Synthesis, structure elucidation, determination of antiproliferative activities, lipophilicity indices and pharmacokinetic properties of novel fused azaisocytosine-like congeners

2.1 Instrumentations and materials

Ethyl 4-nitrophenylglyoxylate of 98% purity was supplied from Alfa-Aesar GmbH (Germany) and was sufficiently pure to be used directly without further purification. Triethylamine (Et₃N) (Sigma–Aldrich, Germany) was dried over anhydrous potassium hydroxide and distilled before use. Anhydrous solvents for the synthesis (e.g., n-butanol, N,N-dimethylformamide, methanol) were purchased from Roth (Germany) and Merck (Germany) companies. Additionally, the polar solvents, such as a distilled water and methanol, were used to remove successfully a soluble by-product derived from the synthetic approach, e.g., triethylammonium iodide which was the explicitly better soluble in these solvents than were the isolated solid final products (10–18). A Heidolph rotary evaporator (Germany), working under reduced pressure, was used to concentrate the volatile components of the reaction mixture to a suitable volume after the completion of the synthesis process. Melting point (m.p.) of each novel congener was determined on a Boetius apparatus (Germany) and the given value is uncorrected. The purity of each compound was checked by thin-layer chromatography. For this aim the routine preparative TLC analysis was carried out on commercial Merck 0.2 mm SiO₂ gel aluminium sheets (60 F-254) with a fluorescence indicator that allows to visualize of each final product spot under UV light at an excitation wavelength of 254 nm. The combustion microanalyses for each element analyzed (C, H, Cl, N) were consistent within ±0.4% of the theoretical values. ¹H NMR spectra (that were observed at 300 MHz for ¹H nuclei) and ¹³C NMR spectra (that were observed at 75 MHz for ¹³C nuclei) for all novel compounds of this class (10–18) dissolved in deuterated dimethyl sulfoxide (DMSO-d₆) were carried out on a Bruker Avance 300 MHz spectrometer. DEPT-135 experiments were done additionally to fully assign carbon multiplicities for all the molecules that were analyzed. All the chemical shifts are given as δ values that were downfield in relation to an internal standard, e.g., tetramethylsilane (TMS) (δ = 0.00 ppm). Chemical shifts are provided in parts per million (ppm), whereas the magnitudes of coupling constants are reported in frequency units, e.g., hertz (Hz). Signal multiplicities, used to describe signals, are indicated as follows: s, singlet; d, doublet; t, triplet and m, multiplet. The reported retention times (t_R) of solutes 10–18 were determined on IAM column in 20 °C (293 K), using as a mobile phase 50% acetonitrile in the buffer adjusted to pH 7.4.

2.1.3

2.1.3 An assessment of DNA synthesis and cytotoxicity

The carried out determination is based on the measurement of the amount of the labelled synthetic nucleoside analogue, e.g., 5-bromo-2′-deoxyuridine (BrdUrd), incorporated into the cellular DNA in place of thymidine during the nascent synthesis phase in a panel of cells that were treated with the compounds of interest. Cells treated only with the culture media served as controls. A panel of four tumour cell lines (such as: A549 (ECACC 86012804) – human Caucasian lung carcinoma cells, HeLa (ECACC 93021013) – human Negroid cervix epitheloid carcinoma cells, T47D (ECACC 85102201) – human breast carcinoma cells and TOV112D (ATCC CRL-11731) – human ovarian malignant adenocarcinoma cells) and one normal cell line (e.g., Vero (ECACC 88020401) – African Green Monkey kidney cells) of the same epithelial origin were employed. The used cell lines were sourced and cultured as described in detail previously (Sztanke et al., 2015).

End-point determinations were performed on an enzyme-linked immunosorbent assay (ELISA) reader (BIO-TEC Instruments, USA) using a commercially available chemiluminescent BrdUrd labelling and detection kit III (Roche Diagnostics, Indianapolis, USA) (Muir et al., 1990; Ellwart and Dormer 1985; Magaud et al., 1988; Huong, 1991). Each freshly prepared stock solution of the screened compound in dimethyl sulfoxide (Sigma–Aldrich, Germany) of above 99.9% purity and filtered through an aseptic 0.22 μm-pore-size syringe filter (GE Water and Process Technologies, USA) was finally diluted into the cell culture media to obtain its 50 μg mL⁻¹ concentration. The effect of each investigated solute at the tested concentration on the BrdUrd incorporation was assessed spectrophotometrically after 24-, 48- and 72-h periods of incubation at standard conditions (37 °C, 5% CO₂, 90% humidity) and compared to the untreated control cells. Table 1 specifies the percentage values of growth inhibition over a time in malignant and normal cells for all novel compounds of interest. These values were determined in the following manner: at first mean absorbance levels of at least three measurements taken from the repeated true independent BrdUrd-based evaluations were derived; then these mean values were converted into the per cent of growth inhibition. A well-known folate antimetabolite – pemetrexed (Lilly, Slovakia) at a concentration of 0.23 mM (100 μg mL⁻¹) was used as a positive control (Sztanke et al., 2015) for the DNA synthesis and cytotoxicity measurements.

Table 1 Growth inhibition ratios of normal and tumour cells by an effective concentration of novel small molecular weight fused azaisocytosine-like congeners.

Compound	R	Incubation time	Growth inhibition (%)a in cell lines
			Vero	A549	HeLa	T47D	TOV112D
10	H	24 h	40 ± 4.5	70 ± 7.2	80 ± 7.8	100 ± 12.2	80 ± 7.8
		48 h	95 ± 10.6	90 ± 8.9	90 ± 10.1	100 ± 9.9	100 ± 12.0
		72 h	100 ± 15.2	100 ± 12.4	100 ± 9.8	100 ± 10.6	100 ± 9.6
11	2-CH3	24 h	50 ± 6.5	75 ± 8.0	80 ± 8.2	60 ± 5.8	80 ± 6.6
		48 h	75 ± 5.2	80 ± 7.9	90 ± 10.2	80 ± 7.9	90 ± 8.4
		72 h	95 ± 9.8	100 ± 11.4	100 ± 12.4	90 ± 8.9	100 ± 9.5
12	4-CH3	24 h	10 ± 1.8	25 ± 2.4	85 ± 8.2	90 ± 10.2	80 ± 6.2
		48 h	80 ± 7.5	50 ± 5.6	90 ± 8.6	100 ± 10.6	90 ± 5.2
		72 h	100 ± 15.4	80 ± 7.9	100 ± 10.6	100 ± 12.1	95 ± 7.2
13	2,3-(CH3)2	24 h	15 ± 1.9	25 ± 3.0	75 ± 6.9	80 ± 7.8	50 ± 3.5
		48 h	50 ± 3.8	75 ± 8.1	85 ± 8.1	95 ± 8.6	90 ± 6.7
		72 h	90 ± 12.5	85 ± 8.0	90 ± 8.9	100 ± 10.8	100 ± 11.2
14	2-OCH3	24 h	10 ± 1.5	30 ± 2.7	85 ± 7.9	80 ± 6.8	80 ± 6.8
		48 h	40 ± 5.6	75 ± 8.2	90 ± 10.5	90 ± 7.8	90 ± 8.8
		72 h	75 ± 5.2	80 ± 8.8	100 ± 9.6	100 ± 9.5	100 ± 9.7
15	2-Cl	24 h	15 ± 1.2	50 ± 4.8	65 ± 5.9	90 ± 6.8	80 ± 5.9
		48 h	50 ± 5.8	90 ± 8.9	80 ± 7.8	100 ± 9.2	100 ± 9.5
		72 h	80 ± 8.2	100 ± 11.8	100 ± 12.1	100 ± 8.6	100 ± 10.8
16	3-Cl	24 h	75 ± 6.8	80 ± 7.9	85 ± 7.9	95 ± 5.8	80 ± 6.9
		48 h	85 ± 7.9	100 ± 11.2	95 ± 9.5	100 ± 8.8	100 ± 8.8
		72 h	100 ± 9.2	100 ± 9.8	100 ± 10.9	100 ± 9.8	100 ± 9.4
17	4-Cl	24 h	50 ± 4.8	75 ± 6.8	85 ± 8.2	95 ± 8.2	90 ± 6.5
		48 h	100 ± 11.2	100 ± 9.8	90 ± 8.8	100 ± 11.7	100 ± 8.5
		72 h	100 ± 10.6	100 ± 10.6	100 ± 10.5	100 ± 10.4	100 ± 9.2
18	3,4-Cl2	24 h	30 ± 3.2	80 ± 8.4	90 ± 7.8	95 ± 6.8	90 ± 7.5
		48 h	70 ± 7.4	100 ± 12.2	100 ± 8.9	100 ± 8.8	100 ± 9.2
		72 h	100 ± 12.5	100 ± 11.5	100 ± 10.5	100 ± 12.6	100 ± 7.5
A reference drug		24 h	5 ± 0.2	10 ± 0.9	5 ± 0.7	10 ± 0.8	nd
		48 h	10 ± 1.1	30 ± 2.4	10 ± 0.9	20 ± 1.2	nd
		72 h	20 ± 1.9	50 ± 2.8	20 ± 1.5	30 ± 2.2	nd

Normal epithelial cell line: Vero – (ECACC 88020401) – African Green Monkey kidney cells.

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

An effective concentration of 50 μg mL⁻¹ corresponds to the following millimolar concentration values: 0.149 mM (10), 0.143 mM (11, 12), 0.138 mM (13), 0.137 mM (14), 0.135 mM (15, 16, 17), 0.124 mM (18).

A reference drug – pemetrexed at a concentration of 100 μg mL⁻¹ (0.23 mM) (Sztanke et al., 2015).

nd – not determined.

a Each value is the mean of at least three independent experiment values ± standard deviation.

3.1 Synthesis and spectroscopic characterization of a novel class of fused azaisocytosine-containing congeners (10–18)

The patented biologically active 1-aryl-2-hydrazinylideneimidazolidines (1–9) (Sztanke, 2012) were exploited (as excellent nucleophilic centred building blocks) in synthesis of novel 3-(4-nitrophenyl)-8-aryl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (10–18).

