Cytotoxic substituted indolizines as new colchicine site tubulin polymerisation inhibitors

Abstract A potential microtubule destabilising series of new indolizine derivatives was synthesised and tested for their anticancer activity against a panel of 60 human cancer cell lines. Compounds 11a, 11b, 15a, and 15j showed a broad spectrum of growth inhibitory activity against cancer cell lines representing leukaemia, melanoma and cancer of lung, colon, central nervous system, ovary, kidney, breast, and prostate. Among them, compound 11a was distinguishable by its excellent cytostatic activity, showing GI50 values in the range of 10–100 nM on 43 cell lines. The less potent compounds 15a and 15j in terms of GI50 values showed a high cytotoxic effect against tested colon cancer, CNS cancer, renal cancer and melanoma cell lines and only on few cell lines from other types of cancer. In vitro assaying revealed tubulin polymerisation inhibition by all active compounds. Molecular docking showed good complementarity of active compounds with the colchicine binding site of tubulin.


Introduction
Recognised as key dynamic structural components in cells, microtubules play an important role in cellular shape organisation, intracellular movement, cell division, and mitosis 1,2 . Thus, they have been considered an attractive target for the development of new antiproliferative agents in the past few years [3][4][5][6][7] .
Commonly, agents targeting tubulin are divided function of the site they interact with tubulin. The three major sites of tubulin are: the paclitaxel binding site (compounds showing microtubule stabilising effects), vinblastine binding site, and colchicine binding site (compounds showing inhibition of tubulin polymerisation) 1, [3][4][5]8 . The microtubule-targeting strategy in the drug development field is validated by the use of microtubule-targeting agents such as paclitaxel, docetaxel, vinblastine, colchicine, combretastatin A-4 (CA-4), nocodazole, and many others in cancer chemotherapy 1,[8][9][10] . However, these compounds also present many drawbacks (high toxicity that induces many side effects, low oral bioavailability, and development of drug resistance) that limit their efficiency [11][12][13] . Therefore, there is a huge demand of novel antimitotic agents to overcome the abovementioned inconveniences.
Among the large number of microtubule-targeting agents with diverse scaffolds investigated during last decades 1,2,11 , Phenstatin ( Figure 1) stands out as one of the simplest molecules that significantly inhibit tubulin polymerisation by binding to the colchicine site of tubulin 14,15 . Phenstatin is also known for its outstanding antitumor activities on a wide variety of human cancer cells 14,15 . Its biological properties are comparable to CA-4, currently investigated in clinical trials 16 , but in contrast to CA-4, Phenstatin has a greater pharmacological potential due to improved metabolic stability and requires an easier synthesis for large-scale production 17 . In the process of drug discovery, this kind of compounds are lead scaffolds for the development of improved bioactive analogues, and Phenstatin continues to be a source of inspiration for researchers in designing new potential anticancer drugs [18][19][20][21][22][23][24][25] .
To increase the structural diversity, the combination of various types of bioactive moieties and rings has become a practical strategy in the field of medicinal chemistry. Thus, our study was inspired by several reported anticancer active Phenstatin analogues having the 3 0 -hydroxy-4 0 -methoxyphenyl ring replaced with an indolizine moiety 21 (compound A, Figure 1) or a pyrrolo [1,2-b]pyridazine group 24 (compound B, Figure 1).
At the same time, the pyridine ring, present in many natural products and even genetic material, has been noted for its role in many biological processes as well as in cancer pathogenesis, which makes it a privileged scaffold in anticancer agents discovery 26 . Thus, there are reports of a variety of monosubstituted pyridines with cytotoxic activity [26][27][28][29] , and there are also several pyridyl-containing drugs introduced on the market for their antitumor properties 26 (Figure 1, right).
In our continuous efforts 22,24 to discover more effective microtubule destabilising agents, we have applied a structural combination strategy to design and synthesise a new series of indolizinebased Phenstatin analogues and evaluated their anticancer activity. Thus, we considered unsubstituted indolizines (at the pyridine ring) (compounds C, Figure 1), as well as several pyridyl-substituted indolizines (compounds D, E, and F, Figure 1) in order to investigate a possible beneficial influence of this group to the binding properties of generated compounds due to the lone pair electrons in this moiety. In order to increase biological activity, we also designed an indolizine series that contains two 3,4,5-trimethoxybenzoyl groups (compounds G, Figure 1). Furthermore, we modified the 3,4,5-trimethoxyphenyl ring of Phenstatin for each series, replacing it with either a 3,5-dimethoxyphenyl, 3,4-dimethoxyphenyl, or a 4-bromophenyl ring. The structure-activity relationships (SARs), effects on tubulin assembly and theoretical binding interactions were also explored in our study.

