Regioselective synthesis, physicochemical properties and anticancer activity of 2-aminomethylated estrone derivatives

The unique estrogen receptor (ER)-independent antiproliferative and apoptotic activity of 2-methoxyestradiol (2ME2) is well known, however, its use has been limited because of its poor oral bioavailability. In this study, novel 2-aminomethylated estrone (E) and estradiol (E2) derivatives structurally related to 2ME2 were synthesized, and their physicochemical properties as well as their in vitro cytotoxic effects were investigated in the hope to find more selective antiproliferative agents with improved pharmacokinetic profile. The target compounds were synthesized from 2-dimethylaminomethylated E obtained regioselectively by a three-component Mannich reaction. Quaternization with methyl iodide followed by reacting the ammonium salt with various dialkyl and alicyclic secondary amines afforded the desired products in good yields. The reactions proceeded via a 1,4-nucleophilic addition of the applied secondary amines to the ortho-quinone methide (o-QM) intermediates, generated in situ from the salt by base-promoted β-elimination. The compound library has been enlarged with structurally similar E2 analogues obtained by stereoselective reduction and with some 17β-benzylamino derivatives prepared by reductive amination. The potential values of the novel E and E2 derivatives were characterised by means of three different approaches. At the first step compounds were virtually screened using physicochemical parameters. Physicochemical characterization was completed by kinetic solubility and in vitro intestinal-specific permeability measurement. Antiproliferative effects were additionally determined on a panel of malignant and non-cancerous cell lines. The evaluation of the pharmacological profile of the novel E and E2 derivatives was completed with the calculation of lipophilic efficacy (LiPE).


Introduction
One of the most important tasks of medicinal chemistry is to create compound libraries that contain a large number of molecules with high structural diversity [1,2]. Multicomponent Mannich-type aminoalkylation reactions in which a product can be formed by the simultaneous reaction of three reactants (an active hydrogen-containing agent, an aldehyde and an amine reagent) serve as an effective means of achieving this goal [3][4][5][6]. Their great advantage is that the order of addition of the components is arbitrary and their structure can be varied independently, vastly increasing the number of molecules that may be obtained in a short amount of time with minimal effort. Because formaldehyde is most commonly used as the aldehyde component, the applicability of miscellaneous types of substrates with an activating functional group (e.g. carbonyl compounds, phenols, naphthols, terminal alkynes, different heterocycles) and the variability of amine reagents generally ensures the structural diversity of the aminomethylated products (Mannich bases) [5,6]. The importance of Mannich bases stems from their many practical applications, especially in the field of medicinal chemistry and drug design. In addition to their therapeutic potential in many kinds of diseases, such as cancer (Fig. 1), improvements in the physicochemical and pharmacokinetic properties are also expected due to the presence of the polar aminomethyl moiety [5].
Although different reaction mechanisms have been suggested for the three-component modified Mannich reaction of an electron-rich aromatic compound (phenol or naphthol), an amine and an aldehyde [7][8][9], the theory of the primary formation of an ortho-quinone methide (o-QM) from the phenolic compound and the aldehyde partner appears to be the most accepted. 1,4-Conjugate addition of the amine to this very short-lived reactive intermediate followed by rearomatization leads to the aminoalkylated aromatic product. It is important to note that the o-QMs formed from naphthols are more stable due to their partially retained aromatic character than those generated from phenols [8]. Nevertheless, several recent reviews [8,10] and articles [11] have addressed the properties, applicability, and biological activity of o-QMs derived from naphthols and phenols.
The phenolic A-ring of natural steroids estrone (E) and estradiol (E2) seems suitable for carrying out the Mannich reaction, however, aminomethylation can occur on both carbon atoms (C2 and C4) adjacent to the C -OH group. Therefore, achieving high regioselectivity toward the C2-substituted product may be challenging despite the fact that for steric reasons, the availability of the C2 position appears to be more favourable. However, the preference for the C2 substitution in Mannich reactions is contradicted by the results of Lande and coworkers, who studied the regioselectivity of the aminomethylation of bicyclic phenols [12]. It was found that in the Mannich reaction of phenol condensed with a six-membered ring, which structurally mimics the A-and B-rings of estrogens, the incorporation of the aminomethyl group occurs mainly at the carbon between the phenolic C -OH and the ring junction. This would correspond to position C4 of the sterane skeleton.
