B-Ring-Aryl Substituted Luotonin A Analogues with a New Binding Mode to the Topoisomerase 1-DNA Complex Show Enhanced Cytotoxic Activity

Topoisomerase 1 inhibition is an important strategy in targeted cancer chemotherapy. The drugs currently in use acting on this enzyme belong to the family of the camptothecins, and suffer severe limitations because of their low stability, which is associated with the hydrolysis of the δ-lactone moiety in their E ring. Luotonin A is a natural camptothecin analogue that lacks this functional group and therefore shows a much-improved stability, but at the cost of a lower activity. Therefore, the development of luotonin A analogues with an increased potency is important for progress in this area. In the present paper, a small library of luotonin A analogues modified at their A and B rings was generated by cerium(IV) ammonium nitrate-catalyzed Friedländer reactions. All analogues showed an activity similar or higher than the natural luotonin A in terms of topoisomerase 1 inhibition and some compounds had an activity comparable to that of camptothecin. Furthermore, most compounds showed a better activity than luotonin A in cell cytotoxicity assays. In order to rationalize these results, the first docking studies of luotonin-topoisomerase 1-DNA ternary complexes were undertaken. Most compounds bound in a manner similar to luotonin A and to standard topoisomerase poisons such as topotecan but, interestingly, the two most promising analogues, bearing a 3,5-dimethylphenyl substituent at ring B, docked in a different orientation. This binding mode allows the hydrophobic moiety to be shielded from the aqueous environment by being buried between the deoxyribose belonging to the G(+1) guanine and Arg364 in the scissile strand and the surface of the protein and a hydrogen bond between the D-ring carbonyl and the basic amino acid. The discovery of this new binding mode and its associated higher inhibitory potency is a significant advance in the design of new topoisomerase 1 inhibitors.


Introduction
Cancer continues to be one of the leading causes of death worldwide. According to the latest data from the International Agency for Research on Cancer, in 2012 there were 14.1 million new cancer cases, 8.2 million cancer deaths and 32.6 million people living with cancer (within 5 years of diagnosis). Cancer is no longer a disease of the developed world, with 57% of newly diagnosed cases of cancer and 65% of deaths being associated with less developed regions [1].
Cancer therapy is still founded on the pillars of surgery, radiotherapy and chemotherapy, with immunotherapy having recently entered the stage as a fourth approach [2]. Nevertheless, the development of new anticancer drugs continues to be essential in the fight against the disease [3].
Topoisomerases are present in all living organisms and are crucial for relieving torsional tension in supercoiled DNA in the course of DNA replication, transcription and reparation [4]. Topoisomerases, and topoisomerase 1 in particular, are among the most relevant anticancer targets [5,6].
The camptothecins, specially irinotecan, topotecan and belotecan (Figure 1), are the main family of clinically relevant topoisomerase 1 inhibitors [7]. These compounds have a planar, pentacyclic core comprising a lactone functional group in ring E and containing a stereocenter at C-20, which must be in the S configuration for camptothecins to be active. Their pharmacologic target is the covalent topoisomerase 1-DNA binary complex, where they can bind non-covalently at the interphase formed between both macromolecules during the enzimatic catalytic cycle. This binding stabilizes the complex and retards its dissociation, finally leading to irreversible DNA damage and cell death [6,8].
In spite of their widespread use, they show severe undesired effects, their main dose-limiting toxicities being myelosupression, diarrhea and bone marrow toxicity [9]. Another serious problem with the camptothecins is their low stability [10], which is due to the easy opening of its lactone E ring to give an inactive hydroxy acid form (camptothecin carboxylate) that is sequestered by seric albumin [11]. The lactone-hydroxy acid equilibrium is relevant to bladder toxicity, another problem associated to the clinical use of the camptothecins that is due to accumulation of the active dlactone form owing to lactonization at the acidic pH of urine [12].
