Synthesis, biological evaluation, and modeling studies of 1,3-disubstituted cyclobutane-containing analogs of combretastatin A4

Abstract

With the aim of circumventing the adverse cis/trans-isomerization of combretastatin A4 (CA4), a naturally occurring tumor-vascular disrupting agent, we designed novel CA4 analogs bearing 1,3-cyclobutane moiety instead of the cis-stilbene unit of the parent compound. The corresponding cis and trans cyclobutane-containing derivatives were prepared as pure diastereomers. The structure of the target compounds was confirmed by X-ray diffraction study. The title compounds were evaluated for their cytotoxic properties in human cancer cell lines HepG2 (hepatocarcinoma) and SK-N-DZ (neuroblastoma), and the overall activity was found in micromolar range. Molecular docking studies and molecular dynamics simulation within the colchicine binding site of tubulin were in good agreement with the obtained cytotoxicity data.

Graphical Abstract

1. Introduction

Since 1982, when highly oxygenated stilbene derivatives were isolated from the bark of the African willow tree Combretum caffrum [1], these compounds have come into the spotlight of the medicinal interest as tumor-vascular disrupting agents (VDAs). In particular, the isolated compounds, which later received the common name combretastatins, were found to cause apoptotic tumor cell death via selective binding to tubulin at the colchicine binding site. This prevents cancer cells from producing microtubules and induces cell cycle arrest at the transition of meta- to ana-phase [2,3]. Combretastatin A4 (CA4) (Figure 1) is the most studied member of the combretastatin family [4,5] and is among the most potent cytotoxic compounds known to date [6,7].

Figure 1.

Combretastatin A4 and investigational drugs based on its structure

The framework of CA4 is characterized by three principal features: a 3,4,5-trimethoxy-substituted benzene ring A, a ring B containing C3’-OH and C4’-OMe substituents, and –CH=CH– bridge between the two rings that provides structural rigidity. Importantly, the cis orientation of the double bond, the presence of methoxy substituents at the ring A and free hydroxyl functionality or other hydrogen bonding donor at the ring B were found to be essential for binding to tubulin and possessing a high level of cytotoxicity [8,9].

A water-soluble phosphate prodrug of CA4, fosbretabulin, completed multiple clinical trials for the treatment of patients with anaplastic thyroid carcinoma [10] and other advanced cancers [11] in 2017. At the same time, the development of another CA4 analogue, the serine amide ombrabulin [12], was discontinued in 2013 after phase III clinical trials in patients with advanced soft tissues sarcoma.

Since the discovery of the antitumor profile of CA4, numerous structure-like derivatives have been prepared and subjected to biological activity testing [13–16]. However, a great challenge which the researchers faced was isomerization of such compounds into their inactive trans form upon storage and administration. Therefore, structural modifications were undertaken aimed at locking the compounds in the desired conformation by replacement of the double bond with carbo- and heterocyclic moieties. In this regard, a series of works were published describing synthesis and biological evaluation of CA4-like derivatives with fused heterocyclic rings instead of the cis-stilbene unit [17–24]. At the same time, incorporation of the heterocyclic fragment altered overall polarity of compound that adversely affected the binding to tubulin. In this view, using small cycloalkane moieties as the double bond replacements is of greater interest as it allows fine-tuning of the angle and position of the aromatic units without changing the overall polarity.

Taking into account that cyclopropane is generally considered as an alkene bioisostere, the cyclopropyl CA4 analogs 1 (Figure 2) have been synthesized and evaluated by several groups [25–27]. Thus, racemic derivative 1 (with no stereochemistry specified) displayed submicromolar cytotoxicity towards HeLa (cervical adenocarcinoma) and MCF-7 (breast adenocarcinoma) cells, although was much less active than the parent compound [25]. Moreover, the (1R,2S)-isomer 1b was as potent as CA4 when assessed by the ability to inhibit tubulin polymerization (IC₅₀ = 0.3 μM) (also much less potent in the cytotoxicity assays) [26].

Figure 2.

Analogs of combretastatin A4 containing small cycloalkane rings.

In this work, we present results on the synthesis of novel CA4 analogs 2a and 2b bearing a 1,3-disubstituted cyclobutane unit instead of the double bond (Figure 2), their biological evaluation, as well as analysis of binding mode to tubulin according to molecular docking and molecular dynamics studies.

2. Results and Discussion

2.1. Chemistry

Retrosynthetic analysis of the target compounds 2 led to the appropriately substituted cyclobutanone 3 as a key intermediate (Scheme 1). The ketone 3 can be obtained via the [2+2] cycloaddition of the corresponding styrene which in turn is prepared by the Wittig olefination of a benzaldehyde derivative. Thereby, we opted to use isovaniline as the starting compound for the synthesis of 2.