Chemistry
All commercially available reagents and solvents employed were used without further purification. Melting points were recorded on a A. Kr€ uss Optronic Melting Point Meter KSPI and are uncorrected. Analytical thin-layer chromatography was performed with commercial silica gel plates 60 F254 (Merck, Darmstadt, Germany) and visualised with UV light (k max ¼254 or 365 nm). The NMR spectra were recorded on an Avance III 500 MHz spectrometer (Bruker, Vienna, Austria) operating at 500 MHz for 1 H and 125 MHz for 13 C. Chemical shifts were reported in delta (d) units, part per million (ppm), and coupling constants (J) in Hz. The following abbreviations were used to designate chemical shift multiplicities: bs ¼ broad singlet. Infrared (IR) data were recorded as films on potassium bromide (KBr) pellets on an FT-IR (Shimadzu, Kyoto, Japan) Prestige 8400s spectrophotometer or a Jasco 660 plus FTIR spectrophotometer. Analyses indicated by the symbols of the elements or functions were within ±0.4% of the theoretical values.
General procedure for preparation of compounds 11a-l and 14a-d The cycloimmonium salt (8a-l and 13a-d) (1 mmol, 1 equiv.) and ethyl propiolate (1.1 mmol, 1.1 equiv.) were added to dichloromethane (DCM) and the obtained suspension was stirred at r.t. Then, a solution of triethylamine (TEA) (3 mmol, 3 equiv.) in DCM (3 ml) was added drop-wise over 1 h (magnetic stirring) and the resulting mixture was then stirred overnight at r.t. Methanol (10 ml) was added and the resulting solid was collected by filtration and washed with 5 ml methanol. The product was then purified by crystallisation from DCM/methanol (1/1, v/v) and/or column chromatography using DCM/methanol (99.5/0.5, v/v).

Anticancer activity
The compounds were tested against a panel of 60 human cancer cell lines at the National Cancer Institute (NCI) (Rockville, MD). The cytotoxicity experiments were performed using a 48 h exposure protocol which consisted of a sulphorhodamine B assay [33][34][35] .
The polymerisation was monitored using FLUOstar Omega multi-mode microplate reader (BMG LABTECH, Ortenberg, Germany). The first step before the analysis is the pre-warming of the plate to 37 C for 30 min. Plate temperature is essential for high polymerisation activity and reproducible results. The tubulin polymerisation buffer will be composed of general tubulin buffer, tubulin glycerol buffer, and GTP stock, at 4 C. Another volume of 500 ml of general tubulin buffer is necessary for dilutions, at r.t. Ten microlitres of GTB will be pipetted into each well. Two of the wells will remain with GTB only, as controls. Into the rest, 10 ml of compounds of Â10 strength or Phenstatin, also of Â10 strength will be added. The final concentration of the compounds and Phenstatin will be 10 mM. The plate will be incubated at 37 C for 2 min. Meanwhile, proceed to the dilution of 10 ml of the Paclitaxel Stock solution with 190 ml of r.t. general tubulin buffer, to be used in quantities of 10 ml of this per well. The tubulin will be defrosted until r.t., placed on ice and diluted with 420 TP cold buffer, reaching a final concentration of 3 mg/ml in 80 mM PIPES (piperazine-N,N 0 -bis(2-ethanesulfonic acid) sesquisodium salt), pH ¼ 6.9, 2 mM MgCl 2 , 0.5 mM EGTA (ethylene glycol-bis (b-amino-ethyl ether) N,N,N 0 ,N 0 -tetra-acetic acid, 1 mM GTP and 10.2% glycerol. The diluted tubulin will be used immediately by adding 120 ml into each of the wells with a multichannel pipette. The absorbance was measured at 340 nm for 1 h at 1 min intervals using a plate reader at 37 C. Representative experiment (n ¼ 3) is shown in Figure 2.