Against this background, camptothecin analogues lacking the lactone ring seem attractive, but during decades this structural feature was believed to be essential for activity. The discovery of the stabilization of the human topoisomerase 1-DNA covalent binary complex by the alkaloid luotonin A, closely related to the camptothecins but having an aromatic E ring, brought about a landmark change in this paradigm [13]. Furthermore, luotonin A is achiral, a feature that can be viewed as an advantage in terms of its potential role as a lead compound in drug discovery and development because it simplifies synthetic access to new analogues and their analytical control. In summary, although the potency of luotonin is much lower than that of the Camptothecin is a natural topoisomerase 1 inhibitor that has been used as a lead for the development of a family of anticancer agents in clinical use, exemplified by topotecan. Nevertheless, the camptothecins suffer severe limitations because of their low stability, which is associated with the hydrolysis of the d-lactone moiety in their E ring that leads to an inactive carboxylate form. Luotonin A is a plant alkaloid whose structure strongly resembles that of camptothecin but lacks the lactone moiety. The discovery that luotonin A is also a topoisomerase 1 inhibitor, although less potent than the camptothecins, provided a unique opportunity for drug discovery in the anticancer area. doi:10.1371/journal.pone.0095998.g001 camptothecins [14], it does provide an excellent opportunity for the discovery of improved topoisomerase 1 inhibitors.
The low potency of the natural luotonin A has prompted the synthesis and biological evaluation of a considerable number of analogues [15][16][17][18]. Nevertheless, these studies have been limited by the constraints found in known synthetic methodologies, specially in terms of A and B ring modification. Among the methods allowing access to the luotonin A framework, the Friedlä nder reaction between precursors corresponding to the C-D-E fragment and aromatic o-aminoketones or aldehydes provides the best opportunity for this type of structural manipulation. We have recently reported a protocol that improves the yields of this classical reaction by employing Ce(IV) ammonium nitrate (CAN) as a catalyst [19], and describe now its application to the synthesis of A and B ring-modified luotonin A analogues, together with studies of their activity as topoisomerase 1 poisons and cytotoxic agents.

Synthesis
General experimental information. All reagents were of commercial origin (Aldrich, Fluka, Acros) and were used as received. Melting points reported in this work were measured in open capillaries. Flash chromatography was performed on silica gel (230-400 mesh) or neutral alumina having the activity grade reported in each case. 1 H and 13 C NMR spectra were measured at 250/300 and 63/75 MHz, respectively, with Bruker 250 and 300 MHz (Avance) instruments maintained by the CAI de Resonancia Magnética Nuclear, Universidad Complutense. The solvents used were CDCl 3 and CF 3 CO 2 D. Chemical shifts are reported as d values (ppm). All one-and two-dimensional NMR spectra were obtained using Bruker Topspin software. IR spectra were recorded on a Perkin-Elmer FT-IR Paragon 1000 spectrometer in film form, prepared by evaporation of a few drops of sample solution over a sodium chloride window. Combustion elemental analyses were obtained by the CAI de Microanálisis, Universidad Complutense, using a Leco CHN 932 Elemental Analyzer.
For preparation of compounds 1f and 1g, a solution of the corresponding 2-aminonitrile (5 mmol, 1 eq) in anhydrous THF (10 mL) was added dropwise over a solution of 3,5-dimethylphenylmagnesium bromide (15 mmol, 3 eq) and the reaction mixture was refluxed for 2.5 h. After cooling, the reaction mixture was slowly poured onto ice water (30 mL). A 3 M aqueous solution of hydrochloric acid (21.5 mL) was then added, the mixture was stirred at room temperature for 1 h, and the organic phase was separated. The aqueous layer was extracted with ethyl ether (2620 mL). The combined organic layers were washed with water (1620 mL), brine (1620 mL), dried over MgSO 4 and filtered. The filtrate was concentrated in vacuo and the residue was either crystallized from ethanol or purified by column chromatography on silica gel eluting with petroleum ether:ethyl acetate to give compounds 1f or 1g.