Scheme 1.

Retrosynthetic analysis of cyclobutyl analogs of CA4.

Notably, cyclobutane derivatives 2 are achiral due to the presence of a symmetry plane; this structural feature should simplify the synthetic scheme significantly.

Following the selected strategy, the target cyclobutane derivatives 2 were synthesized starting from 3-benzyloxy-4-me-toxystyrene (3), prepared from isovaniline in high overall yield (78%) using the previously published method [28]. Briefly, isovaniline was protected with benzyl bromide, and the resulting intermediate was subjected to the Wittig reaction (Ph₃PMe⁺Br^–, t-BuOK). The construction of the cyclobutane ring was achieved through [2+2]-cycloaddition with ketene iminium salt generated in situ from dimethyl acetamide (DMA) and trifluoromethanesulfonic anhydride in the presence of collidine [29] (Scheme 2). The resulting cyclobutanone 5 was introduced into the nucleophilic addition reaction with 3,4,5-trimethoxyphenyllithium (prepared by lithiation of 5-bromo-1,2,3-trimethoxybenzene with n-butyllithium) affording properly substituted alcohol 6, which was converted into a mixture of the target compounds through palladium-catalyzed hydrogenolysis. Finally, the HPLC separation of the racemic mixture afforded the target diastereomerically pure cyclobutane-containing analogs of combretastatin 2a and 2b.

Scheme 2.

Synthesis of cyclobutyl analogs 2 of combretastatin A4.

The structure and the configuration of the target compounds were confirmed by X-ray diffraction studies (Figures 3 and 4).

Figure 3.

Molecular structure of the compound 2a according to X-ray diffraction study.

Figure 4.

Molecular structure of the compound 2b according to X-ray diffraction study.

2.2. Biological evaluation

The target compounds 2a and 2b were evaluated for in vitro cytotoxicity in human cancerous cell lines HepG2 (hepatocarcinoma) and SK-N-DZ (neuroblastoma) using a resazurin-based cell viability assay [30–32]. Combretastatin A4 and doxorubicin were included as the control substances. Results of these biological activity studies were represented as dose-response curves and estimated half-maximal inhibitory concentration (IC₅₀) (Table 1, Figures 5 and 6).

Table 1.

IC₅₀ values (mean; n=4) of reference and studied compounds to HepG2 and SK-N-DZ cell lines.

Cells	IC50 (μM)
	CA4	doxorubicin	2a	2b
HepG2	0.009	0.36	7	≈13
SK-N-DZ	0.005	0.4	6	1.7

Figure 5.

Dose-response curves for the compounds evaluated in resazurin-based cell viability assay against HepG2 cells (mean ± SD, n = 4).

Figure 6.

Dose-response curves for the compounds evaluated in resazurin-based cell viability assay against SK-N-DZ cells (mean ± SD, n = 4).

Unfortunately, the compounds 2a and 2b exhibited weak cytotoxicity (micromolar range) against both tested cell lines as compared to CA4 (nanomolar range) – a situation similar to that observed for the cyclopropane analogs 1. This may suggest a loss of specific tubulin-blocking activity of 2 while retaining non-specific cytotoxicity. In order to confirm or refute this assumption and simulate the difference in the binding modes of the tested compounds with tubulin, modeling studies were performed as described hereafter.

2.3. Molecular docking

The molecular docking of compounds 1a and 1b was reported in the previous study; importance of the compound stereochemistry was outlined therein [28]. We have performed the docking with all isomers of 1 (1a–1d), as well as with 2a and 2b. The compounds were docked into the binding interface of tubulin protein α/β-chains (also called colchicine binding pocket). Tubulin – colchicine complex (PDB entry 1SA0) [27] was used as a template for molecular docking of the studied compounds. In order to validate molecular docking procedure colchicine was extracted from binding site and then re-docked. RMSD (Root-mean-square deviation) between best-docked poses and crystal structure was less than 0.5 Å. Based on the structure of tubulin – colchicine complex (1SA0) and our results of the molecular docking, the residues involved into the most important interactions were identified, namely, the T5 loop (Thr 179, Ala 180, Val 181, and Ser 178), the T7 loop (Glu 245 and Leu 246), and Cys 241, Met 259 [33–35]. Therefore, according to our binding model, the potential ligand should be located close to Cys 239 (possibly forming a hydrogen bond with this residue) and Met 259; in addition to that, a hydrogen bond should be formed with one of the residues of the T5 loop. Figure 7 shows results of molecular docking for the compounds 2a and 2b.