Nevertheless, 2-substituted E-or E2-based derivatives are of great interest, since several compounds of this type have been demonstrated to have less or no ability to bind to the estrogen receptor (ER) and thus no hormonal activity [13]. Perhaps the best known and most studied member of this family of compounds is 2-methoxyestradiol (2ME2, Fig. 1), which has an ER-independent mechanism of action to inhibit the proliferation of human cancer cells of diverse origins. Although it lacks hormonal effects, its oral bioavailability is low due to its poor solubility and intestinal absorption as well as rapid metabolism in the body [14].
Its unfavourable pharmacokinetic profile has prompted the production of several modified analogues [15]. Among others, 2-aminomethylated E2 derivatives were also prepared by regioselective 2-formylation of E and subsequent reductive amination with different primary amines [16]. These compounds and their analogues modified by Wittig reaction at C17 position, which are formally considered to be Mannich bases, showed significant antiproliferative activity on various human cancer cell lines (Fig. 1).
Based on the literature background and our research on the synthesis of cytotoxic A-ring modified E derivatives [17][18][19], the aim of the present study was to develop an efficient method for the highly regioselective aminomethylation of E and E2 at C2 position and to characterize in silico and in vitro physicochemical properties (solubility, permeability) and cytotoxic activity of our novel derivatives in comparison with 2ME2 as reference compound. According to the results of cytotoxicity studies performed on cancerous (HeLa, A2780, MDA-MB-231) and non-cancerous (NIH/3T3) cell lines, we further aimed to identify the structureactivity/selectivity relationships of the new derivatives and to select them taking into account the potential promiscuous property classified by lipophilic efficiency index (LiPE).

Synthetic studies
For the introduction of an aminomethyl group onto C2 position of the estrane skeleton, first, a one-step three-component Mannich-type method was carried out using E, dimethylamine (40 % aqueous solution) and formaldehyde (35 % aqueous solution) as reactants (Scheme 1). The amine and the aldehyde were used in excess, and the mixture was stirred in refluxing EtOH under regular thin-layer chromatography (TLC) control. To our delight, complete conversion was achieved within 2 h, and the transformation proved to be highly regioselective toward the formation of the 2-substituted product (1a), which was obtained in 85 % yield after chromatographic purification. Trace amounts of the 4substituted isomer (1'a) were thus removed successfully during purification. When the same reaction was performed in a closed vessel at 100 Fig. 1. Design of 2-aminomethylated E and E2 derivatives based on the structures of 2ME2 [14] and some cytototoxic phenolic Mannich bases [5,6,16].
• C under microwave (MW) irradiation, the reaction time was reduced to 10 min. Although C16 is also active due to the proximity of the adjacent carbonyl group for the Mannich reaction [20], aminomethylation did not occur at this position despite the use of reagent excess. Encouraged by the initial result, other secondary amines (bis(2-methoxyethyl)amine, piperidine and morpholine) were also tested (Scheme 1), but in addition to a significant decrease in the reaction rate (24 h, reflux or 45 min, MW), the desired products proved to be mixtures of the corresponding C2 (1b− d) and C4 regioisomers (1′b− d), to which the latter contributes to a much greater extent than in the reaction with dimethylamine. Nevertheless, in all cases, the results supported the formation of the C2-substituted derivative (1) as the major product, contrary to previous findings for bicyclic phenols [12], although the ratio of 1b and 1'b was already close to 1:1 when bis(2-methoxyethyl)amine was used.
