non-tumoural cells after 24 h of incubation. In turn, the para-methyl substituent attached to the phenyl ring was responsible for the observed increase in cytotoxicity towards HeLa cells.

Moving the chloro substituent from the para to the ortho position of the phenyl ring was tolerated since it resulted in the highly antiproliferative active structure (15). In adding, this modification led to a slight increase in cytotoxicity against HeLa cells after 24 h of incubation, and evoked decreases in cytotoxicity towards A549 and TOV112D cells after 24–72-h incubation periods and against T47D cells only after 48 h of incubation, comparatively to the compound 16. Fortunately, this modification was responsible for a twofold decrease in cytotoxicity towards non-tumoural cells after 24-h incubation period, when compared to 16. Primarily, it was confirmed that the chloro group had to be introduced in the ortho position of the phenyl ring in order to retain the more selective cytotoxicity for HeLa, T47D and TOV112D over non-tumoural cells. Secondly, it was shown that the chloro substituent had to be retained in the para position of the phenyl ring in order to obtain the more selective cytotoxic activity for neoplastic A549 cells over non-tumoural cells.

Structure–activity relationship (SAR) studies around the unsubstituted and an optionally substituted phenyl moiety in these two novel classes of fused azaisocytosine-containing congeners demonstrated significant antitumour activities for all the compounds (7–17). However, substitutions with chlorine atoms in meta and meta-para (in the case of 8-aryl-3-ethyl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones) as well as in ortho and para (in the case of 8-aryl-3-(2-phenylethyl)-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones) positions of the phenyl moiety were found to be the best choices of substitution patterns in these classes of compounds. Thus, a great attention should be focused on four highly antiproliferative potent lead structures (10, 12, 15 and 16), disclosing the explicitly marked lower cytotoxicity towards non-tumoural over cancer cells (both of the same epithelial origin) after 24-h incubation period. Therefore, these ones might be utilised in further modern designing novel less toxic to use anticancer agents.

3.4 IC₅₀ values and selectivity indices of novel fused azaisocytosine-like congeners (7–17)

Table 3 (comparing the IC₅₀ values of all the original small molecules (7–17)) as well as Table 4 (reporting their selectivity indices) is attached. The IC₅₀ values were calculated from concentration-dependent inhibition curves, whereas the selectivity indices (SI) were calculated as follows:

Table 4 Selectivity indices (SI) for all novel fused azaisocytosine-like congeners.

R	Incubation time (h)	SI
		A549		HeLa		T47D		TOV112D
		R′	R″	R′	R″	R′	R″	R′	R″
H	24	1.62	<1.00	<1.00	1.78	1.24	1.78	1.75	1.60
	48	1.55	1.30	1.31	1.63	1.42	1.63	1.70	1.63
	72	1.10	1.00	1.00	1.00	1.00	1.00	1.10	1.00
4-CH3	24	1.00	1.00	1.00	1.88	1.43	1.88	1.54	1.67
	48	1.00	1.44	1.33	1.63	1.14	1.63	1.45	1.63
	72	0.93	0.89	1.30	1.00	1.00	1.00	1.30	1.00
2-Cl	24	>1.00	>1.00	>1.00	>1.56	>1.64	>1.56	>1.64	>1.00
	48	1.00	0.86	1.50	1.33	1.36	1.50	1.50	1.33
	72	1.00	0.75	1.22	1.13	1.10	1.29	1.22	1.13
3-Cl	24	>1.50	–	>1.20	–	>1.64	–	>1.64	–
	48	1.64	–	1.64	–	1.64	–	1.80	–
	72	1.40	–	1.40	–	1.40	–	1.56	–
4-Cl	24	1.00	>1.56	1.50	>1.27	1.50	>1.56	1.64	>1.75
	48	1.00	1.25	1.36	1.11	1.50	1.43	1.50	1.43
	72	1.10	1.14	1.10	1.00	1.22	1.14	1.10	1.14
3,4-Cl2	24	>1.00	1.00	>1.23	1.00	>1.60	1.00	>1.33	1.00
	48	>1.23	1.00	>1.45	1.00	>1.78	1.00	>1.78	1.00
	72	1.33	1.00	1.50	1.00	1.50	1.00	1.50	1.00
Pemetrexed	24	1.00		1.00		1.00		nd
	48	1.00		1.00		1.00		nd
	72	1.00		>1.00		1.00		nd

SI – selectivity indices are expressed as ICs₅₀ of compounds in non-tumoural Vero cells/ICs₅₀ of compounds in cancer cells.

selectivity index (SI) = IC₅₀ of the tested compound in a non-tumoural Vero cell line/IC₅₀ of the tested compound in a cancer cell line.