Figure 7.

Results of molecular docking for the compounds 2a and 2b (tubulin is shown in grey, T7 loop – in violet, T5 loop – in wheat, 2a – in yellow, and 2b – in green).

The aforementioned complexes are illustrative examples which contain typical interactions characteristic to our model. In particular, the hydrogen bonds were formed between Cys 239 and one of the methoxy groups of the ring A for both isomers of 2. An additional interaction between Cys 239 and the ring A included tight π-stacking. For cis isomer 2a, a significant steric effect was observed between the cyclobutane ring and the T7 loop. Nevertheless, stacking interaction was possible between the hydrophobic side chain of Leu 246 and the ring B. An additional interaction with the T5 loop (Thr 179) was also observed. In the case of trans-2, the free OH group could make a number of hydrogen bonds with T7 loop and Lys 252. In addition, a possibility of π-cation interaction with Lys 252 also existed. Additional interactions were also possible with Asn 101, Asn 256, Ser 178, and Gln 11.

The molecular docking results suggested that all the docked compounds occupied almost identical space in the colchicine binding pocket of tubulin. The 3,4,5-trimethoxyphenyl ring A was located closely to the Cys 239 residue; their interactions were more or less similar for all the compounds studied. For the ring B parts of the molecules, some binding differences were observed.

2.4. Molecular dynamic (MD) simulation

To obtain more precise results, 25 ns molecular dynamics (MD) simulations were performed for the complexes obtained after the molecular docking. The results are shown in Figure 8.

Figure 8.

MD simulation results for: (A) 1a; (B) 1d; (C) 1b; (D) 1c (E) 2a; (F) 2b (tubulin is shown in grey, all compounds – in yellow (hydrogen bonds – in blue), not final but common conformation during simulation of tubulin and compounds – in cyan (in this case, hydrogen bonds – in yellow)).

Characterization of cis isomers.

Analysis of the tubulin – ligand complexes for the cis isomers 1a, 1b, and 2a showed that in all cases, the hydrogen bonds with the T5 loop were observed (although their stability was different, see below). The rings A were located in nearly the same part of the binding pocked for all three compounds, so that π-stacking interaction with Cys 239 and Met 257, or even hydrogen bonding interaction with Cys 239 could be observed. In addition to that, π-cation interactions of the rings B with Lys 252/300 were also possible.

While the root-mean-square deviation (RMSD) of tubulin was almost the same in both cases (Figure 9A, C), the RMSD values of the ligands during the MD simulation were rather different (Figure 9B, D). Thus, colchicine (used as a reference) and 1a showed considerable conformational rigidity (0.5 Å and 0.7 Å, respectively), whereas 1b and 2a were much more flexible (1.1 Å and 1.2 Å, respectively). This might be one of the reasons behind the small activity of 2a observed experimentally.

Figure 9.

RMSD trajectories of tubulin and investigated compounds. A – RMSD of tubulin in complex with compounds in cis conformation. B – RMSD of tubulin in complex with compounds in trans conformation. C – RMSD of compounds in cis conformation. D – RMSD of compounds in trans conformation. Green – compound 1a or 1d, brown – 1b or 1c, black – compound 2a or 2b, violet – colchicine.

The compound 1a formed strong hydrogen bond with Thr 179 (Figure 8A; Figure 10A, line 10). That interaction was further stabilized by formation of short-time hydrogen bonds with two adjacent amino acid residues Lys 350 and Ser 178 (Figure 10A, lines 1, 2, and 9).

Figure 10.

Hydrogen bond existence map between tubulin and compounds 1a (A), 1b (B), and 2a (C).

As in the case of 1a, 1b initially formed hydrogen bond with Thr 179; then, this bond was lost (Figure 8C; Figure 10B, line 10), and another one with Ser 178 was formed instead (Figure 8C; Figure 10B, lines 0, 1, 2). Both these interaction were additionally stabilized by short-range hydrogen bonds with Lys 350 (Figure 8C; Figure 10B, line 8). However, taking into account conformational flexibility of 1b, there is a possibility of losing any hydrogen bond interactions with the T5 loop.

In turn, compound 2a did not form such stable hydrogen bonds with the T5 loop (Figure 8C). In the first phase of the MD simulation, the interaction between hydroxyl group of the ring B and Thr 179 dominated (Figure 10B, lines 7 and 3). In the later phases, the loss of that interaction was observed. This might be partially related to higher flexibility of 2a as compared to 1a. Then, a new hydrogen bond with Ser 178 was formed (Figure 10B, line 2), which could be stabilized by short-time interactions with Lys 350 (line 6) and Thr 179 (lines 3 and 7). Finally, ashift of the T5 loop (1.54 Å) was observed, which resulted in the loss of any hydrogen bond interactions.