Pharmacological Studies
DNA relaxation assay. DNA relaxation assay was carried out to investigate the inhibitory effect of luotonin A derivatives on the activity of the human topoisomerase 1 enzyme. A final reaction volume of 20 mL containing 200 ng of pGEM-5Zf(+) plasmid (Promega, Madison, Wisconsin) in its native supercoiled form, the corresponding drug diluted in 1 mL DMSO for a final concentration of 10 mM, 1 unit of human topoisomerase 1 (Sigma-Aldrich, St. Louis, Missouri) and the enzyme reaction buffer (TopoGEN, Port Orange, Florida) was incubated over 15 minutes at 37uC. Lanes containing native supercoiled and fully relaxed forms were run from a mixture containing the reaction buffer and plasmid, for the first one; and the reaction buffer, plasmid and human topoisomerase 1, for the latter. After incubation, the reactions were stopped by addition of a mixture of electrophoresis loading buffer and 20% SDS. The reaction products were loaded into a 1% agarose gel (Bio-Rad, Hercules, California) and separated for 3 h at 60 V in Tris-borate-EDTA buffer. The gels were stained using GelRed dye (Biotium, Hayward, California) according to the manufacturer's protocol, and photographed in 254 nm UV light using a ChemiDoc XRS+ (Bio-Rad) system. The bands were analysed using OriginPro 8.6 software (OriginLab, Northampton, Massachusetts) and the inhibitory activity of the drugs was calculated for each one as the supercoiled to unwinded bands ratio, and normalised to the activity of camptothecin (CPT) (Sigma-Aldrich, St. Louis, Missouri).
Cell culture and in vitro anti-proliferation assay. Three human cell lines were used; HeLa (cervical carcinoma), A549 (lung adenocarcinoma) and SW480 (colon adenocarcinoma) were obtained from the American Type Culture Collection (ATCC, Manassas, USA). All cell lines were seeded in 96-well plates at a density of 19000 (HeLa) and 5000 (A549 and SW480) cells/well in 200 ml of Dulbeccos modified Eagles medium (DMEM, Life Technologies, Carlsbad, USA) supplemented with 10% fetal bovine serum (FBS) (Life Technologies). Cell cultures were maintained at 37uC under 5% CO 2 .
After 5 h, cells were treated with compounds 3a-g: 1 mL DMSO containing a compound to yield final concentrations of 25.0, 12.5, 6.3 and 3.1 mM per well, the final concentration of DMSO being 0.5%. Every concentration of the drugs was assayed in triplicate. In addition, every 96-well plate contained wild type cells, negative control wells (0.5% DMSO) and positive control wells containing CPT (Sigma-Aldrich) to final concentrations of 25.0 and 2.5 mM, which caused 100% cell mortality. The plate was incubated for 72 h before the fixation step.
Moreover, a no-growth control 96-well plate was seeded in every experiment. The cells on this plate were allowed to adhere for 5 h as in the experiment plate and then directly fixed, dried and stored until dying as described below, with no exposure to any compound. This no-growth plate was used to calculate the cell growth in every well of the experiment plate by subtraction of the no-growth values to the 72 h time values of the experiment plate. Experiments for every cell line were performed in triplicate.
Cells were fixed by addition of 100 mL of cold 30% (w/v) trichloroacetic acid into every well and incubated at 4uC for 1 h. Plates were then washed with water four times and dried. Protein dying was carried out with 100 mL per well of 0.057% (w/v) sulforhodamine B in 1% (v/v) acetic acid, kept at room temperature for 30 minutes. After that, excess of sulforhodamine B was removed by quickly rinsing the plates four times with 1% (v/ v) acetic acid. After drying, sulforhodamine B was redissolved by adding 100 mL unbuffered Tris base solution to each well and shaking for 45 minutes on a gyratory shaker. Sulforhodamine B was quantified by its absorbance measured at 510 nm [22].

Docking Studies
Crystal structure preparation. The crystal structure of topotecan bound to the topoisomerase 1-DNA covalent complex was downloaded from Protein Data Bank (PDB code 1K4T) [23]. Its choice as the starting point was dictated by the fact that topotecan is the only camptothecin in clinical use that has been crystallized in its site of action. To prepare the crystal structure for docking, firstly an atom of Hg, a molecule of PEG and the two forms of topotecan (the closed lactone and the opened carboxylate) were deleted; later the Dockprep tool of the UCSF Chimera package [24] was used. This tool allows to automatically delete water molecules, add hydrogens to both the enzyme and the DNA, repair truncated side chains (Dunbrack rotamer library [25]) and assign Gasteiger charges with the AMBER ff12SB force field [26]. The input file for docking was generated with AutoDock Tools 1.5.6 [27]. The docking site was defined as a box with dimensions 15615620 Å , whose centroid was calculated using the coordinates of the closed lactone form of topotecan (x = 21.377, y = 24.068, z = 28.192).