Since the C2 and C4 products (1 and 1′) obtained by the reactions of amines other than dimethylamine were difficult to separate by column chromatography, an alternative route for the preparation of 2-aminomethylated derivatives of E had to be developed instead of the direct Mannich-type synthesis. For this purpose, 1a was used as the starting material, which could be efficiently synthesized regioselectively even in larger quantities. Next, 1a was subjected to salt formation with methyl iodide (MeI) in diethyl ether (Et 2 O) / acetonitrile (MeCN) = 4:1 to obtain tetraalkylammonium iodide (2) in a nearly quantitative yield by stirring at room temperature (r.t.) for 24 h (Scheme 2). Since compound 2 proved to be light-sensitive [21], the reaction was performed in dark. The use of a solvent mixture was of great importance: while MeCN aided in the dissolution of the starting material (1a), diethyl ether promoted the precipitation of the salt (2) formed. After dilution of the reaction mixture with ether and subsequent filtration, the quaternary ammonium iodide 2 was converted to Mannich bases (1b− j) with different dialkyl and alicyclic secondary amines (Scheme 2). The transformations were carried out in MeCN under reflux in the presence of a base, which induced the in situ formation of an o-QM intermediate (3) by β-elimination [11]. 1,4-aza-Michael addition of the amines applied resulted in the desired 2-aminomethylated products in good yields (Table 1). It is important to note that organic bases, such as 1,8-diazabicyclo(5.4.0) undec-7-ene (DBU) or tetramethylethylenediamine (TMEDA) effectively catalysed the reaction to occur towards the formation of 2-aminomethylated E derivatives (1b− j), while in the presence of KOH or NaOH a considerable amount of by-product, presumably 2-hydroxymethyl-E was formed beside the desired products. Further advantage of this protocol that there was no need to protect the COOH group of proline during the synthesis of 1j contrarily to the three-component Mannich-type reaction [21]. Additional novel compounds with higher polarity than 1 were also synthesized either by stereoselective reduction of 1a− j to afford 2-aminomethylated E2 derivatives (4a− j) or by reductive amination of the C17-carbonyl group in 1a− c, 1h and 1i with benzylamine (BnNH 2 ) in the presence of sodium cyanoborohydride to provide some 17β-benzylamino derivatives (5a¡c, 5h, 5i) (Scheme 2, Table 1). It is worth mentioning that the regioselective functionalization of E2 and its 17β-benzylamino analogue (easily accessible from E by reductive amination) at C2 position with different aminomethyl groups is also possible by the described method, but the early-stage derivatization of E and subsequent conversions of C17 is more practical due to easier handling of the less polar E-based Mannich bases.
The structures of all synthesized compounds were confirmed by 1 H and 13 C NMR spectroscopy as well as by MS spectrometry (Supplementary Material). The chemical shifts, splitting, and integrals of the signals on the 1 H NMR spectra of the products clearly supported the structures. For all derivatives, the substitution at C2 position is evidenced by a singlet of two non-coupled protons (1-H and 4-H)  H) would be observed in the same range. In case of the 17β-benzylamino derivatives (5), the signals between 7.2 and 7.4 ppm are indicative for the incorporation of the aromatic ring, while the diastereotopic CH 2 protons belonging to the Bn group give a complex signal at around 3.7 ppm. Proton peaks characteristic of the various aminomethyl substituents incorporated at C2 can also be identified in the spectra. The 13 C NMR spectra, which were recorded using J-MOD pulse sequence, were also in agreement with the proposed structures.

Characterization of in silico and in vitro physicochemical properties
Following the classical medicinal chemistry approach, physicochemical characterization of E (1), E2 (4) and 17β-benzylamino derivatives (5) was the first step to select possible candidates with lower attrition risks for drug adverse and toxicological effects. Regarding the rules of thumb for clinical candidates (e.g. Lipinski's rule of five: Ro5) [22,23] the most relevant physicochemical parameters (molecular weight: Mw, pK a , logP/D 7.4 , topological polar surface area: TPSA) of synthesized compounds were predicted, kinetic solubility (in phosphate buffer saline, pH 7.4) and intestinal-specific PAMPA (pH 6.5 (donor) -pH 7.4 (acceptor)) permeability were also experimentally determined. Reviewing the in silico data summarized in Table 2, Ro5 violations were identified only for 17β-benzylamino derivatives. The lipophilicity (logP > 5) for 5a, 5c, and 5i, the molecular weight (Mw > 500) for 5b, and both parameters for 5h exceeded the thresholds determined by Ro5. In terms of lipophilicity, it is worth highlighting the E derivatives 4c and 4e, which have relatively higher predicted logP values (4.15 and 4.21, respectively) and the logD 7.4 values of the two proline derivatives (1j, 4j: 0.33, 0.64) are lower (explained by the presence of ionized COOH group at pH 7.4) than the optimal lipophilicity range for oral absorption.