Characterization of trans isomers.

Unlike the case of the cis isomers, no stable hydrogen bond interactions with tubulin were observed for the trans isomers (Figure 11). For the compounds 1c and 1d, the rings B were found in a different part of the binding pocket, so that interactions with Lys 252, Asn 101, Asn 256, Gln 11 and GDP were more likely; on the contrary, 2b had similar orientation of the ring B to that of cis isomers (Figure 8B,D,F). The rings A had nearly the same position as for the cis isomers (close to Cys 239 and Met 257) (except for 1c). The RMSD of tubulin in all cases were almost the same (Figure 9B), whereas the RMSD values for the compound molecules varied considerably (Figure 9D).

Figure 11.

Hydrogen bond existence map between tubulin and compounds 1c (A), 1d (B), and 2b (C).

Unlike for other compounds, the ring B of 1c was located too far from Cys 239 (distance between the nearest atoms was ca. 8 Å). Instead of that, this part of the molecule stuck to Lys 350, Asn 256 and Met 257 (Figure 8D). Due to this arrangement, compound 1c was capable to create stronger interactions with the T5 loop (compared to 1d and 2b). In the initial phase of the MD simulation, compound 1c formed hydrogen bond with Gln 11 (Figure 8D (in cyan); Figure 11B, lines 10 and 11). Later, 1c formed hydrogen bonds with Val 181 (Figure 11B, line 5), Ser 178 (lines 3 and 4), Asn 101 (line 7), and Gln 11 (line 1). Although these interactions were not stable, they were numerous, so that the compound 1c was located near the T5 loop during the whole MD trajectory.According to the MD simulation, the compound 1d formed hydrogen bonds with Ser 178 (Figure 8B; Figure 11A, lines 0, 7 and 4) and Thr 179 (Figure 8B; Figure 11A, line 5); both interactions were not stable enough. Furthermore, in the last phases of the MD simulation, the distance between the closest atoms of compound 1d and T5 loop was about 4 Å. In addition to that, compound 1d was located near Lys 252, Leu 246, and Asn 256, and had steric interactions with these residues (Figure 8B). Also, there was a possibility of π-cationic interaction between the ring B and Lys 252.

The most stable interactions of compound 2b identified in the MD simulation were hydrogen bonds with Asn 247 (Figure 8D; Figure 11B, lines 5 and 12) and Asn 101 (Figure 8D; Figure 11B, line 1). This compound did not take part in any interactions with the T5 loop (except steric ones).

The obtained results showed that all the cis isomers were capable of interactions with the T5 loop, whereas the trans isomers had only short-time ones. This allowed assuming higher activity for cis isomers, which is indirectly confirmed by the literature data for the compound 1b. In the case of 2a and 2b, the predicted interactions with tubuline (especially with its T5 loop) were less stable as compared to those for the cyclopropane counterparts. This could be partially addressed to higher conformational flexibility of the cyclobutane derivatives. These results correlate with experimental data on the biological activity of 2a and 2b.

3. Conclusions

To address the problem of cis-trans-isomerization inherent in naturally occurring combretastatins leading to loss of biological activity, cyclobutane-containing analogs of CA4 were synthesized. The key step of the synthesis included addition of 3,4,5-trimethoxyphenylmagnesium reagent to isovaniline-derived cyclobutanone derivative, followed by catalytic deoxygenation. The target products were obtained as single diastereomers after chromatographic separation. Both cis and trans isomers demonstrated modest cytotoxicity in human cancerous cell lines HepG2 (hepatocarcinoma) and SK-N-DZ (neuroblastoma). These results were rationalized thought molecular docking studies and molecular dynamics simulation within the colchicine binding site of tubulin. The modeling studies showed that cis isomers of cyclopropane- and cyclobutane-containing analogs of CA4 demonstrate better capability to form hydrogen bonds with the T5 loop of tubulin, whereas trans isomers had only short-time interactions. For the cyclobutane derivatives, these interactions were less stable, which might be partially addressed to higher conformational flexibility of the compounds.

4. Experimental section

4.1. Chemistry

Commercially available reagents were used without purification unless otherwise stated. Unless otherwise noted, all materials were obtained from commercial sources and were used without purification. THF was dried over sodium/benzophenone. All air- and moisture-sensitive reactions were carried out under a dry argon atmosphere. Analytical thin-layer chromatography was carried out on aluminum backed plates coated with silica gel (Merck, 60, F254). ¹H and ¹³C NMR spectra were obtained on a Bruker Avance 500 instrument at 500.13 MHz and 125.76 MHz respectively, using CDCl₃ as solvent and Me₄Si as an internal standard. HRMS were obtained on Agilent 6200 TOF/6500 Q-TOF B.08.00 instrument.