Molecular modelling
Flexible-ligand docking experiments were performed as previously reported 24 , with slight modifications. Briefly, the 3D structures of the ligands were constructed in Avogadro v1.2.0 38 and were energetically optimised in the MMFF94 force field until a local energy minimum was achieved. Autodock Vina 39 was used for all docking experiments, using a 22 Â 22 Â 22 Å 3 gridbox centred on the colchicine binding site of the a,b-tubulin heterodimer (PDB: 4O2B) 40 . The co-crystallised colchicine ligand and water molecules were removed during protein preparation for docking, and the target protein was kept rigid. Twenty poses were generated for each ligand, which were then ranked based on theoretical binding energy. The best ranked models were visually inspected in order to assess the consistency of the generated docking solutions relative to the docking poses of the known inhibitors colchicine and Phenstatin. In order to evaluate the quality of the docking protocol, colchicine was extracted from the crystal structure and redocked into the binding site. RMSD between re-docked ligand and co-crystallised conformation was computed in PyMOL. Visual inspection, molecular graphics and analyses were made in the PyMOL Molecular Graphics System, Version 1.
Indolizines 15a-i were obtained in good yields using the substitution of halides 4-6 generated by the indolizines 11e-g (Scheme 3) 32,42 . We also synthesised the previously reported compound 15j 32 using a similar procedure.
The structures of all new target compounds were fully confirmed by 1 H and 13 C NMR, IR and elemental analyses.

Biological activity
Anticancer activity All synthesised compounds were submitted to the NCI, and 13 compounds (11a, b, d, e, f, i, j, k, l, 14a, 14d, 15a, and 15j) were selected for single dose (10 À5 M) screening against a panel of 60 human tumour cell lines, representing leukaemia, melanoma and cancer of lung, colon, central nervous system, ovary, kidney, prostate, and breast 33 . Representative results for 11 of the compounds are summarised in Table 1.
Indolizines 11a, b showed a very good inhibition effect on almost all 60 lines, the best results being registered on leukaemia HL-60 (TB) cells, colon cancer COLO 205 cells, SNC cancer SF-539 cells, melanoma M14 and MDA-MB-435 cells, ovarian cancer cell OVCAR-3, renal cancer A498 and RXF393 and breast cancer MDA-MB-468 cells. Compound 11a also exhibited a cytotoxic effect on all these lines, the best one being registered on melanoma MDA-MB-435 cells.
Interestingly, the substitution of indolizine heterocycle at position 7 with a pyrid-4-yl or pyrid-2-yl ring resulted in the loss of the activity, compounds 11e-f and 11i, 11j, and 11l (data not shown for compound 11l) presenting almost no inhibition effect on the tested cell lines. As an exception, compound 11k selectively inhibited the growth of NCI-H522 and NCI-H460 nonsmall cell lung cancer, SK-OV-3 ovarian cancer cells and T-47D breast cancer cells.
Substitution of the indolizine heterocycle at position 5 with a pyrid-2-yl ring also led to a loss of growth inhibition effect. Thus, compounds 14a (Table 1) and 14d showed no inhibition on tested cancer cells (data not shown for compound 14d). The presence of two 3,4,5-trimethoxybenzoyl groups in compound 15a led to the best growth inhibition effect on the tested cancer cells. Compound 15a is also distinguished by the high cytotoxic activity displayed on 18 cell lines, including cell lines from each panel except prostate cancer. Of the same serious, compound 15j showed similar behaviour to 15a on most line cells, but with much lower GI % values on NCI-H522 lung cancer cells, SF-268 CNS cancer cells, MDA-MB-231/ATCC and BT-549 breast cancer cells.
As shown in Table 1, the substitution of the 3,4,5-trimethoxyphenyl ring produced different effects in series of compounds C and D (Figure 1). Thus, in series C (11a-d), replacing the 3,4,5-trimethoxyphenyl ring with 3,5-dimethoxyphenyl does not alter the potency substantially, while substitution with 4-bromophenyl ring causes a dramatical loss of inhibitory properties. In series D (11i-l), the 3,4-dimethoxyphenyl ring appears to be the only one to confer selective inhibitory properties against the above mentioned cancer cell lines.
Showing the most significant growth inhibition, compounds 11a, 15a, and 15j were selected for evaluation against 60 cell lines at five concentrations [33][34][35] . Results from the NCI-60 5-dose screen are shown in Table 2.
All three compounds displayed good antiproliferative properties. The best candidate in terms of growth inhibition properties was compound 11a, with GI 50 values < 100 nM against 47 cell lines, most notably on melanoma MDA-MB-435 (GI 50 <10 nM) and UACC-62 (GI 50 ¼31 nM) cells, leukaemia SR cells (GI 50 ¼23 nM), and renal cancer A498 cells (GI 50 ¼27 nM). Compound 11a displayed selective cytotoxic activity on the melanoma MDA-MB-435 cell line (LC 50 ¼20.4 mM). Even if the overall antiproliferative activity of compound 11a is lower in comparison with control Phenstatin, there are 11 cancer cell lines against which compound 11a showed better GI 50 values and several other lines against which compound 11a had comparable GI 50 values.
Although displaying excellent growth inhibitory effects at the 10 À5 M single dose evaluation (Table 1), compounds 15a and 15j did not exhibit submicromolar GI 50 values, except compound 15j with GI 50 ¼0.33 mM against leukaemia SR cell line. Notably, both compounds showed considerable cytotoxic activity against all colon cancer, CNS cancer, renal cancer, and melanoma cell lines, but also on some cell lines from non-small cell lung cancer, ovarian cancer, breast cancer, and prostate cancer, respectively. This different behaviour of compounds 15a and 15j in comparison with compounds 11a and Phenstatin, suggests different mechanisms of action.