The IC₅₀ values in epithelial cancer cell lines for all novel fused azaisocytosine-containing congeners (7–17) were found to be lower than these determined for pemetrexed.

Some of innovative structures demonstrated also better selectivity indices than pemetrexed in A549 (7, 10, 12, 16), HeLa (9, 10, 11, 12, 15) and T47D (9, 10, 11, 12, 15, 16) cell lines after all incubation periods. Simultaneously, the majority of SI values for closely related molecules indicated that the safety window between healthy and cancer cell lines is not big. Notwithstanding it should be noted that the selectivity index for 4-hydroxytamoxifen (the clinically approved anticancer agent in an antiestrogen treatment for a human breast cancer) was found to be 1.29 (Badisa et al., 2009). The better selectivity revealed the designed azaisocytosine-like congeners which have selectivity indices near to ca. 2. They were almost twice lower cytotoxic to non-tumoural as compared to malignant cells. This allows to assume a lower damage to the epithelium of the gastrointestinal tract, justifying their further in vivo testing. This finding seems to be significant in the field of the preliminary antitumour search.

3.5 The retention factors of the investigated compounds

Lipophilicity parameters of the investigated solutes (7–17) are expressed as their isocratic retention factors, logs k, (Yamagami et al., 1990, 1994; Klein et al., 1988) that were measured directly for a given RP-HPLC system. These ones were established on two LC columns employing such stationary phases as an immobilised artificial membrane (IAM) and an internal surface reversal phase (ISRP) in suitable buffer/MeCN RP mobile phases. The retention factors ranged from log k_IAM,0.4 = −0.44 for the compound 9 with an ortho-chloro substitution at the phenyl ring to log k_IAM,0.3 = 1.43 for the most lipophilic 3,4-dichloro-substituted derivative (17). Analytes from the group I (7–12) were found to be explicitly less lipophilic than compounds belonging to the set II (13–17) because of the presence of the ethyl substituent instead of a more hydrophobic phenylethyl moiety (Table 5).

Generally, logs k of the para- or/and the meta-substituted solutes were higher than those of the ortho-substituted ones. In addition, the determined lipophilicity parameters of the ortho-substituted compounds were found to be distinctly lower than those predicted by computer-assisted programmes. The decreased lipophilicity parameters of the ortho-substituted molecules can be explained by taking into consideration the ortho effect. The most probably hypothesis is that the substituents in the ortho position, through-space interactions, display the steric hindrance forcing-out-of-plane free rotation, which explicitly decreases lipophilicity parameters (Hsieh and Dorsey, 1993; van de Waterbeemd et al., 1996; Lambert, 1993; Carrupt et al., 1997).

The para as well as meta substitution at the phenyl ring evoked an increase in the determined log k factors of our fused azaisocytosine-like congeners. In turn, the introduction of the second chlorine atom to the para- and meta-chloro derivatives resulted in an increase in log k parameters of two 3,4-dichloro-substituted analytes (12 and 17).

3.6 Correlations between calculated log P indices and isocratic log k values

The important chromatographic retention parameter, expressed as log k, in the partition RP-HPLC system is strictly related to the solute’ lipophilicity/hydrophobicity. Therefore this empirical retention factor is considered to reveal a correlation with the logarithm of the partition coefficient, log P, in a reference n-octanol-water binary system (Valko, 2004). The relationships between calculated log P values (by using an ACD/Percepta software) and logs k of our fused azaisocytosine-containing congeners were carefully studied. These are described by the equations provided underneath Fig. 3. These correlations were found to be common for two new classes of compounds (groups I and II) that were investigated on IAM and ISRP type columns in buffer/MeCN RP mobile phases. The predicted values of log P by using an ACD/Percepta programme plotted against the experimental log k _ISRP factors of the studied compounds revealed the most satisfactory statistical significance levels (p = 0.00002) and higher square correlation coefficients (R² ranged from 0.8811 to 0.8814).

Close

Fig. 3 Correlations between log P values for the investigated compounds and their retention factors, log k. Statistical terms for the derived regression equations: R² – are the squared correlation coefficients, s – denote the standard errors of estimate, F – are the values of the Fischer test of significance, p – denote the significance levels of the whole equations.