3-(3-(Benzyloxy)-4-methoxyphenyl)cyclobutan-1-one (5)

(CF₃SO₂)₂O (63.17 g, 0.22 mol) was added dropwise to a stirred ice-cold solution of dimethylacetamide (16.7 g, 0.19 mol) in CH₂Cl₂ (80 mL) maintaining the temperature below 5°C. This was accompanied by the formation of a white precipitate. The mixture was allowed to stir for 30 min at 5°C and then the solution of styrene 6 (38.0 g, 0.16 mol) and 2,4,6-trimethylpyridine (27.1 g, 0.22 mol) in CH₂Cl₂ (20 mL) was added dropwise at this temperature. The resulting mixture was refluxed for 18 h, after which the volatiles were removed in vacuo. The residue was treated with water (250 mL) and CCl₄ (250 mL) and the obtained mixture was refluxed under stirring for another 18 h. The cooled mixture was divided, the water layer was extracted with CCl₄ (3×150 mL), the organic layers were combined, dried (Na₂SO₄) and evaporated in vacuo. The residue was subjected to silica gel column chromatography (hexane – EtOAc (4:1) as eluent) affording spectrally pure title compound 5 (9.10 g, 32.2 mmol, 21%) as light yellow crystals; mp = 94°C; R_f = 0.36 (hexane – EtOAc (4:1)); ¹H NMR (400 MHz CDCl₃) 7.45 (d, J 7.5 Hz, 2H, H-2′ and H-6′), 7.37 (t, J 7.4 Hz, 2H, H-3′ and H-5′), 7.31 (t, J = 7.2 Hz, 1H, H-4′), 6.88 (d, J = 8.2 Hz, 1H, H-6), 6.84 (d, J = 8.2 Hz, 1H, H-5), 6.82 (s, 1H, H-2), 5.16 (s, 2H, OCH₂), 3.89 (s, 3H, OCH₃), 3.57 (quint, J = 7.8 Hz, 1H, cyclobutanone), 3.49–3.35 (m, 2H, cyclobutanone), 3.19–3.06 (m, 2H, cyclobutanone); ¹³C NMR (126 MHz, CDCl₃) 206.9, 148.6, 148.3, 137.0, 136.1, 128.6, 127.9, 127.4, 119.1, 113.1, 112.0, 71.3, 56.1, 54.7, 28.0; MS (APCI): m/z = 283 [M+H]⁺; HRMS (ESI): MNa⁺, found 305.1148. C₁₈H₁₈NaO₃ requires 305.1154.

3-(3-(Benzyloxy)-4-methoxyphenyl)-1-(3,4,5-trimethoxyphenyl)cyclobutan-1-ol (6)

To a cold (–78 °C) stirred solution of 5-bromo-1,2,3-trimethoxybenzene (0.65 g, 2.63 mmol) in THF (10 mL), a solution of n-BuLi (1.1 mL, 2.5 M in hexanes, 2.75 mmol) was added dropwise under argon atmosphere. The reaction mixture was stirred at −78°C for 1 h, then a solution of compound 5 (0.71 g, 2.52 mmol) in THF (3 mL) was added dropwise at the same temperature. The mixture was stirred at −78 °C for additional 1 h, then allowed to warm to rt. and left overnight. Then it was poured onto cold saturated aq NH₄Cl and extracted with EtOAc (2×20 mL). The combined extracts were dried (Na₂SO₄) and evaporated in vacuo yielding the crude product, which was subjected to silica gel column chromatography (hexane –EtOAc – Et₃N (1:1:1) as eluent) affording pure title compound 6 (0.75 g, 1.67 mmol, 66%) as light yellow crystals; m.p. = 123–124°C; R_f = 0.60 (hexane – EtOAc – Et₃N (1:1:1)); ¹H NMR (400 MHz CDCl₃) 7.46 (d, J 7.5 Hz, 2H, H-2′ and H-6′), 7.37 (t, J = 7.4 Hz, 2H, H-3′ and H-5′), 7.31 (t, J = 7.4 Hz, 1H, H-4′), 6.91 – 6.84 (m, 3H, H-2, H-5 and H-6), 6.82 (s, 2H, H-2′′ and H-6′′), 5.16 (s, 2H, OCH₂), 3.91 (s, 6H, 3′′-OCH₃ and 5′′-OCH₃), 3.88 (s, 3H, 4′′-OCH₃), 3.87 (s, 3H, OCH₃), 3.05 – 2.90 (m, 3H, cyclobutane), 2.48 – 2.42 (m, 2H, cyclobutane), 2.09 (s, 1H, OH); ¹³C NMR (126 MHz, CDCl₃) 153.3, 148.3, 148.2, 141.0, 137.24, 137.18, 128.5, 127.8, 127.4, 119.2, 113.1, 111.9, 102.9, 72.6, 71.2, 60.8, 56.3, 56.2, 44.8, 29.8; MS (APCI): m/z = 433 [M+H–H₂O]⁺. HRMS (ESI): MNa⁺, found 473.1935. C₂₇H₃₀NaO₆ requires 473.1940.