The acid-base character of the test compounds was also characterized by estimating the pK a of the strongest functional groups. With the exception of proline derivatives (where the pK a,acid values of COOH group are 2.3), the acidic character of the compounds was determined by the aromatic OH function (pK a,acid values between 10.1 and 10.6). Regarding the basic character, a much more diverse picture emerges from the review of pK a,base values (Table 2), which is also closely related to the synthetic strategy of the new E and E2 derivatives. In the case of 2aminomethylated E and E2, the N-centered mono-and dibasic groups, while in the case of the 17β-benzylamino derivatives, an additional Ncentered basic group determines the basic character of the compounds. It is important to note that the reference compound (2ME2) is a moderately lipophilic monoprotic acid as opposed to the novel amphoteric (in some case amphiphilic) E and E2 Mannich base derivatives.
In correlation with the steroid structure, the TPSA (topological polar surface area) values of the test compounds are relatively low and in the narrow range. This range of polarity is suitable for penetration through cell membranes as well as the blood-brain barrier. Compounds in this polarity range tend to have good permeability across cell membranes (TPSA < 140 Å 2 ) [24] in general and across blood-brain barrier (TPSA < 90 Å 2 ) [25] as well. In accordance with ADME screening in the early stage of drug discovery, we investigated the kinetic solubility the intestinal-specific PAMPA permeability relevant to oral absorption of the new E and E2 derivatives and of 2ME2 as a reference compound.
Considering the heterogeneity of the experimental physicochemical data, both kinetic solubility and permeability data were evaluated according to a three-level classification system (see footnote of Table 2). Based on the class categories obtained in the experimental study and the in silico parameters, we were able to identify primary and secondary candidates. Compounds that did not identify Ro5 violation and were in the good (medium grey) category in both in vitro assays were evaluated as primary candidates. For secondary candidates, moderate (light grey) classification was sufficient based on experimental results. It can be seen from Table 2 that a total of five compounds were selected for E derivatives (primary candidates: 1f, 1g, secondary candidates: 1a, 1c, 1i) and three compounds for E2 derivatives (primary candidate: 4i, secondary candidates: 4c, 4g). 17β-benzylamino derivatives were excluded due to Ro5 violation. It is also important to note that the reference compound (2ME2) would also did not meet the established criteria system due to its poor kinetic solubility (54 μM). Summarizing the results of physicochemical pre-screening, N-substituted cyclic base piperidine (c), N-ethylated piperazine (g) and pyrrolidine (i) derivatives satisfy the general medicinal chemistry rules.

Antiproliferative assay
In vitro antiproliferative capacity of twenty-seven newly synthesised, 2-aminomethylated E analogues (1a-5i) was also evaluated on a panel of human adherent cancer cell lines. The compounds were tested against cervical (HeLa), ovarian (A2780) and breast (MDA-MB-231) carcinoma cell lines. Additionally, their tumour selectivity was also determined by using non-cancerous mouse embryo fibroblast (NIH/3T3) cells.
2-Methoxyestradiol (2ME2) as positive control in our study is known for its highly effective antiproliferative activity on several cancer cell lines from different tissue origins [27]. In our panel for antiproliferation screen, 2ME2 similarly displayed low IC 50 values (0.70-2.5 μM) on the utilized gynecological cancer cell lines (Supplementary Material,  Table S1), in addition, it was the most effective against A2780 cells and the less effective against MDA-MB-231 breast carcinoma cells. However, 2ME2 inhibited the proliferation of non-cancerous fibroblast cells at the Table 1 Yields of E-derived Mannich bases.
-a After chromatographic purification. b Yield obtained from E by direct Mannich reaction (Scheme 1). c Boc = tert-butyloxycarbonyl. d Yield calculated from 1a after 3 steps (salt formation, preparation of 1e and deprotection with trifluoroacetic acid (TFA) in dichloromethane (DCM) for 24 h at r.t. same concentration range. Consequently, 2ME2 might not be considered as a tumour-selective agent in our experimental circumstances. Similar to 2ME2, amongst the newly synthesized derivatives, 17βbenzylamino derivatives (5a-i) also exerted high antiproliferative effectiveness on the investigated cancer cell lines (IC 50 = 1.07-5.09 μM), however the most susceptible cell line cannot be determined due to the different effectivity order of each analogue in this group. The IC 50 values of these compounds on non-cancerous fibroblast cells were between 1.15 and 5.32 μM, which demonstrated their non-selective antiproliferative effect towards cancer cells. Considering the promiscuous properties of the 5a-i derivatives and their associated increased lipophilicity, hereinafter, the antiproliferative activity of the test compounds was characterized by their lipophilic efficiency (LiPE = pIC 50 -logP) [23,28] in order to highlight their selectivity and to reduce potential adverse effects (Table 3). Therefore, to compare the antiproliferative effects, we used pIC 50 values (pIC 50,2ME2 = 5.60-6.15, pIC 50,5a-i = 5.27-5.94) and paired them with LiPE to use a three-class system to select potentially selective and drug-like derivatives of novel E and E2 compounds. The individual effect of data pairs was divided into a classification system as follows: medium grey (elevated effect and low promiscuity risk): pIC 50 ≥ 5 and LiPE ≥ 1.5 or pIC 50 ≥ 4.5 and LiPE ≥ 2.0, light grey (moderate effect and moderate promiscuity risk): 5 > pIC 50 ≥ 4.5 and 2.0 > LiPE ≥ 1.0, dark grey (elevated effect and high promiscuity risk): pIC 50 ≥ 5 but LiPE < 1.0.