In vitro tubulin polymerisation inhibition activity
In order to confirm if the observed anticancer activity of the above mentioned compounds is conferred by a microtubule-targeting mechanism, we evaluated the effect of the active compounds 11a, 11b, 11k, 15a, and 15j, but also of the inactive compound 14a (for comparison) on the assembly of tubulin. To confirm their influence on microtubule dynamics, two positive controls were used: paclitaxel (a tubulin stabiliser) and Phenstatin (a tubulin polymerisation inhibitor). As presented in Figure 2, Paclitaxel was found to stimulate tubulin polymerisation, while Phenstatin and all six tested compounds appeared to inhibit tubulin polymerisation. Four compounds, namely 11k, 14a, 15a, and 15j showed a similar strong inhibitory behaviour on tubulin polymerisation, superior to the one obtained for compounds 11a and 11b. The obtained data clearly indicated that all tested target compounds effectively inhibit tubulin polymerisation in vitro.

Molecular modelling
Molecular docking experiments revealed similar docking conformations of the tested compounds to previously reported anticancer pyrrolo[1,2-b]pyridazines (compounds type B, Figure 1), which are also thought to achieve anticancer activity by binding to the colchicine binding site of tubulin 24 . The used docking protocol was validated by computing the RMSD between redocked colchicine and its co-crystallised conformation, which was 0.16 Å in our case. Generally, an RMSD value below 2 Å (the average resolution of a crystal structure) is considered acceptable 43 .
Compound 11a is stabilised in the colchicine binding site of tubulin through H-bonding between its carbonyl moiety and bAsn258 and through extensive hydrophobic contacts with the accommodating pocket formed by bCys241, bLeu248, bLeu255, bAla250, bAla316, bIle318, bAla354, and bIle378, similar to Phenstatin. Moreover, the indolizine heterocycle extends towards the a subunit with the help of two polar interaction partners: bLys254 and aAsn101, and is further stabilised by the nonpolar chain of bLys352, which is engaged in an H-bond with aThr179 through its e-NH 3 þ group. The 4-methoxy substituent does not engage in an H-bond with bCys241, as seen in the case of Phenstatin, but the binding orientation of this compound suggests that this bond could form if the binding site would be optimised. Further molecular dynamics simulations could be performed in order to investigate the formation and stability of this H-bond. Removal of the 4-methoxy group leads to a slight shift of the methoxy-substituted ring compared to 11a, possibly to optimise hydrophobic contacts, as reflected by the similar theoretical binding energy (Table 3), but the formed H-bonds with bAsn258, bLys254, and aAsn101 are enough to maintain a conformation roughly overlapping with 11a, which could account for the observed biological activity exhibited by this compound. The bromo-substituted compound 11d was accommodated more deeply in the colchicine binding pocket, exclusively through hydrophobic interactions, having one of the lowest theoretical binding energies of all docked compounds. The absence of polar contacts upon accommodation at the tubulin binding site could explain the low biological properties exhibited by this compound. Docking experiments for compounds 11e, 11f, 11i, 11j, and 11k did not reveal any conformations in which the methoxy-substituted cycle would overlap with the one in the colchicine binding site as to permit a polar interaction with bCys241. Instead, this aromatic moiety was oriented towards a subunit of the binding pocket, being stabilised by H-bonds with bLys254 (11e, 11i) or aAsn101 (11j), as well as weak hydrogen bonding interactions with aSer178, bGln247 and the nonexchangeable GTP molecule. The indolizine moiety is stabilised by extensive hydrophobic contacts and, in the case of 11f, 11k, by additional amide stacking with the backbone of bLys254. The pyridyl ring is positioned deep in the colchicine binding site, away from the a/b interface, and is stabilised through hydrophobic interactions with residues in the b subunit. Since all these compounds have good binding energies, yet lack activity, it could be postulated that a polar interaction with bCys241, or at least the positioning of possible polar interaction partners in the proximity of this residue is crucial for the observed anticancer activity, as has been seen for other colchicine binding site inhibitors [44][45][46] . However, an exception can be seen at compound 11k, which showed selective activity against nine cancer cell lines (GI > 70%), and also inhibited tubulin polymerisation in vitro, but did not form a favourable contact with this residue