2-Methoxy-5-(3-(3,4,5-trimethoxyphenyl)cyclobutyl)phenols (2a and 2b)

Compound 6 (1.50 g, 3.33 mmol), Pd/C (10%, 0.20 g), 36% aq HCl (0.2 mL), and MeOH (20 mL) were placed in an autoclave, and the mixture was hydrogenated at 10 MPa and rt for 48 h. Then the resulting mixture was filtered, washed with MeOH (5 mL) and evaporated in vacuo yielding the crude product, which was subjected to silica gel column chromatography (hexane:EtOAc=4:1) affording spectrally pure mixture of 2a and 2b (3:2 ratio, 0.41 g, 1.19 mmol, 36%); R_f 0.42 (hexane – EtOAc (4:1)).

A sample of the mixture (100 mg) was subjected to preparative HPLC (silica gel column, eluent: hexane – i-PrOH (99.5:0.5), flow rate: 15 mL/min) giving pure 2a (44 mg) and 2b (41 mg).

2-Methoxy-5-((1S,3S)-3-(3,4,5-trimethoxyphenyl)cyclobutyl)phenol (2a)

Yield 0.043 g (0.13 mmol), white crystals; m.p. = 107–109°C; ¹H NMR (500 MHz CDCl₃) 6.86 (d, J = 2.1 Hz, 1H, H-6), 6.81 (d, J = 8.1 Hz, 1H, H-4), 6.73 (dd, J = 8.2, 2.1 Hz, 1H, H-3), 6.47 (s, 2H, H-2′ and H-6′), 5.59 (s, 1H, OH), 3.89 (s, 3H, 4′-OCH₃), 3.88 (s, 6H, 3′-OCH₃ and 5′-OCH₃), 3.84 (s, 3H, OCH₃), 3.46 – 3.37 (m, 2H, cyclobutane), 2.76 (qd, J = 8.0, 2.8 Hz, 2H, cyclobutane), 2.19 (qd, J = 8.3, 2.9 Hz, 2H, cyclobutane); ¹³C NMR (126 MHz, CDCl₃) 153.1, 145.5, 144.8, 141.2, 138.8, 136.2, 117.7, 112.8, 110.5, 103.4, 60.8, 56.11, 56.05, 37.2, 36.3, 35.3; MS (APCI): m/z = 345 [M+H]⁺; HRMS (ESI): MNa⁺, found 367.1516. C₂₀H₂₄NaO₅ requires 367.1522.

2-Methoxy-5-((1R,3R)-3-(3,4,5-trimethoxyphenyl)cyclobutyl)phenol (2b)

Yield 0.041 g (0.12 mmol), white crystals; m.p. = 88–90°C; ¹H NMR (400 MHz CDCl₃) 6.97 (d, J = 2.0 Hz, 1H, H-6), 6.85 (d, J = 8.1 Hz, 1H, H-4), 6.81 (d, J = 8.2 Hz, 1H, H-3), 6.55 (s, 2H, H-2′ and H-6′), 5.63 (s, 1H, OH), 3.91 (s, 6H, 3′-OCH₃ and 5′-OCH₃), 3.90 (s, 3H, 4′-OCH₃), 3.86 (s, 3H, OCH₃), 3.59 (m, 2H, cyclobutane), 2.63 – 2.54 (m, 4H, cyclobutane); ¹³C NMR (126 MHz, CDCl₃) 153.2, 145.6, 144.7, 141.8, 139.3, 136.1, 117.8, 112.8, 110.6, 103.5, 60.9, 56.13, 56.06, 36.5, 35.7, 35.4; MS (APCI): m/z = 345 [M+H]⁺; HRMS (ESI): MNa⁺, found 367.1516. C₂₀H₂₄NaO₅ requires 367.1522.