The E2 derivatives (4a-k) displayed more diverse antiproliferative effect against the selected gynecological cancer cell lines. Since four analogues from this group (4a, 4b, 4i and 4e) possess pIC 50 values higher than 5.0, HeLa cells can be considered as the most susceptible cell line. Only 4e was able to inhibit ovarian cancer cell (A2780) proliferation with a pIC 50 value higher than 5.0. Three compounds (4a, 4b and 4h) exerted moderate or weak antiproliferative activity against this cell line. The novel E2 derivatives, except 4b, were not able to inhibit substantially the proliferation of hormone-insensitive breast cancer cell line, MDA-MB-231. On the other hand, 4b demonstrated good or moderate antiproliferative activity against all investigated cancer cell lines. Moreover, 4b inhibited the proliferation of non-cancerous fibroblast cells at 1.6-3.2 times higher concentration compared to its IC 50 values on cancerous gynecological cell lines. Among the remaining 2-aminomethylated E2 compounds, 4a exerted the most selective antiproliferative effect, because it possesses 15 times higher IC 50 value on NIH/3T3 cells compared to that on HeLa cells. 4i and 4g can be considered as tumourspecific compounds, because they were able to suppress HeLa cell proliferation only. Regarding the antiproliferative effects (pIC 50 ) and the LiPE values of our classification system (Table 3), the primary hits (medium grey) were 4a and 4i, and 4b and 4g as the secondary hits (light grey) showed good selectivity on cancerous cell lines. Compound 4e can be also considered as a potential lead molecule, although the LiPE values are lower than 1.0 (due to increased logP: 4.21), but it also showed an increased antiproliferative effect selectively (pIC 50,HeLa/A2780 > 5.0 and pIC 50,NIH/3T3 = 4.73).
The newly synthesized E derivatives (1a-k) displayed lower antiproliferative activities against the investigated cancer cells compared to the compounds of the other two scaffolds. 1a, 1b and 1c demonstrated moderate antiproliferative activity (pIC 50 = 4.70-4.98) on cervical HeLa cells. Five compounds (1a, 1b, 1c, 1e and 1i) exerted moderate or weak antiproliferative activity (pIC 50 = 4.49-4.81) against ovarian cancer cell Table 2 In silico and experimental physicochemical parameters of E-derived Mannich bases. a Lipinski's Ro5 violation for molecular weight (Mw > 500) or (logP > 5), b pK a values calculated for the strongest acidic or basic moiety, c decreased predicted pK a due to the proline COOH group, d Classification system for kinetic solubility, kin. sol. (μM): good (medium grey ≥ 300), moderate (300 > light grey ≥ 100), poor (dark grey < 100), e Classification system for PAMPA permeability, P e (10 − 6 cm/s): good (medium grey ≥ 20), moderate (20 > light grey ≥ 15), poor (dark grey < 15), f ND: not determined (the compound cannot be detected in the acceptor side), g increased lipid partition (MR% > 80), indicating a strong interaction between the active substance and the PAMPA artificial lipid membrane. h primary candidates (medium grey): no Ro5 violation and belongs to the good class based on experimental data, secondary candidates (light grey): no Ro5 violation and belongs to at least moderate class based on experimental data. i Calculated parameters using ACD/Labs Percepta software package [26].