Ligand preparation. Individual PDB files of ligands were prepared by ab initio energy minimization with Spartan '10 at the 3-21G level. Hydrogens were added to all ligands and the root of torsion tree was detected. The input file for docking was generated with AutoDock Tools 1.5.6.
Docking studies. Docking was performed with AutoDock Vina [28] using the same parameters described above. The best pose for each ligand was selected and analyzed. To validate the docking protocol, we docked the closed lactone form of topotecan into the target binding site, after preparing both files as described above. The best resulting pose was compared with the pose of the crystal structure, and a RMSD value of 0.492 Å was calculated for the differences between the positions of the atoms in both pentacyclic scaffolds.

Synthesis
As mentioned in the Introduction, the planned Friedlä nder disconnection of the target compounds 3 required the preparation of two kinds of starting materials, compounds 1 and 2. Compound 2, which can be regarded as an oxidised derivative of the alkaloid vasicinone, was prepared using a literature method [29]. Some of the compounds 1, namely 1a-d, were commercially available, while access to 1e was secured by a literature method involving the bromination of 1c with potassium bromide under oxidative conditions [20]. Finally, compounds 1f and 1g were prepared by treatment of the corresponding derivatives of 2-aminobenzonitrile with 3,5-dimethylphenylmagnesium bromide followed by acidic hydrolysis (Figure 2).
With compounds 1 and 2 in hand, we studied their Friedlä nder coupling under conditions previously established by our group [19] that involved heating of the starting materials in refluxing ethanol in the presence of cerium(IV) ammonium nitrate (CAN) as a Lewis acid catalyst [30] (Figure 3). As shown in Table 1, the reaction proceeded in good to excellent yields and furnished a small library of luotonin analogues with a variety of substituents at the A and B rings. Compound 3b had been previously prepared by a related method, albeit as a side product in a poor 14% yield, and was not tested as a topoisomerase 1 inhibitor [31].
During the initial optimization studies, we carried out two of the reactions for shorter reaction times (3 h) and obtained the open luotonin analogues 4, finding subsequently that they were transformed into the corresponding final products 3 by use of longer reaction times. This result proves that compounds 4 are intermediates of the CAN-catalyzed Friedlä nder reaction, a conclusion that is of some mechanistic interest because there is not a general agreement on whether the Friedlä nder reaction is initiated by an aldol reaction or by the formation of a menaminone like compounds 4 ( Figure 4) [32].

Pharmacological Studies
DNA relaxation assay. A representative electrophoresis gel showing the separation of DNA topomers after incubation of the plasmid in the presence of human topoisomerase 1 and the inhibitors under assay is shown in Figure 5.
When normalised to the activity of CPT, luotonin A (3a) showed a lower inhibitory activity (++), as expected. Interestingly, all the newly synthesized analogues exhibited an inhibitory activity similar or better than that of the reference compound, luotonin A (3a). Compounds 3b, 3c and 3e exhibited an activity similar to that of 3a (++). The chlorine-containing compound 3d and the dimethylphenyl derivative 3f showed an inhibitory potential (+++) higher than the lead compound 3a. Finally, the methyl dimethylphenyl derivative 3g demonstrated an excellent activity and achieved the best poisoning capacity (++++), which was close to that of camptothecin. This level of potency is remarkable in that it has very rarely been found in the inhibition of human topoisomerase by luotonin analogues [16].
In order to rationalize the high activity as topoisomerase poisons of some of our luotonin analogues, and also with a view of using the information thus obtained for planning improved analogues in the future, we undertook in silico studies. There is no crystal structure available for the luotonin-DNA-topoisomerase 1 ternary complex and, furthermore, to the best of our knowledge, this is the first time that docking of luotonin A derivatives onto the DNAtopoisomerase 1 complex has been carried out. For this reason, we started this part of our work by studying the behaviour of luotonin A itself (3a), the lead compound in this series. As shown in Figures 6A and 6B, it stacked between the base pairs (21) and (+1) and also gave hydrogen bonding interactions between N5 and Arg364. A comparison between the docking pose of luotonin with the crystal structure of the complex formed by topotecan showed that the orientation of both molecules was the same. The docking of compound 3c is also shown (Figure 6C), showing a similar  docking pose but a stronger interaction due to hydrogen bonding of its two nitrogen atoms, N5 and N6. This is a feature common to all our analogues, as will be discussed below.