The target fused azaisocytosine-like congeners (10–18), comprising in their structures the incorporated 6-azaisocytosine template, were successfully prepared employing the general synthetic pathway, as shown in Scheme 1. An efficient and simple procedure has been developed that allows a patent pending synthetic approach of that novel class of compounds (Sztanke and Sztanke, 2014a,b). It is clear that in this case a new 1,2,4-triazinone template was annelated to the existing four atoms’ fragment (N—C—N—N) of nucleophilic centred hydrazinylideneimidazolidines employing ethyl para-nitrophenylglyoxylate (as an excellent electrophilic donor of the two carbon atoms’ fragment (C—C⚌O)) according to the [4+2] pattern (Rusinov et al., 2008). For this purpose, the molar equivalents of the available hydroiodide salts of hydrazinylideneimidazolidines bearing various substitution patterns on the phenyl ring (1–9) were successfully treated with ethyl para-nitrophenylglyoxylate in the suitable reaction media under basic conditions. Initially, the condensation process was conducted by gently heating the above starting materials in n-butanol medium containing a small molar excess of triethylamine leading to the formation of suitable ketimine intermediates (drawn in square bracket in Scheme 1). In this case the nucleophilic attack of the most reactive exocyclic amino-type nitrogen of hydrazono moiety to the reactive electrophilic carbonyl carbon at the position next to the ester group took place. The step of condensation would be expected to proceed with a concomitant removal of water molecule, and additionally led to an evolution of the hydrogen iodide from the reaction mixture which had reacted with Et₃N to give triethylammonium iodide (Et₃NH⁺I⁻) as a by-product of alkylation. Subsequently, ketimine-type intermediates, containing the protonated endocyclic N3 nitrogen, are cyclized on heating in an inert refluxing n-butanol/DMF mixture to novel fused azaisocytosine-like congeners (10–18) with a concomitant loss of ethanol molecule. The results of combined ¹H NMR/¹³C NMR investigations have fully defined the structures of final products that are based on the privileged 7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-one template. Moreover, the opposite mode of cyclization to 3-(4-nitrophenyl)-7-aryl-5H-6,7-dihydroimidazo[2,1-c][1,2,4]triazole derivatives (that has also been thoroughly investigated) proved to be impossible under the established reaction conditions since no traces of such type of compounds were detected in the clear spectroscopic data of the final products. It is now believed that thermal cyclization of intermediates to the desired products proceeds in the direction which yields the most stable cyclic products possessing the least energy.

Close

Scheme 1

A scheme for the synthesis of 3-(4-nitrophenyl)-8-aryl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (10–18).

Export to PPT

The complete spectroscopic characterization of novel fused azaisocytosine-like congeners (10–18) is specified in Subsection 2.1.2. describing the experimental protocols. The title compounds (10–18) exhibited ¹³C NMR chemical shifts of the C7 atoms at δ in the region 45.4–48.3 ppm, whereas the corresponding lower shifts of the C6 atoms were found to be at δ in the region 40.5–41.7 ppm. These findings proved that there are two endocyclic methylene groups (in the five-membered imidazolidine ring) that reveal an unequal chemical character. Obviously these secondary methylene carbon atoms (C6 and C7) were unambiguously identified as the signals that go down, when analysing the results of DEPT-135 experiment. ¹H NMR chemical shifts of the aromatic proton signals (derived from two rings attached to the bicyclic scaffold) were found to be at δ in the region 7.04–8.46 ppm, as indicated in the case of all the compounds (10–18). However, the compound 13 revealed the proton signals derived from the para-nitrophenyl ring (as two doublets in the aromatic region at δ 8.29 ppm and δ 8.38 ppm with coupling constants of J = 9.1 Hz and 9.1 Hz, respectively, integrating with two protons each) that were shifted the most downfield from those proton signals of the second 2,3-dimethylsubstituted phenyl ring (7.19–7.32 ppm).

In the present study we have optimized the general synthetic process by an adequate choice of reagents, solvents and suitable conditions that were required to synthesize the desired final products. It has been proved that the synthesis of novel fused-azaisocytosine-containing congeners (10–18), although conducted under mild conditions, proceeded with relatively good overall yields. Higher yields were obtained when the cyclization of the hypothetical intermediate ketimine-type Schiff bases was carried out for the suitable reaction period in refluxing n-butanol/DMF solvent mixture containing a small molar excess of Et₃N. The general synthetic procedure was developed in the present study on a multi-gram scale that could be applicable to a number of identifiable derivatives of this novel class.

Preparative procedures of the title compounds as well as reactions of 1-aryl-2-hydrazonoimidazolidine derivatives with ethyl para-nitrophenylglyoxylate have not been reported in the scientific literature yet, including Chemical Abstracts Service (CAS). Thus, this not studied in the past approach for the synthesis of an unknown class of fused azaisocytosine-like congeners (with the common 4-nitrophenyl moiety at the C-3 and different substitution patterns at the N-8) is the first time presented. As the synthesized compounds contain a nitro substitution on the aromatic phenyl ring they could be utilized in further synthesis of arylamines as well as aryldiazonic salts and their related derivatives.

The initial formation of the suitable ketimine-type Schiff bases, e.g., ethyl (2Z)-(4-nitrophenyl)[(2E)-(1-arylimidazolidin-2-ylidene)hydrazinylidene]acetates (II), as intermediates in the synthesis of final products (10–18), probably proceeds through an intermediate stage according to the mechanism that may be represented by Scheme 2. A nucleophilic attack of an amine-type nitrogen of the suitable hydrazinylideneimidazolidine to an electrophilic α-carbonyl carbon of ethyl para-nitrophenylglyoxylate results in the formation of hemiaminal (I), e.g., ethyl hydroxy(4-nitrophenyl)[(2E)-2-(1-arylimidazolidin-2-ylidene)hydrazinyl]acetate. Its further conversion into the suitable ketimine (II) followed with concomitant dehydration is accomplished by deprotonation of the nitrogen in hemiaminal molecule and pushing the oxygen off of the carbon by the electrons of the same nitrogen atom. The first reversible step of this reaction is believed to be largely dependent on the rate of dehydration of I.

Close

Scheme 2

The proposed stepwise mechanism for the formation of ketimine-type Schiff bases (II).