Table 3
In vitro antiproliferative activity (pIC 50 ) and lipophilic efficiency (LiPE) of E-derived Mannich bases. Classification system for antiproliferative effect: medium grey (elevated effect and low promiscuity risk): pIC 50 ≥ 5 and LiPE ≥ 1.5 or pIC 50 ≥ 4.5 and LiPE ≥ 2.0, light line (A2780). Similar to the E2 derivative 4b, bis(2-methoxyethyl)amino derivative 1b (modified at C2 position) was able to inhibit MDA-MB-231 cell proliferation to a large extent. The result indicates that regardless of the chemical environment of the C17 position, the presence of the bis(2methoxyethyl)amino group at the C2 position may be responsible for this selective antiproliferative effect. It can be supported by the experience that among the E and E2 derivatives, no other substituent at position C2 provided similar antiproliferative pattern. In this group of E derivatives, 1i and 1e can be considered as tumour specific compounds due to their selective inhibition on proliferation of ovarian cancer cells. Moreover, 1e displayed the highest selectivity toward cancer cells, as a difference four times higher between the IC 50 values calculated for A2780 and fibroblast cells can be observed.
Regarding the potential structureantiproliferative activity relations for C2-substituent of E and E2 compounds, a N-containing branched-chain acyclic derivatives (1a ↔ 4a and 1b ↔ 4b) might be considered to possess more favourable antiproliferative characteristics against cervical and ovarian cancer cell lines than a corresponding Ncontaining cyclic analogues (1d ↔ 4d, 1j ↔ 4j, 1k ↔ 4k). Another prominent QSAR pattern can be identified for piperazine derivatives of E2, where the antiproliferative effect on HeLa cells increases with increasing lipophilicity of non-substituted (4k) and N-aliphatic (4e-h) analogues (logP: 2.92, 3.05, 3.31, 3.57, 4.21 and pIC 50 : -, -, 4.68, 4.46, 5.17, respectively). This observation supports the consideration of LiPE for compound selection in our medicinal chemistry project as well. Furthermore, C2-pyrrolidine derivatives 1i and 4i represented cancerspecific antiproliferative activity with moderate or good pIC 50 values against A2780 and HeLa cells, respectively. Overall, based on the pIC 50 -LiPE classification system, among E derivatives, 1b was identified as a primary and 1i was selected as a secondary hit.
Finally, to collect the primary and secondary candidates of in silico / in vitro physicochemical screening and the primary and secondary hits of the antiproliferative assay panel, the most advantageous compound is the E2 derivative 4i, but 1i and 4g may also be preferred as a secondary candidate for in vivo pharmacology studies. Although, focusing on the results of antiproliferative studies, 4a and 4b can be also served as in vivo preclinical candidates due to their HeLa-selective activity.

Conclusions
In summary, regioselective synthesis of novel A-ring-modified derivatives in the estrone series was efficiently accomplished. Phenolic Mannich bases substituted with various tertiary aminomethyl groups at C2 position were obtained by aza-Michael addition of different secondary amines to the o-QM intermediate generated in situ from quaternized 2-dimethylaminomethyl-E. Additional 17β-hydroxy and 17βbenzylamino analogues were also prepared to compare physicochemical properties and pharmacological activities.
Two independent approaches were used to characterize the compounds, considering the rules of thumb for drug-likeness related to pharmacokinetic behaviour and the risk of promiscuity in the evaluation of the antiproliferative effect. Based on the rigorous classification system, 17β-benzylamino derivatives (5) were excluded due to their increased molecular weight and lipophilicity. The result of the primary in silico screen is supported by the fact that the 5 derivatives, although they have an enhanced antiproliferative activity, however, their effect is not selective. Promiscuous behaviour of this scaffold (5) was also supported by their low LiPE values (< 1). The selectivity-LiPE relationship was also identified for the piperazine derivatives formed at C2 position. In this closed chemical space, the positive correlation between lipophilicity and antiproliferative effect was confirmed by our observations. Overall, only one estrone (E: 1i) and two estradiol (E2: 4i, 4g) derivatives complied with our rules on in silico drug-likeness, on in vitro drug absorption and on selective antiproliferative activity. The result is consistent with the general finding of antiproliferative screen, which showed that E2 derivatives exhibited more enhanced effect than Ederived compounds. However, the role of the substituent at C2 position in the antiproliferative effect was also observed, given that the pyrrolidine derivatives of E and E2 satisfy the two-stage classification system. It is also important to highlight that the reference compound, 2ME2 has a substantial antiproliferative effect pIC 50 (> 5.6) with optimal LiPE (> 2.2) values. However, similar to 5 derivatives, its effect is not selective, and it has poor aqueous solubility. In contrast, 4g, 4i and 1i derivatives may be advantageous clinical candidates in the treatment of cervical and ovarian carcinoma due to their more favourable physicochemical and selective antiproliferative properties.