The affinity of the examined molecules was quantified as the free-energy difference between the binary DNA-top1 complex and the ternary complexes formed upon binding of the drugs (''scores'' in Table 2). Importantly, a good correlation was found between the ability of the molecules to stabilize the binary complex and their in vitro topoisomerase 1 inhibitory effects, as shown in Table 2. It is interesting to note that the two compounds bearing 3,5dimethylphenyl substituents (3f and 3g), which were the most potent of the series, were also the ones showing the highest scores.
Our docking studies revealed that the binding modes of the molecules under study could be grouped into two different types. Derivatives 3a-3e ( Figure 7A) bound the target site in the same fashion as the original crystallographic topotecan. All these compounds stacked between the (21) and (+1) base pairs, perpendicular to the DNA main axis and with the two nitrogens at positions 5 and 6 facing the minor groove of the DNA. The aromatic rings in the pentacyclic structure established p-p stacking interactions, within a 3.6 Å range, with the four DNA bases flanking the molecule. A single direct protein-drug interaction was detected through hydrogen bonding between the Arg364 residue and the two basic nitrogen atoms in the molecule. Only in the case of halogenated compounds 3d and 3e a third class of interaction was found, as halogen bonding was possible between the chlorine or bromine atoms and the Glu356 side chain. This type of bonding is a subset of the so-called m-hole bonds [33] and can play a significant role in molecular recognition in biological environments, as often evidenced by crystallographic studies of ligandbinding site interactions [34]. All compounds in this group pointed their E rings towards the scissile strand, although in the case of compound 3b, another pose was found with a similar score (2 11.7 vs. 211.6 kcal/mol) that stacks the opposite way, with the ring A on the scissile strand side.
Interestingly, compounds 3b-3e showed in all cases a better affinity (''score'') than luotonin itself in our computational studies. Upon examination of the 3D docking models, this was attributed to the fact that luotonin binds to the Arg364 residue by a single hydrogen bond by the nitrogen belonging to its quinoline fragment, while in the other cases the quinoline and quinazoline fragments are both within hydrogen bonding distance of Arg364. An overlay of all five compounds at their binding sites ( Figure 7A) clearly shows that the docking pose of luotonin A is different from the others. This difference may be attributed to the steric effect of the R 1 substituent in compounds 3b-3e, which would give repulsive interactions with Asn352 if their binding poses were superimposable with that of luotonin A (Figures 7B and 7C).
Compounds 3f and 3g showed a different binding mode ( Figure 8B). Both molecules stacked between the (21) and (+1) base pairs, but the orientation of the rings was modified. Both molecules bear a 3,5-dimethylphenyl substituent at the C-14 position of ring B that confers them a high lipophilicity [35]. Although the topotecan-like binding mode previously described allows the accommodation of large groups on position 14 of the luotonin B ring [36], nevertheless such substituents remain exposed to the highly polar environment. In this regards, it is worth noting that solvent accessible surface area for compounds 3f (530.45 Å 2 ) and 3g (562.76 Å 2 ) is at least 10% higher than the values calculated for the other luotonin derivatives [37]. The binding mode of 3f and 3g almost totally shields the hydrophobic dimethylphenyl moiety from the aqueous environment by burying  it between the base pairs of the scissile strand and the surface of the protein, more specifically between Arg364 and the deoxyriboses belonging to G(+1) and T (21). This minimizes the entropic effect of an ordered layer of water molecules covering the dimethylphenyl moiety [38] and enhances the stability of the ternary complex ( Figure 8C). As a consequence of this binding mode, the hydrogen bonds between the basic nitrogens and Arg364 are lost. Nevertheless, this is compensated by the formation of a new, shorter hydrogen bond between the D-ring carbonyl and the basic amino acid. The dimethylphenyl group ceases to be orthogonal to the pentacyclic system in order to achieve a better fit with this binding pocket.