Export to PPT

3.2 In vitro antiproliferative activities of novel 3-(4-nitrophenyl)-8-aryl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (10–18)

The designed and synthesized fused azaisocytosine-like congeners (10–18) were screened thoroughly for their in vitro antiproliferative effects (measured by BrdUrd incorporation) against a panel of five reference cell lines (A549, HeLa, T47D, TOV112D and Vero) of the same epithelial origin. An efficient, reliable and reproducible BrdUrd-ELISA method is a derivative of the traditional radioactive [³H]-thymidine incorporation assay. The main advantage of this approach is that it permits a quantitative estimate of the incorporation of the labelled nucleoside analogue (as the fraudulent thymidine precursor) without the possibility of working with radioactive isotopes (Sztanke et al., 2015). The employed assay is commonly accepted in the preliminary phase of the discovery process directed towards development of novel anticancer agents (Vega-Avila and Pugsley, 2011). The BrdUrd-based method proved to be as sensitive as [³H]-thymidine assay in previous in vitro investigations (Muir et al., 1990; Hawker, 2003). According to the manufacturer’s protocol the results obtained with these two assays are strongly correlated together.

Table 1 specifies the percentage growth inhibition ratio of both tumour and normal cells by the 50 μg mL⁻¹ concentration of all fused azaisocytosine-like congeners (10–18) after 24-, 48- and 72-h periods of incubation. Table 2 lists the IC₅₀ values (the half maximal inhibitory concentrations calculated from the concentration-dependent inhibition curves) for the examined compounds.

All novel 3-(4-nitrophenyl)-8-aryl-7,8-dihydroimidazo[2,1-c][1,2,4]triazin-4(6H)-ones (10–18) exerted a cytotoxic effect in the majority of recruited tumour cells. It should be noted that all molecules at an effective concentration (50 μg mL⁻¹) revealed significantly higher antiproliferative effects in three malignant cells (A549, HeLa and T47D) than those evoked by the 100 μg mL⁻¹ concentration of commonly used folate antimetabolite – pemetrexed (Sztanke et al., 2015). Pemetrexed was employed in our studies as a standard drug, since this congener of folic acid is clinically used in patients suffering from non-small cell lung cancer. The inhibition of proliferation of tumour cells by the investigated congeners having incorporated fused azaisocytosine template is thought to be due to competition for nucleotides required for DNA biosynthesis (Rusinov et al., 2008).

Analysing the antiproliferative activities of the target fused azaisocytosine-containing congeners with the phenyl, alkylphenyl, dialkylphenyl and alkoxyphenyl substitution pattern at the N-8 (10–14) it has been noticed that the parent compound (10), with an electron-rich phenyl ring, is more cytotoxic for T47D, A549 and TOV112D cells than are either its substituted derivatives. However, the most promising compounds seemed to be 12 and 14 (bearing the electron-donating para-CH₃ and ortho-CH₃O group, respectively, at the phenyl ring) because they possess the highest cytotoxic activity against HeLa cells, displaying only a slight toxicity on normal cells after a 24-h incubation period.

Amongst the compounds bearing the mono- and dichlorophenyl substitution at the N-8 of the template (15–18), the derivatives 16, 17 and 18 (that have electron-withdrawing groups attached at meta-Cl, para-Cl and 3,4-Cl₂ position, respectively, of the phenyl ring) were chosen as capable of producing the highest cytotoxic effects in T47D cells. The highest antiproliferative activities in TOV112D cells were achieved by introducing the 4-Cl and 3,4-Cl₂ group at the phenyl ring (compounds 17 and 18, respectively), whereas in A549 cells – via the 3-Cl and 3,4-Cl₂ substitution (compounds 16 and 18, respectively). The derivative 15 with the Cl group at an ortho position on the phenyl moiety, although displaying cytotoxic effects in tumour cells, proved to be the lowest toxic for normal cells in the series of chloro-substituted congeners, particularly when investigated after 24 and 48 h of incubation. Moving the Cl group to the meta position of the phenyl ring resulted in still antiproliferative active compound 16 (against T47D and TOV112D cells), with an increased cytotoxicity for A549 and HeLa cells after 24 and 48 h of incubation, respectively, when compared to 15. Almost none of the enhancement in the antiproliferative activity against tumour cells was attained by introducing the chlorine atom into the position para of the phenyl ring, as proved for 17 when compared to 16. In the case of the antiproliferative active compound with the 3-Cl substitution (16), introducing a further Cl substituent at position para of the phenyl moiety resulted in still antiproliferative active 3,4-Cl₂ substituted molecule (18), although revealing the decreased cytotoxicity for normal cells after 24- and 48-h incubation periods.

The bioisosteric replacements of the CH₃ group (susceptible to oxidation in vivo (Patrick, 2009)) that is placed both in the ortho and in the para position of the phenyl ring (compounds 11 and 12, respectively) with a Cl group (resistant to oxidative biotransformation in vivo (Patrick, 2009)) led to in vitro antiproliferative active bioisosteres (molecules 15 and 17, respectively). The best balance of cytotoxicity versus more selectivity was achieved by introducing a Cl group (compound 15) in the case of the ortho-substituted isosteres (11 and 15) or a CH₃ group (compound 12) in the case of para-substituted isosteres (12 and 17).

Introducing the methoxy group at the ortho position of the phenyl ring turned out to be the successful structural modification leading to a decrease in the cytotoxicity towards normal cells, which was particularly seen after 24 and 48 h of incubation as confirmed in the case of the compound 14, when compared to the parent structure (10).

Placing a methyl group ortho to the phenyl ring led to a decrease in cytotoxicity towards normal cells only after 48- and 72-h incubation periods, as confirmed in the case of the structure 11 (when compared to 10). Introducing a further methyl group to the phenyl ring resulted in a decrease of toxic effects of the derivative 13 for normal cells mostly after 24 and 48 h of incubation, when compared to 11. Placing a methyl group para to the phenyl ring led to a decrease in cytotoxicity against normal cells after 24- and 48-h incubation periods, as confirmed in the case of the molecule 12 (when compared to 10).

Experimental results proved that the attachment of mostly electron-donating moieties such as 4-CH₃, 2,3-(CH₃)₂, 2-CH₃O and electron-withdrawing 2-Cl and 3,4-Cl₂ groups at the phenyl ring, enhanced the selectivity for malignant over normal cells of the compounds 12, 13, 14, 15 and 18 after 24 h of their incubation. However, they showed rather little selectivity after 48- and 72-h incubation periods.

3.3 Assessment of the drug-likeness of the investigated fused azaisocytosine-containing congeners (10–18)

According to the Lipinski’s rule of five (Lipinski et al., 2001), all the reported in this study unknown congeners (10–18) have the increased chances of good oral bioavailability by the reason of fact that they have the following: the calculated log P values less than 5 and the overall molecular weights below 500 Da, not more than 5 H-bond donor and 10 H-bond acceptor sites. Furthermore, the analyzed compounds predicted to be conformationally stable as their molecules have less than 10 rotatable bonds, whereas their topological polar surface areas are less than 140 Ų. These predictable molecular descriptors, shown in Table S1 in Supplementary material, were in the optimum range as that reported for the orally and intestinally bioavailable pharmaceutics (Veber et al., 2002; Eartl et al., 2000; Palm et al., 1997).

3.4 Chromatographic behaviour and the individual retention factors of novel solutes (10–18)

The chromatographic behaviour of this novel class of the compounds (10–18) was established. The isocratic retention factors, logs k (Yamagami et al., 1990, 1994; Klein et al., 1988), were measured directly by reversed phase liquid chromatography (RPLC) on three different columns in 20 °C (293 K) as well as 37 °C (310 K). Chromatographic analyses were carried out using buffered aqueous eluent systems containing as an organic modifier 40% (v/v) (log k _{0.4 MeCN}) and 50% (v/v) acetonitrile (log k _{0.5 MeCN}). An immobilized artificial membrane (IAM), an immobilized cholesterol (that imitate biological systems) and an internal surface reversal phase (ISRP), respectively, were employed as three different stationary materials. The choice of the endcapped stationary phases was aimed at minimizing secondary retention mechanisms. It is well known that HPLC columns without endcapping have significant amounts of silanol groups capable of secondary silanophilic interactions which interfere with the partitioning mechanism. This is especially noticeable when employing acetonitrile as the mobile phase component.

The primary advantages of the performed direct RPLC measurements should be underlined. The retention parameters, logs k, were employed as the descriptors of lipophilicity. Thus this allowed not only to significantly shorten the retention times of compounds under study but also significantly reduce the overall time and costs of RPLC measurements that were carried out. It should be noted that the reliable and reproducible logs k of the compounds of interest are different flavours of their lipophilicity. All these isocratic retention factors that are the first time presented will be indispensable in further bioactivity predictions.

The correlation matrix concerning the results obtained on three different columns is listed in Table S2 in Supplementary material. The plot correlating log k _{0.4 MeCN, 310K} and log k _{0.4 MeCN, 293K} values measured on three different stationary phases is presented in Fig. S1A in Supplementary material. On the other hand, the plot of logs k _{0.5 MeCN, 310K} versus logs k _{0.5 MeCN, 293K} is shown in Fig. S1B in Supplementary material. The obtained results showed the column with an immobilized cholesterol is the most selective in relation to the analyzed compounds because their experimental logs k are the most varied. Furthermore, it has been shown that a bit weaker selectivity to the analyzed solutes reveals the IAM column, whereas the ISRP column seems to be the distinctly less selective for these types of analytes as their determined log k factors are the least varied.

It has been shown that the structure-dependent lipophilic properties of the individual compounds have a remarkable influence on their chromatographic behaviour and retention factors (Table 3). This is, as expected, taking into account the background theory of retention explaining that in the RPLC measurements the retention is governed by hydrophobic forces and therefore the lipophilicity of the individual solute is governing the retention (Cimpan et al., 2000; Dill and Dorsey, 1994; Dorsey et al., 1989; Dorsey and Khaledi, 1993; Valko, 2004; Gocan and Cimpan, 2006; Kaliszan, 2007). It is obvious that the structure-retention behaviour of solutes in reversed phase chromatographic investigations (that are of dynamic nature) determinates their fundamental processes of pharmacokinetic phase, e.g., liberation, absorption, distribution, metabolism and elimination (LADME) that occur in a living organism (Dorsey and Khaledi, 1993; Kaliszan, 2007; Perisic-Janjic et al., 2011; Janicka et al., 2013).

The combined electronic effects of the groups within the molecules investigated and resulting steric interactions may have influence on their determined log k parameters (Table 3). It seems that these ones although in part were capable of significantly perturbing the reported earlier hydrophobicity constant increment of some substituents attached to the aromatic systems (Hansch and Leo, 1979; Fujita et al., 1964).