in our docking experiments. Further mutagenesis experiments could be performed in order to describe the impact of this residue on the binding properties of the tested compounds. Despite its potent anticancer activity, the low theoretical binding score of 11a compared to 11e-k suggests that its cytotoxicity may involve other cellular targets or pathways other than the ab-tubulin heterodimer. At the same time, since cancer cells preferentially express different b-tubulin isoforms, it would be possible that this compound binds with greater affinity to other isoforms. This aspect could be further studied in silico, as has been done for DAMA-colchicine 47 and other colchicine binding site microtubule depolymerising agents 48 .
Compound 14a was accommodated in a similar fashion to compounds 11a,b,d, preserving the interactions of the methoxysubstituted moiety with the hydrophobic pocket formed by bCys241, bLeu248, bLeu255, bAla250, bAla316, bIle318, bAla354, and bIle378, but its indolizine ring rotated as to permit the interaction between the pyrid-2-yl ring and bMet259 through pi-sulphur stacking. This rotation also led to the formation of an Hbond between the carboxylate moiety of this compound and bGln247. While the in vitro tubulin polymerisation assay results are in agreement with the docking observations, the lack of anticancer activity in the case of compound 14a remains to be elucidated.
Compound 15a occupied the colchicine binding site similar to compounds 11a,b,d and Phenstatin, engaging in H-bonds with bCys241 and bAsn258, and forming hydrophobic contacts with bLeu248, bLeu255, bAla250, bAla316, bIle318, and bAla354. The additional 3,4,5-trimethoxybenzoyl group reached towards the H10 helix of the b-tubulin subunit to form H-bonds with bThr353 and bGln336, as well as pi-anion stacking with the sidechain of bAsp329. The indolizine moiety was stabilised by amide-pi stacking with the backbone of aThr179. This compound had the one of the lowest binding energies of all tested molecules (-10.2 kcal/ mol), being surpassed only by compound 15j (-10.7 kcal/mol). Interestingly, compound 15j forms a hydrophobic interaction with bCys241 and maintains many of the polar and hydrophobic contacts observed at 15a, being accommodated in the same extended binding site-spanning conformation. Additional Table 3. Binding orientation, energy, and amino acid contacts for tested compounds, as predicted by molecular docking experiments.
(continued) For binding orientation, the a,b-tubulin heterodimer is shown as ribbons; aminoacids and ligands are represented as sticks; for 2D interaction diagrams, colours are as follows: conventional hydrogen bondsgreen, carbon-hydrogen bondspale green; hydrophobic interactionslight pink; amide-p stackingdark pink; anion-p stacking: orange; p-sulphur stacking: dark yellow. molecular dynamics experiments should be performed in order to confirm the stability of the observed interactions, especially with bCys241.

Conclusions
Twenty-six new substituted Phenstatin analogues with an indolizine core were synthesised and submitted to NCI for anticancer activity evaluation. Thirteen compounds were selected and tested against a panel of 60 human cancer cells. Tubulin polymerisation assays and docking studies were also performed for the active compounds. Compounds 11a, 11b, 15a, and 15j showed excellent inhibitory properties on a broad range of cancer cell lines, and tubulin polymerisation assays revealed significant inhibitory effects on tubulin assembly for these compounds. This mechanism of action is further supported by docking experiments, which showed that all four compounds fit well to the colchicine binding site of tubulin. Interestingly, substitution of the indolizine heterocycle at position 7 with a pyrid-4-yl or pyrid-2-yl, or at position 5 with a pyrid-2-yl ring resulted in the loss of anticancer activity. As an exception, compound 11k showed a good inhibitory profile on tubulin polymerisation, but only selectively inhibited the growth of NCI-H522 and NCI-H460 non-small cell lung cancer, SK-OV-3 ovarian cancer cells, and T-47D breast cancer cell lines. Interestingly, inhibitory tubulin polymerisation properties, as well as a good compatibility for the colchicine binding pocket of tubulin are shown by 14a, but this compound is basically inactive against the tested cancer cells. Taken together, these results offer new SAR insights into this class of compounds and prove that using a strategy of structural combination can generate new colchicine site tubulin polymerisation inhibitors, as well as highly cytotoxic molecules against various cancer cells, which could aid the general research community in their ongoing anticancer efforts.