The main ligand-binding site interactions found in our docking studies are summarized in Table 3.
Cell culture and in vitro anti-proliferation assays. The in vitro cytotoxic activity of the studied compounds was deduced from the percentage of cell growth observed when several cell cultures were exposed to 25 mM drug concentration, normalized with regard to the negative control (DMSO-treated), with the results shown in Figure 9. In the case of HeLa cells, cell growth  The fact that the topoisomerase inhibition data showed a less than perfect correlation with the cytotoxicity results can be attributed to the participation in the latter of physicochemical characteristics and pharmacokinetic phenomena such as solubility and transport across membranes.
When examined together with the results from the topoisomerase 1 inhibition assay, the outcome from the cell cytotoxicity experiments supports the conclusion that compounds 3f and 3g are the most promising therapeutic candidates, showing a much higher potency than the reference luotonin A in the three cell lines assayed. Also compounds 3d and 3e, which showed a moderate activity in the topoisomerase 1 inhibition experiments, have been revealed as interesting alternatives, with good results in the colon adenocarcinoma cell line. The high selectivity of these chloro-and bromo-derivatives towards SW480 cells is remarkable and shall be the subject of further research.

Conclusions
The CAN-catalyzed Friedlä nder reaction provides a reliable and efficient method for the synthesis of luotonin A analogues modified at their A and B rings. All analogues showed an activity similar or higher in topoisomerase 1 inhibition in comparison to the reference compound luotonin A, and some compounds even had an activity comparable to camptothecin, which is an unusually good result for series of analogues of this alkaloid. Furthermore, almost all compounds showed a better activity than luotonin A in cell cytotoxicity assays. In order to rationalize these results, docking studies of luotonin A-topoisomerase 1-DNA ternary complexes were undertaken for the first time, and led to the conclusion that most compounds bound in a manner similar to standard topoisomerase poisons such as topotecan and also to luotonin A, although the presence of the substituent at the C-14 position of ring B induced in most cases a better fit to the binding site due to improved hydrogen bonding to Arg364. Interestingly, the two most promising compounds, bearing a 3,5-dimethylphenyl substituent at ring B, docked in a different orientation, with this In the case of compounds 3a-e, the docking pose was similar to the one found for topotecan by X-Ray crystallography. B. Compounds 3f and 3g, on the other hand, changed their mode of binding in order to shield the highly hydrophobic dimethylphenyl moiety from the aqueous environment by burying it between the base pairs of the scissile strand and the surface of the protein. C Detail of the newly described binding feature of the human topoisomerase 1-DNA complex uncovered by our study. doi:10.1371/journal.pone.0095998.g008 Table 3. Main interactions located for compounds 3a-g at their binding sites.
hydrophobic moiety shielded from the aqueous environment by being buried between the deoxyribose belonging to the G(+1) nucleotide in the scissile strand and the Arg364 on the surface of the protein, additionally forming a hydrogen bond between the Dring carbonyl and the basic amino acid. The discovery of this new binding mode and its associated higher inhibitory potency is a significant advance that will find application in the design of new topoisomerase 1 inhibitors.

Supporting Information
File S1 SI1, 1 H and 13 C -NMR spectra of all compounds. Figure  S1, Two views of compound 3a (luotonin A) docked in the topoisomerase 1 active site. Figure S2, Two views of compound 3b docked in the topoisomerase 1 active site. Figure S3, Two views of compound 3c docked in the topoisomerase 1 active site. Figure S4, Two views of compound 3d docked in the topoisomerase 1 active site. Figure S5, Two views of compound 3e docked in the topoisomerase 1 active site. Figure S6, Two views of compound 3f docked in the topoisomerase 1 active site. Figure S7, Two views of compound 3g docked in the topoisomerase 1 active site. (ZIP) Figure 9. Results of the cytotoxicity assays. The cytotoxicities of compounds 3a-g are expressed as percentage of cell growth relative to the negative control. Error bars account for the standard error. Asterisks indicate statistically significant differences between a drug and the model compound 3a. Camptothecin was employed as a positive control (0% growth) at 25 and 2.5 mM concentrations. doi:10.1371/journal.pone.0095998.g009