Effects of a number of electron-donating and electron-withdrawing substituents attached at the phenyl moiety on decreasing and increasing logs k were thoroughly investigated. The parent compound (10) has a non-substituted electron-rich phenyl as the first aromatic moiety attached at the N-8 of the common template and the electron-withdrawing para-NO₂ group at the second phenyl ring. All the solutes 11–14 possess electron-donating substituents located at the first phenyl ring in different positions and additionally the electron-withdrawing NO₂ group in the para position at the second phenyl moiety. In turn, all the compounds 15–18 have electron-withdrawing substituents located at the first phenyl ring in different positions as well as the electron-withdrawing NO₂ group in the para position at the second phenyl moiety. It has been confirmed that electron-donating substituents located on the phenyl moiety decreased the logs k of the investigated compounds in the order: 2-CH₃ > 2-CH₃O > 4-CH₃ > 2,3-(CH₃)₂, when the RPLC measurements were carried out on the cholesterol column at both temperatures and both acetonitrile volume concentrations as well as on the IAM column at 20 °C using 50% acetonitrile in the mobile phase. In turn, the log k parameters of the same compounds decreased in the order of 2-CH₃ > 2-CH₃O > 2,3-(CH₃)₂ > 4-CH₃, when the chromatographic measurements were performed on the IAM column at 37 °C employing 40% acetonitrile in the mobile phase, on the IAM column at both temperatures employing 50% acetonitrile and on the ISRP column at both temperatures and both acetonitrile volume concentrations. Furthermore, it has been observed that electron-withdrawing substituents attached at the phenyl moiety decreased the log k factors of the studied solutes in the order: 3,4-Cl₂ > 4-Cl > 2-Cl > 3-Cl, when the chromatographic measurements were carried out at both temperatures on the column with an immobilized cholesterol as the stationary phase employing both acetonitrile volume concentrations, on the IAM column using 50% acetonitrile in the mobile phase and on the ISRP column employing 40% acetonitrile. In turn, the logs k of the same compounds decreased in the order of 3,4-Cl₂ > 4-Cl > 3-Cl > 2-Cl, when the chromatographic measurements were performed at both temperatures on the IAM column employing 40% acetonitrile in the mobile phase and on the ISRP column using 50% acetonitrile.

Introducing two electron-donating methyl groups to the parent molecule (10) resulted in a noticeable decrease of the determined logs k of the 2,3-(CH₃)₂-substituted compound (18). The most probable explanation would account for the steric hindrance of the modified molecule, which decreased its lipophilicity. Generally, it has been confirmed that the log k factors of the positional isomer with the para-Cl substitution (17) are higher than those of the ortho-Cl and meta-Cl substituted isomers (15 and 16, respectively). Introducing the second chlorine atom to the para-Cl and meta-Cl derivatives (16 and 17, respectively) evoked an increase in the log k parameters of the 3,4-Cl₂-substituted solute (18). Additionally, the determined retention factors of the ortho-substituted compounds proved to be distinctly lower than those predicted by an ACD/Percepta software. The decreasing and increasing experimental log k values of the molecules with an ortho substitution at the phenyl ring can be explained by taking into account the ortho effect. The most probable hypothesis is that 2,3-(CH₃)₂ and 2-Cl substituents in the ortho position caused the steric hindrance forcing-out-of-plane free rotation, whereas 2-CH₃ and 2-CH₃O groups were engaged in internal H-bond interactions (Hsieh and Dorsey, 1993; Van de Waterbeemd et al., 1996; Lambert, 1993; Carrupt et al., 1997). This could be the cause why in our measurements 2,3-(CH₃)₂ and 2-Cl substituted molecules revealed decreased, whereas 2-CH₃ and 2-CH₃O ones had increased retention factors in relation to the parent structure (10).

3.5 Chromatographic retention factors versus partitioning lipophilicities

The relationships between computed log P values (by using an ACD/Percepta software) and experimentally obtained log k factors were carefully investigated.

Logs P versus logs k dependences were found to be explicitly linear for five compounds (10, 11, 14, 17 and 18) of nine investigated (Fig. 2). The remaining four solutes (12, 13, 15 and 16), as typical outliers, had to be excluded from the regression equation. These outliers, although revealing the explicitly overestimated log P values in in silico calculations, proved to be distinctly less lipophilic in the empirical RPLC measurements. The same figure also shows that in the case of these outliers there are significant objections against their calculated lipophilicity indices. This means that the prediction of reliable log P values of molecules 12, 13, 15 and 16 is impossible and there is a clear need to verify the computed results experimentally.

Close

Figure 2

Log P versus log k relationships obtained for different HPLC systems at two temperatures (20 °C (293 K) and 37 °C (310 K)).

Export to PPT

Statistics of the relationships obtained by plotting log P versus log k values is presented in Table 4. The statistically significant very strong linear correlations were disclosed between the logarithms of predicted 1-octanol/water coefficients (expressed as log P values) and the logarithms of chromatographic retention parameters (expressed as the isocratic log k factors) measured on IAM (R² from 0.9684 to 0.9847; p < 0.003) and ISRP (R² from 0.9492 to 0.9923; p < 0.005) columns. Logs P versus logs k correlations proved to be also strong (R² from 0.8895 to 0.9056) and statistically significant (p < 0.002) when chromatographic measurements were performed on the column with an immobilized cholesterol as the stationary phase. This means that novel lipophilicity indices, giving a consistent scale of lipophilicity with the trustworthy partitioning descriptors (expressed as log P values) of the compounds 10, 11, 14, 17 and 18, may be regarded as their sustainable values of reference. However, the computed log P values of solutes 12, 13, 15 and 16 cannot be used as the trustworthy descriptors of their lipophilicity, as confirmed above.