Materials and methods
Chemicals, reagents, and solvents were purchased from commercial suppliers (Sigma-Aldrich and Alfa Aesar) and used without further purification. Reactions under MW irradiation were carried out with a CEM Corporation Focused Microwave System, Model Discover SP. Melting points (Mp) were determined on an SRS Optimelt digital apparatus and are uncorrected. The transformations were monitored by TLC using 0.25 mm thick Kieselgel-G plates (Si 254F, Merck). Compound spots were detected by spraying with 5 % phosphomolybdic acid in 50 % aqueous phosphoric acid. Flash chromatographic purifications were carried out on silica gel 60, 40-63 μm (Merck). All eluent and solvent system compositions are given in volume percent (v/v%). NMR spectra were recorded with a Bruker DRX 500 instrument at room temperature in CDCl 3 or DMSO-d 6 using residual solvent signals as an internal reference. Chemical shifts are reported in ppm (δ scale), and coupling constants (J) are given in Hz. Multiplicities of the 1 H signals are indicated as a singlet (s), a doublet (d), doublet of doublets (dd), a triplet (t) or a multiplet (m). 13 C NMR spectra are 1 H-decoupled and the J-MOD pulse sequence was used for multiplicity editing. In this spin-echo type experiment, the signal intensity is modulated by the different coupling constants J of carbons depending on the number of attached protons. Both protonated and unprotonated carbons can be detected (CH 3 and CH carbons appear as positive signals, while CH 2 and C carbons as negative signals). Elemental analysis data were obtained with a Perkin Elmer CHN analyzer model 2400 and FT-IR spectra were recorded on a FT/IR-4700 spectrometer (Jasco) using ATR. Automated flow injection analyses were performed by using an HPLC/MSD system. The system comprised an Agilent 1100 micro vacuum degasser, a quaternary pump, a micro-well plate autoinjector and a 1946A MSD equipped with an electrospray ion (ESI) source operated in positive ion mode. The ESI parameters were: nebulizing gas N 2 , at 35 psi; drying gas N2, at 350

General procedure for the synthesis of E-derived Mannich bases (1b− j)
Compound 1a (327 mg, 1.00 mmol) and MeI (0.6 mL, 10 equiv.) were dissolved in Et 2 O/ MeCN = 4:1 and the mixture was stirred in dark at room temperature for 24 h. After TLC control using MeOH / EtOAc = 30:70 as eluent, the suspension was diluted with diethyl ether and the quaternary ammonium iodide salt (2) was filtered off and dried. The salt was dissolved in MeCN (20 mL), then secondary amine (2 mmol) and DBU (0.30 mL, 2 equiv.) were added. The mixture was kept at reflux temperature for 1 h until complete conversion (TLC monitoring). The pale yellow homogeneous mixture was poured onto ice-cold water and the precipitate formed was filtered off, washed with cold water and dried. The crude product was purified by column chromatography.

General procedure for the synthesis of E2-derived Mannich bases (4a− k)
E-based Mannich base (1a− k, 0.50 mmol) was dissolved in EtOH (10 mL) and NaBH 4 (45 mg, 1.20 mmol) was added. The solution was stirred at room temperature for 2 h, then poured into water and neutralized with diluted HCl to decompose the excess of the reagent. After adjusting the pH to 8 with 1 N KOH solution, the resulting precipitate was filtered, washed with water, dried, and purified by flash chromatography.

General procedure for the synthesis of 17β-benzylamino Mannich bases (5a− c, 5h, 5i) by reductive amination
To a stirred solution of E-based Mannich base (1a− c, 1 h, 1i, 0.50 mmol) in MeOH (10 mL), benzylamine (0.22 mL, 4 equiv.) and NaCNBH 3 (47 mg, 1.75 mmol) was added, and the pH was adjusted to 6 by the addition of glacial acetic acid. The solution was stirred at room temperature for 16− 24 h, then poured into NaHCO 3 solution and extracted with EtOAc (2 × 10 mL). The combined organic layers were washed with water, dried over anhydrous Na 2 SO 4 and evaporated in vacuo. The resulting crude product was purified by flash chromatography.  maintenance was executed by sub-culturing in 75 cm 2 cell culture flasks. Rapidly growing cells were used for cell viability assay.