Export to PPT

3.7 Correlations between predicted log P values by milog P algorithm and the retention factors

Noteworthy is that the highly statistically significant linear correlations (with R² values ranging from 0.9619 to 0.9933; p < 0.000004) were described between the predicted log P coefficients by using milog P algorithm and the retention factors, logs k, (Fig. 4) after the exclusion from the regression analysis two evident ortho-chloro outliers (9 and 15). These highly significant correlations show the reversed phase HPLC-based approaches with IAM and ISRP type columns to be good techniques for investigating the lipophilicity of our antiproliferative active fused azaisocytosine-like congeners.

Close

Fig. 4 Correlations between predicted milog P values for the investigated compounds and their retention factors, log k. The evident outliers marked in red (9 and 15) were excluded from the regression analysis. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

3.8 Principal component analysis

PCA-based approach was employed to compare the retention factors of the investigated solutes (log k_IAM,0.3, log k_IAM,0.4, log k_ISRP,0.3, log k_ISRP,0.4) as well as their predicted lipophilicity parameters (log P, milog P) with their in silico bioactivity descriptors, such as: the oral bioavailability (%F), the permeability through the intestinal epithelium of Caco-2 cells (Caco-2), the fraction unbound in brain tissue (f_u,brain), the blood-brain barrier coefficient (log BB), the distribution in water-human serum albumin (log P_HSA) and an effective permeability in human jejunum (P_e,jejunum).

The overall twelve data set variables were submitted to the PCA. The first two principal components accounted for 96.5%, while the first three principal components accounted for 98.2% of the total variance. The relationships between all the variables are depicted in the loading plot (Fig. 5A). The pharmacokinetic descriptors such as log P_HSA, Caco-2, log BB, determined experimentally retention factors (log k_IAM,0.3, log k_IAM,0.4, log k_ISRP,0.3, log k_ISRP,0.4) and calculated partition coefficient logarithms (log P, milog P) proved to be similar to each other because of forming an “acute arrow” on the right side of the chart, showing that their first principal component revealed a positive value. In this case the in silico bioactivity descriptors were found to be proportional to the determined experimentally and calculated retention parameters. In turn, such bioactivity descriptor variables as P_e,jejunum, f_u,brain and %F were found to be inversely proportional to the empirical logarithms of retention factors as well as to the calculated logarithms of partition coefficients. These variables are placed on the left side of the chart, showing that their first principal component had a negative value. In turn, similarities and dissimilarities existed between molecular structures of the compounds that were tested according to the chromatographic, partitioning and bioactivity variables are clearly seen in Fig. 5B. The results obtained by the principal component (the loading plot shown in Fig. 5A) and cluster variables (the dendrogram presented in Fig. 5C) analyses are consistent, confirming the high degree of similarity between chromatographic and bioactivity descriptors.

Close

Fig. 5 (A) The loading plot. (B) The score plot. (C) Dendrogram.

Export to PPT

3.9 Correlations between chromatographic lipophilicity parameters and in silico bioactivity descriptors of the compounds investigated

We have decided to employ the data sets of experimentally derived retention factors (logs k) as well as in silico bioactivity descriptors (e.g., %F, Caco-2, f_u,brain, log BB, log P_HSA, P_e,jejunum) for evaluating the pharmacokinetic properties of novel antiproliferative active compounds and disclosing the most promising drug candidates.

It is commonly accepted that bioavailability of each novel pharmaceutical substance is very important in its pharmacokinetics and therefore must be considered when calculating its dosages for non-intravenous routes of administration. The in silico oral bioavailabilities after 50 mg dose administration (referred as %F values listed in Table 5) changed from 12% for the most lipophilic 3,4-Cl₂ compound (17) up till 99% for three substances (7–9) that appear to be the most promising orally bioavailable candidates. The statistically significant correlations (with R² in the range from 0.8145 to 0.9359; p < 0.02) were found between %F values and log k factors of the solutes studied (Fig. 6) after exclusion from the regression equations four evident outliers (9, 12, 15 and 17). The reversed parabolic shapes of these correlation lines suggested the existence of an optimum lipophilicity range for a satisfactory oral bioavailability of solutes investigated.