3.6 PCA

The principal component analysis-based approach was employed to visualize the systematic trends in existed similarities and dissimilarities between the chromatographic, partitioning and pharmacokinetic parameters characterizing fused azaisocytosine-like congeners of interest (10–18). The overall twenty data set variables including: the chromatographic retention factors (log k _{IAM 0.4, 293K}, log k _{IAM 0.5, 293K}, log k _{IAM 0.4, 310K}, log k _{IAM 0.5, 310K}, log k _{Cholester 0.4, 293K}, log k _{Cholester 0.5, 293K}, log k _{Cholester 0.4, 310K}, log k _{Cholester 0.5, 310K}, log k _{ISRP 0.4, 293K}, log k _{ISRP 0.5, 293K,} log k _{ISRP 0.4, 310K}, log k _{ISRP 0.5, 310K}), the predicted n-octanol/water partitioning parameters (expressed as log P values) and the in silico calculated molecular descriptors (log BB_permeation, log BB_distribution, log K_HSA,%F, f_{u, plasma}, Caco-2, P_{e, jejunum}) that are relevant to pharmacokinetics of the investigated compounds were submitted to a rational PCA-based approach.

Three principal components explained 64.9%, 84.6% and 95.5% of the total variance, respectively. It should be noted that the results obtained by the principal component (see the loading plot presented in Fig. 3A) and cluster variables (see the dendrogram shown in Fig. 3B) analyses are consistent, confirming high trends in similarity between the analyzed chromatographic, partitioning and bioactivity parameters.

Close

Figure 3

Loading plot (A), dendrogram (B) and score plot of PC2 against PC1 for the PCA-based approach (C).

Export to PPT

The loading plot (Fig. 3A) presents the relationships between all the variables after subjecting them to PCA. The majority of data set variables, including valuable bioactivity descriptors such as Caco-2, log K_HSA, log BB_distribution, all determined experimentally retention factors as well as the calculated partitioning coefficient logarithms, logs P proved to be similar to each other. In this case the in silico bioactivity descriptors were proportional to the retention parameters both determined experimentally and calculated in silico. All these data set variables formed an “acute arrow” on the right side of the chart. In turn, the other pharmacokinetic descriptor variables such as P_{e, jejunum}, %F, f_{u, plasma}, log BB_permeation proved to be inversely proportional to the determined logarithms of retention factors as well as to the predicted logarithms of partition coefficients. All these data set variables formed an “acute arrow” on the left side of the chart.

Furthermore, the score plot (Fig. 3C) reveals the trends in similarities and dissimilarities between molecular structures of the compounds that were tested according to the chromatographic, partitioning and bioactivity variables. The investigated compounds have been divided into two separate clusters due to differences in their molecular structures. The first explicitly linear cluster is associated with five more lipophilic solutes (10, 11, 14, 17 and 18) and the second cluster (with the scattered data points) is formed for four less lipophilic analytes (12, 13, 15 and 16). Such grouping of the tested solutes is most likely directly due to similarities in their retention mechanism (as evidenced by their logs k) obtained on different RPLC materials. The diverse grouping on the analyzed compounds may also suggest that the retention of each group of substances is controlled by the different net effect of intermolecular forces between the solute on subsequent reversed phase materials and the mobile phase (Kaliszan, 2007; Perisic-Janjic et al., 2011). It cannot be excluded that clustering of the compounds 12, 13, 15 and 16 may be presumably evoked by more complicated steric and electronic effects within their molecules that have a significant influence on their retention behaviour and consequently on bioactivity descriptors.

3.7 Correlations between chromatographic lipophilicity parameters and in silico calculated pharmacokinetic descriptors

The data sets of the determined retention factors (log k) on three different RPLC columns as well as in silico bioactivity descriptors (e.g.,%F, f_{u, plasma}, log K_HSA, log BB_distribution, log BB_permeation, Caco-2, P_{e, jejunum}) were correlated with the goal of evaluating whether the lipophilicity parameters have influence on the pharmacokinetic properties of our novel compounds. This approach was performed with the purpose of preselecting the lead candidates for further in vivo studies with optimum lipophilicity ranges relevant to satisfactory pharmacokinetics. Knowledge about pharmacokinetic properties of compounds with sufficient in vitro biological activities is especially important before subjecting them to further in vivo investigations. It is commonly known that poor pharmacokinetics, in the majority cases, results in drug candidate failure. Therefore, in silico predictive models are very helpful at an early discovery phase of new biologically active compounds that are considered as potential drug candidates (Guerra et al., 2010; Matysiak et al., 2019).

All the in silico calculated pharmacokinetic descriptors of the investigated solutes are listed in Table 3. %F value is used in an early phase of the lead candidate selection process for predicting the bioavailability of the solute after per os administration, whereas the f_{u, plasma} predictor is used for assessing a fraction unbound of the compound of interest in the plasma (e.g., actually active). Log K_HSA value is employed for predicting the solute binding with human serum albumin as the most significant protein component. Caco-2 and P_{e, jejunum} predictors are used for assessing the intestinal absorption and an effective transport in human jejunum, respectively. Log BB_distribution value predicts the distribution of the individual solute between the blood–brain, whereas log BB_permeation value predicts the ability of the solute to penetrate the blood–brain barrier.

Plots of the in silico pharmacokinetic descriptors against the experimental retention factors are shown in Fig. 4A–C and apart from that in Fig. S2 in Supplementary material. The described correlations between physico-chemical properties strictly related to pharmacokinetics and retention factors of the analyzed new compounds on different reversed phase stationary materials are presented in Table 5 and apart from that in Table S3 in Supplementary material. The determined log k factors proved to be applicable in predictions of some fundamental pharmacokinetic properties of the studied solutes such as%F, f _{u, plasma}, log K_HSA values.

Close

Figure 4

Relationships between pharmacokinetic descriptors and retention factors: (A) %F predictors versus log k factors; (B) f _{u plasma} predictors versus log k factors; (C) log K _HSA predictors versus log k factors.

Export to PPT

Fig. 4A shows the relationships between the calculated oral bioavailability values (expressed as %F) and the experimental retention factors. These dependences are described by two quadratic equations that were found to be separated for solutes 15–18 and 10–14. The statistically significant and very strong curvilinear relationships (R² in the range from 0.9979 to 0.9999) were identified between the %F predictors and the log k factors of compounds 15–18 established on IAM and ISRP columns. %F versus logs k relationships were again very strong (R² in the range from 0.9965 to 0.9985) when the measurements were performed on the column with an immobilized cholesterol. The disclosed curvilinear shapes of these correlation lines indicate that an enhancement in the solutes’ lipophilicities (expressed as logs k) reduces their bioavailability after per os administration. In turn, %F values of the compounds 10–14 plotted against the log k factors determined on the cholesterol column at the temperature of 20 °C revealed also statistically significant and very strong parabolic correlations (R² = 0.9746 and 0.9797).

Fig. 4B presents the relationships between the pharmacokinetically active fraction unbound in the plasma (f _{u, plasma}) and the determined experimentally retention factors. The parabolic shape of shown correlation lines suggests that there exists an optimum range of lipophilicity providing a proper binding of the solute in the plasma. The relationships between the f _{u, plasma} predictors and the chromatographic retention parameters are described by quadratic equations for solutes 15–18. The strongest and statistically significant correlations were identified by plotting f _{u, plasma} predictors versus the log k factors obtained from measurements that were performed at 20 °C on IAM (R² = 0.9988 and 0.9992) and ISRP (R² = 0.9993 and 0.9996) columns. In turn, in the case of solutes revealing lower binding with plasma proteins and forming the separate group (10–14) no effect of the retention factors on their optimal binding with plasma was observed. All the tested substances (10–18) predicted to have a high protein binding behaviour (according to the known protein binding classification (Idaidek and Arafat, 2012; Amidon et al., 1995)) as their overall fraction unbound in human blood plasma is relatively low (<0.1).

Fig. 4C also shows plots correlating affinities for human serum albumin (log K_HSA) of the compounds 15–18 and their retention factors determined experimentally. These relationships are described by quadratic equations. In this case it is seen that only a significant increase in the logs k of the solutes of interest evokes a significant enhancement in their binding with albumin in the human serum. The strongest and statistically significant correlations were derived by plotting log K_HSA predictors versus the log k values determined on IAM column at 37 °C (R² = 0.9998 and 1.0000) and on ISRP column at 20 °C (R² = 0.9990 and 0.9998). Our results are consistent with some findings presented in earlier studies (Kaliszan et al., 1996; Ong et al., 1996; Caldwell et al., 1998; Nasal et al., 1994; Barbato et al., 2007; Stewart and Chan, 1998; Chan et al., 2005; Janicka and Pachuta-Stec, 2014) that described the advantages in bioactivity predictions of the retention parameters determined on IAM-type columns that imitate biosystems. In contrast, in the case of compounds 10–14 no effect of their lipophilicity on an optimal binding with human serum albumin was observed.

All the investigated small molecules (10–18) predicted to be highly permeable and therefore they are expected to be easily transported through the intestinal epithelium (according to the adopted classification of Caco-2 cell permeability (Yamashita et al., 2000; Yazdanian et al., 1998)). The computed Caco-2 values ranged from 218 × 10⁻⁶ cm s⁻¹ for solute 14 to 241 × 10⁻⁶ cm s⁻¹ for the most lipophilic compound 18. In turn, the calculated P_{e, jejunum} values ranged from 7.71 × 10⁻⁴ cm s⁻¹ for analyte 13 to 7.96 × 10⁻⁴ cm s⁻¹ for the parent compound 10. These values indicate that all these solutes should be fast absorbed in the human jejunum at physiological pH (Stępnik et al., 2014). Nevertheless, in the case of all the compounds that were studied no influence of the logs k on Caco-2 as well as P_{e, jejunum} predictors, respectively, was observed (Fig. S2 and Table S3 in Supplementary material).

Amongst this novel class of fused azaisocytosine-like congeners only molecules revealing the optimum lipophilicity range relevant to satisfactory pharmacokinetics may be regarded as the most promising candidates of the series for further in vivo testing. These lead structures emerged from a rational approach aiming at processing the experimental retention factors in bioactivity predictions. Statistically significant and highly predictive relationships presented in this study confirmed that the liquid chromatography with suitable reversed phase materials of diverse separation properties (including those imitating biosystems) is an excellent in vitro system that may be successfully used for assessing some fundamental pharmacokinetic properties (such as %F, f _{u, plasma}, log K_HSA) of our new solutes from their retention factors measured directly for given systems, as described above in detail. Pharmacokinetic properties of our novel compounds will be of value during their further in vivo investigations and finally all these descriptors of bioavailability could be validated in an animal model.