4.2. X-ray diffraction study of compounds 2a and 2b

Data sets for compounds 2a and 2b were collected with a Bruker D8 Venture CMOS diffractometer. Programs used: data collection: APEX3 V2016.1–0 [37]; cell refinement: SAINT V8.37A [37]; data reduction: SAINT V8.37A [37]; absorption correction, SADABS V2014/7 [37]; structure solution SHELXT-2015 [38]; structure refinement SHELXL-2015 [39]. R-values are given for observed reflections, and wR² values are given for all reflections.

X-ray crystal structure analysis of 2a:

A colorless prism-like specimen of C₂₀H₂₄O₅, approximate dimensions 0.086 mm × 0.192 mm × 0.199 mm, was used for the X-ray crystallographic analysis. The X-ray intensity data were measured. A total of 1931 frames were collected. The total exposure time was 27.10 hours. The frames were integrated with the Bruker SAINT software package using a wide-frame algorithm. The integration of the data using a triclinic unit cell yielded a total of 48144 reflections to a maximum θ angle of 68.35° (0.83 Å resolution), of which 6472 were independent (average redundancy 7.439, completeness = 99.6%, R_int = 3.52%, R_sig = 2.03%) and 5952 (91.97%) were greater than 2σ(F²). The final cell constants of a = 10.9891(4) Å, b = 11.0398(4) Å, c = 14.8472(5) Å, α = 94.6760(10)°, β = 99.7360(10)°, γ = 90.3750(10)°, volume = 1768.99(11) ų, are based upon the refinement of the XYZ-centroids of 9954 reflections above 20 σ(I) with 6.061° < 2θ < 136.6°. Data were corrected for absorption effects using the multi-scan method (SADABS). The ratio of minimum to maximum apparent transmission was 0.911. The calculated minimum and maximum transmission coefficients (based on crystal size) are 0.8640 and 0.9380. The structure was solved and refined using the Bruker SHELXTL Software Package, using the space group P-1, with Z = 4 for the formula unit, C₂₀H₂₄O₅. The final anisotropic full-matrix least-squares refinement on F² with 461 variables converged at R₁ = 3.47%, for the observed data and wR₂ = 8.93% for all data. The goodness-of-fit was 1.037. The largest peak in the final difference electron density synthesis was 0.303 e⁻/ų and the largest hole was −0.231 e⁻/ų with an RMS deviation of 0.044 e⁻/ų. On the basis of the final model, the calculated density was 1.293 g/cm³ and F(000), 736 e⁻. CCDC Nr.: 1968723.

X-ray crystal structure analysis of 2b:

A colorless prism-like specimen of C₂₀H₂₄O₅, approximate dimensions 0.161 mm × 0.251 mm × 0.371 mm, was used for the X-ray crystallographic analysis. The X-ray intensity data were measured. A total of 1321 frames were collected. The total exposure time was 16.53 hours. The frames were integrated with the Bruker SAINT software package using a wide-frame algorithm. The integration of the data using an orthorhombic unit cell yielded a total of 12833 reflections to a maximum θ angle of 70.00° (0.82 Å resolution), of which 2724 were independent (average redundancy 4.711, completeness = 95.5%, R_int = 2.87%, R_sig = 2.99%) and 2711 (99.52%) were greater than 2σ(F²). The final cell constants of a = 12.1924(4) Å, b = 5.4355(2) Å, c = 26.1084(8) Å, volume = 1730.25(10) ų, are based upon the refinement of the XYZ-centroids of 487 reflections above 20 σ(I) with 12.53° < 2θ < 62.38°. Data were corrected for absorption effects using the multi-scan method (SADABS). The ratio of minimum to maximum apparent transmission was 0.885. The calculated minimum and maximum transmission coefficients (based on crystal size) are 0.7630 and 0.8860. The structure was solved and refined using the Bruker SHELXTL Software Package, using the space group Pca2₁, with Z = 4 for the formula unit, C₂₀H₂₄O₅. The final anisotropic full-matrix least-squares refinement on F² with 234 variables converged at R₁ = 2.56%, for the observed data and wR₂ = 6.54% for all data. The goodness-of-fit was 1.069. The largest peak in the final difference electron density synthesis was 0.131 e⁻/ų and the largest hole was −0.139 e⁻/ų with an RMS deviation of 0.030 e⁻/ų. On the basis of the final model, the calculated density was 1.322 g/cm³ and F(000), 736 e⁻. The hydrogen at O1 atom was refined freely. CCDC Nr.: 1968724.

4.3. Cytotoxity assay

Human hepatocellular carcinoma cell line HepG2 and neuroblastoma cell line SK-N-DZ were provided by Promega, USA. HepG2 and SK-N-DZ were cultured in DMEM/F12 supplemented with 10% fetal bovine serum in the presence of L-glutamine and penicillin/streptomycin. Cells were maintained in the logarithmic growth phase in a humidified atmosphere containing 5% CO₂ at 37°C in all cases.

Before plating, the cells were first washed with PBS and then trypsinized with 1X trypsin diluted in DPBS. The appropriate volume of culture medium was then placed in a flask to stop the trypsinization, and cells were counted using a counting chamber after being stained with Trypan Blue. The appropriate volume of the cell solution was placed in a new falcon tube containing completed DMEM medium to give a desired final concentration.

HepG2 and SK-N-DZ cells at 1.25·10⁴ cells/mL and 2.5·10⁴ cells/mL, respectively, were plated in sterile 384-well black wall clear flat-bottom plates 45 μL/well (562.5 and 1125 cells/well). Plates with cells were sealed and left overnight in a humidified atmosphere at 37°C and 5% CO₂ for adaptation and adherence. The appropriate concentrations of cells correspond to 20–30% of confluence before incubation with drugs.

Serial dilutions were performed as follows. 10 points 3-fold serial dilutions in 100% DMSO from 200 uM of CA4 and from 2 mM of doxorubicin and test compounds were made. For each concentration point of serial dilution was made 20-fold intermediate dilution in DPBS: 5μL + 95μL of DPBS.

The compound solutions (5 μL) with concentrations 0.00005–1 μM (CA4) and 0.005–10 μM (doxorubicin, 2a, and 2b) were added in quadruplicates and incubated with cells in a humidified atmosphere containing 5% CO₂ at 37°C and for 72 h. The final concentration of DMSO in assay was 0.5%.

10% of 500 uM resazurin in DPBS (5.5 μL) was added to 50 μM final concentration and incubated in humidified atmosphere containing 5% CO₂ at 37°C for 3 h. The presence of resazurin was quantified by measuring fluorescence, E_x – 555 nm, E_m – 585 nm, cut-off 570, bottom read.

The IC₅₀ values of reference and test compounds were calculated by plotting the % cell toxicity against log (concentration of the compound (μM)) in GraphPad PRISM 7.0 software. The data represent the best fit values of sigmoidal fit with variable slope.

4.4. Molecular docking protocol

For molecular docking, the published tubulin structure [27] and compounds 1a–d, 2a, and 2b were used. Prior to the molecular docking routine, water molecules were removed from the tubulin structure. All the Arg and Lys residues were protonated in order to be capable of forming a large number of hydrogen bonds. To each compound, all the possible low-energy conformations were generated.

Molecular docking was performed using flexible ligands and rigid tubulin molecule. An algorithm of systematic docking (SDOCK+) built-in QXP docking software was used [40, 41]. This method has shown the high reproducing ability of ligand conformation with minimum RMSD in comparison to the crystallographic data [40].

The maximum number of SDOCK+ routine steps was set to 300, and the 10 best complexes based on the built-in QXP scoring function [41] were selected for analysis in the next stages of investigation. In accordance with the defined pharmacophore models, the resulting/selected previously “tubulin-ligand” complex structures were filtered by intrinsic Flo+ filters and multiRMSD software package [42]. Filtering was based on the built-in QXP scoring function, the number of hydrogen bonds, the tubulin-ligand contact surface area and the distance between the ligand and the key points of the corresponding pharmacophore model. As a result of molecular docking to each compound one (best) complex with tubulin was selected.

4.5. Molecular dynamic (MD) simulation protocol

To estimate stability and crucial interactions of obtained complexes after molecular docking, MD simulation was used. The calculations were done using Gromacs 5.1.3 [43] in force field Charmm36 [44]. In all cases, tubulin was protonated according to build-in function in Gromacs 5.1.3. The topology files for ligands were generated by SwissParam [45]. The complexes obtained after molecular docking were used for MD simulation. The system was placed into the center of periodic cubic box which was filled with TIP3P water molecules. Minimum 0.9 nm distance was maintained between the complex and the edge of the simulation box so that complex can fully immerse with water and rotate freely. Then, Na⁺ and Cl^– ions were added to neutralize the system electrostatically, and following this Na⁺ and Cl^– ions were added to bring the ionic concentration to 150 mM (mimicking the cellular environment). In the system solvent molecules randomly replaces with monoatomic ions. Next obtained complex was energy minimized which also relieved any steric clashes. The system was relaxed by applying steepest descent algorithm (the maximum number of steps was 50000). Then the equilibration was computed in two stages: NVT was first equalized at 100 ps, with the second NPT equalization of 1 ns. After that, we launched MD simulation within 25 ns. All calculations were done using the “leap-frog integrator” at the temperature of 300 K and at constant atmospheric pressure.