Close

Fig. 6 Relationships between oral bioavailability predictors (%F) for the investigated compounds and their retention factors, log k. The evident outliers marked in red (9, 12, 15 and 17) were excluded from the regression analysis. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

All the studied fused azaisocytosine-like congeners have been identified as strongly antiproliferative active agents, and therefore their ability for penetrating the intestinal walls could be of great importance. All these screened compounds, in accordance to the permeability classification (Yamashita et al., 2000; Yazdanian et al., 1998), were identified as being highly permeable (Table 5). Hence, these molecules are expected to be easily absorbed through the intestinal walls. The in silico Caco-2 values ranged from 167 × 10⁻⁶ cm s⁻¹ for the solute 7 to 244 × 10⁻⁶ cm s⁻¹ for the most lipophilic compound 17, whereas the P_e,jejunum values ranged from 7.66 × 10⁻⁴ cm s⁻¹ for the most lipophilic compound 17 to 8.34 × 10⁻⁴ cm s⁻¹ for the solute 10. The relationships between Caco-2, P_e,jejunum predictors and logs k are shown in Figs 7 and 8, respectively.

Close

Fig. 7 Relationships between Caco-2 predictors for the investigated compounds and their retention factors, log k. The evident outliers marked in red (9 and 15) were excluded from the regression analysis. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

Close

Fig. 8 Relationships between the P_e,jejunum predictors for the investigated compounds and their retention factors, log k. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

In the case of these highly permeable solutes the highly statistically significant inverse parabolic relationships (with R² ranged from 0.9227 to 0.9786; p < 0.0005) were obtained between Caco-2 and log k factors (Fig. 7) after removing from the regression equations two evident ortho-chloro outliers (9 and 15). The curvilinear shape of these relationships suggested that there is an optimal range of the log k factors for these compounds with regard to the satisfactory intestinal absorption. When solutes reveal lower than optimal retention factors, their intestinal absorption decreases. This may be explained by the fact that agents that are not lipophilic enough would be poorly permeable through a lipid bilayer of biomembranes, including intestinal walls.

When analysing the correlation obtained between P_e,jejunum and log k factors (with R² ranged from 0.6097 to 0.7490; p < 0.005) it is evident that an effective permeability in jejunum weakly increases with the decreasing log k values of the compounds belonging to both classes (7–17) (Fig. 8).

The most of the investigated compounds were predicted to have high brain penetration as well as brain to blood distribution, and simultaneously a lower fraction unbound in the brain tissue. Their log BB values changed from 0.117 for the compound 7 to 0.701 for the most lipophilic analyte (17), whereas their f_u,brain values ranged from 0.62 for the solute 7 to 0.04 for the most lipophilic agent (17) (Table 5). All the investigated compounds have been identified as strongly antitumour active structures, and therefore their affinity to permeate the cerebral barrier seems to be of great importance, for example in an efficient treatment of brain tumours. The highly statistically significant linear correlations of log BB values versus log k factors (with R² ranged from 0.7335 to 0.8424; p < 0.0008), were found for all the compounds that were studied (Fig. 9). The linear shape of these relationships suggested that with increasing log k factors the ability of the investigated compounds to penetrate a blood-brain barrier increases. The relationships between the fraction unbound in brain and the determined empirically retention factors of fused azaisocytosine-containing congeners were also investigated (Fig. 10 and Table 5). In this case the highly statistically significant relationships were obtained (with R² ranged from 0.8941 to 0.9680; p < 0.001) after removing two evident ortho-chloro outliers: 9 and 15 from the regression equations. The parabolic shapes of these correlations suggested that an increase in compounds’ log k indices reduces the concentration of their fraction unbound in the brain tissue.

Close

Fig. 9 Relationships between the log BB predictors for the investigated compounds and their retention factors, log k. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

Close

Fig. 10 Relationships between the f_u,brain predictors for the investigated compounds and their retention factors, log k. The evident outliers marked in red (9 and 15) were excluded from the regression analysis. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

The in silico log P_HSA values, that denote compounds’ distribution in water-human serum albumin, were analysed and correlated with logs k. Two ortho-chloro substituted solutes (9 and 15) were excluded from the regression equation as the evident outliers. They were assessed as having a greater affinity to human serum albumin presumably due to their lower retention resulting from steric effects. The highly statistically significant linear relationships (with R² ranged from 0.8953 to 0.9639; p < 0.0002) were obtained between the log P_HSA predictors and the retention parameters of the compounds (Fig. 11). The linear shape of these correlations suggested that with increasing log k values the distribution of the compounds investigated in water-human serum albumin also increases. The linear relationships between the xenobiotics’ affinities for human serum albumin and retention data determined on IAM type columns have previously been reported by the other researchers (Janicka and Pachuta-Stec, 2014; Kaliszan et al., 1996; Nasal et al., 1994; Barbato et al., 2007). On the other hand, Kaliszan (2007) has thoroughly reviewed and emphasised a number of advantages for modelling important pharmacokinetic properties of biologically active agents from different chemical classes based on their determined retention factors on IAM columns.

Close

Fig. 11 Relationships between the log P_HSA predictors for the investigated compounds and their retention factors, log k. The evident outliers marked in red (9 and 15) were excluded from the regression analysis. All statistical terms for the derived regression equations are explained underneath Fig. 3.

Export to PPT

In conclusion, the statistically significant correlations were obtained between the in silico bioactivity descriptors and the retention parameters measured directly for a given RP-HPLC system. This means that IAM and ISRP columns, as reliable in vitro systems, should be recommended not only for studying the lipophilicity but also for modelling pharmacokinetic properties of the compounds that were investigated from their measured retention factors.

3.10 Assessment of Lipinski and related properties for the investigated compounds

We have employed the available online molinspiration programme (http://www.molinspiration.com/cgi-bin/properties) to gain an information about important molecular properties of two unknown classes of the newly synthesised and investigated fused azaisocytosine-like congeners (Table S1 in Supplementary material). It has been found out that all the derivatives belonging to the discussed here classes obey the rule of five. In addition none evaluated molecule violates although onefold the predictive Pfizer’s rule (Lipinski et al., 2001). This is a significant finding because of numerous examples that a twofold violation of the rule of five may lead to serious bioavailability problems in the classes of commonly used pharmaceutics (Lipinski et al., 2001). It is worth noting that the investigated antitumor active structures possess all four physico-chemical properties that would make them likely orally active pharmaceutics. Hence, each of these compounds has an optimised probability of being well absorbed and therefore to reach easier the place of its action.

Analysing the data obtained from computer-assisted simulation experiments it has been proved that all these molecules reveal optimal ranges of the predicted topological polar surface areas and molecular flexibility, expressed by a number of freely rotatable bonds (Table S1 in Supplementary material), which hypothetically should facilitate their passive permeation through cell biomembranes, including the intestinal walls (Eartl et al., 2000; Veber et al., 2002).

Finally, the aforementioned molinspiration software was employed for assessing the possibility of screened molecules to bind for G-protein coupled receptors (GPCRs), ion channels, nuclear receptors and act as enzyme inhibitors. This available virtual screening service is especially useful in trials aiming to find possible novel ligands modulating G-protein coupled receptors due to a lack of information about this receptor 3D structure. The results of the in silico calculations (Table S2 in Supplementary material) revealed that the binding properties of fused azaisocytosine-containing congeners from class II (compounds 13–17) are explicitly higher than those belonging to class I (compounds 7–12) and this may be due to their higher lipophilicity. Amongst two novel classes of molecules that were investigated the compounds belonging to class II (13–17) predicted to have the highest probability to bind for G-protein coupled receptors and act as enzyme inhibitors. Binding ability for GPCRs of these compounds would be predictably due to their structural similarity to the previously identified bicyclic and tricyclic adenosine receptor antagonists (Szuster-Ciesielska et al., 2012; Taliani et al., 2012), capable of interacting via G-protein coupled receptors, and containing in their structures the privileged 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one template. The target compounds, as likely GPCR ligands, possess both lipophilic substituents (e.g., phenyl/substituted phenyl at the N-8 and 2-phenylethyl substitution at the C-3) and the carbonyl group at the C-4 that presumably could be engaged in hydrogen bonds with G-protein coupled receptors (Taliani et al., 2012).

The majority of the investigated molecules predicted to have higher probability to modulate G-protein coupled receptors than previously disclosed nonselective antagonist of adenosine receptors acting through interactions via specific G-protein coupled receptors, e.g., 8-(4-methoxyphenyl)-4-oxo-4,6,7,8-tetrahydroimidazo[2,1-c][1,2,4]triazine-3-carbohydrazide, namely IMT (Szuster-Ciesielska et al., 2012).

The predicted probability of the molecules investigated to act as enzyme inhibitors may be explained by the fact that these fused azaisocytosine-like congeners as false metabolites may inhibit the activity of enzymes involved in the biosynthesis of DNA (Rusinov et al., 2008; Aapro, 1994).