MTT assay for cell viability
The MTT assay was employed to determine the extent of inhibition of cell proliferation by the test compounds. Exponentially growing HeLa, A2780, MDA-MB-231 and NIH/3T3 cells were seeded into 96-well plates (5 × 10 3 cells per well in 100 μL of media) and incubated for 24 h to allow for cell attachment. First, cells were treated with 10 and 30 μmol concentrations of test compounds formulated in 100 μL of media and incubated for 72 h in a screening experiment. The cells in the control group were treated only with media. After 72 h of incubation, 20 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT-reagent, 5 mg / mL in PBS; Duchefa Biochemie BV, Haarlem, The Netherlands) was added and incubated for 4 h at 37 C in humidified 5 % CO 2 . The medium was then aspirated and 100 μL of DMSO added to dissolve formazan crystals produced by the mitochondrial enzymes of intact cells, followed by 30 min of mechanical shaking. Absorbance measurements were subsequently taken using a microplate UV-VIS reader (SPECTROstar Nano, BMG Labtech GmbH, Offenburg, Germany) at 545 nm, followed by calculation for percentage cell viability [29] μmol. The IC 50 values (the concentration at which the extent of cell proliferation was half that of the untreated control) were calculated using sigmoidal dose response curves fitted to the determined data by means of GraphPad Prism 5.01 (GraphPad Software, San Diego, CA, USA). All in vitro experiments were carried out twice with five parallels. Stock solutions of the tested substances (10 mM) were prepared in DMSO. The highest DMSO content of the medium (0.3 %) did not have any substantial effect on cell proliferation.

In vitro physicochemical test methods
Kinetic solubility was tested with 5 % DMSO as cosolvent in phosphate buffered saline (PBS, 0.01 M phosphate buffer pH 7.4, 0.138 M NaCl, 0.0027 M KCl) at room temperature for 2 h using 96-well Multi-Screen HTS-PCF Filter Plates (Merck Millipore). Target concentration was 500 μM for every investigated compounds. After the incubation period all solutions were filtered by MultiScreen HTS Vacuum Manifold (Merck Millipore). Concentration of filtrates were determined by spectrophotometric detection (AUC 250-500nm ) using Multiskan FC Microplate Photometer (Thermo Scientific).
The intestinal specific PAMPA investigation was based on a 'in house developed' method. Positive and negative control compounds (carbamazepine: 28.7 ± 1.5⋅10 − 6 cm/s and indomethacine and 2.5 ± 0.2⋅10 − 6 cm/s, respectively) were used in every individual PAMPA investigation. Briefly, a 96-well acceptor plate and a 96-well filter plate were assembled into a sandwich. The hydrophobic filter material of the 96 well filter plate was coated by the mixture of phosphatidylcholine / cholesterol = 2:1 (5 μL, 4 % (w/v) in n-dodecan). Subsequently, the acceptor wells at the bottom of the sandwich were filled with 300 μL of 10 mM PBS solution adjusted to pH 7.4. The donor wells at the top of the sandwich were hydrated with 150 μL of test compound solution. The test compound solutions were prepared in 100 times dilution. Firstly, a 10 mM stock solution was prepared from the corresponding compound in DMSO, than it was diluted with PBS (pH 6.5) to get 5 % mixture followed by filtration through a MultiScreen Solubility filter plate. The resulting sandwich was then incubated at 37 • C for 4 h. After the incubation, PAMPA sandwich plates were separated and compound concentrations in starting, donor and acceptor solutions were measured similar to kinetic solubility method. The effective permeability and membrane retention of drugs were calculated using the following equation [30]:

.303
A × (t − τ ss ) •lg where P e is the effective permeability coefficient (cm/s), A is the filter area (0.3 cm 2 ), V D and V A are the volumes in the donor (0.15 cm 3 ) and acceptor phase (0.3 cm 3 ), t is the incubation time (s), τ SS is the time to reach steady-state (s), C D (t) is the concentration of the compound in the donor phase at time point t (mol/cm 3 ), C D (0) is the concentration of the compound in the donor phase at time point zero (mol/cm 3 ), r v is the aqueous compartment volume ratio (V D /V A ), MR is the membrane retention factor, defined as: