Tetrahydroisoquinoline Sulfamates as Potent Microtubule Disruptors: Synthesis, Antiproliferative and Antitubulin Activity of Dichlorobenzyl-Based Derivatives, and a Tubulin Cocrystal Structure

Tetrahydroisoquinoline (THIQ) 6-O-sulfamate-based anticancer agents, inspired by the endogenous steroid 2-methoxyestradiol and its sulfamate derivatives, are further explored for antiproliferative and microtubule disruptor activity. Based on recently designed C3-methyl C7-methoxy-substituted THIQ derivatives, compounds with mono- and dichloro-substitutions on the pendant N-benzyl ring were synthesized and evaluated. Although improved antiproliferative activity was observed, for example, 4a versus 4b and 4b versus 8c, it was relatively modest. Compound 8c, a 2′,5′-dichlorobenzyl derivative was, however, identified as a promising antiproliferative agent with in vitro activities exceeding that of the parent steroid (e.g., GI50 90 nM in DU-145 cells) and was highly potent against a range of tumor cell lines (e.g., GI50 26 nM for OVCAR-3). 8c inhibited the polymerization of tubulin in vitro with an IC50 only twofold less potent than combretastatin A-4 and inhibited colchicine binding to tubulin. Tubulin polymerization assays showed the parent THIQ 4a to be only a very weak inhibitor, but a striking potency difference was seen between compounds with C2′ methoxy and chloro substituents, whereas this was much smaller when these substituents were positioned at C5′. To confirm the target in atomic detail and because 8c is a racemic mixture, an achiral parent THIQ 6-O-sulfamate derivative 10 was successfully cocrystallized with the αβ-tubulin heterodimer. The derivative 10 binds at the colchicine site on tubulin, the first example of this compound class investigated in such detail, with its sulfamate group interacting with residues beyond the reach of colchicine itself, similar to a recently reported quinazolinone sulfamate derivative, 6a. The structure also suggests that for racemic C3-methyl-substituted THIQ derivatives, such as 8c, the (S)-enantiomer is likely to be preferentially accommodated within the colchicine site for steric reasons. The results further confirm the potential of nonsteroidal THIQ sulfamate derivatives for oncology and suggest that the mechanism of microtubule destabilization for the THIQ compound class is to prevent the curved-to-straight conformational transition of tubulin required for polymerization.


■ INTRODUCTION
In previous studies, we described N-benzyl-substituted tetrahydroisoquinoline (THIQ) derivatives as novel microtubule disruptors with potential therapeutic application for the treatment of cancer. 1−5 These compounds were designed to mimic the 2-substituted estratriene class of microtubule disruptors derived from the endogenous steroid 2-methoxyestradiol (2ME) and, in particular, its sulfamate derivatives (e.g., 2a (STX140), Figure 1). 6−10 Incorporation of a phenolic 3-O-sulfamate group is generally observed to be highly beneficial for both activity and oral bioavailability, and STX140 is an optimized anticancer agent with both 3-O-and 17-O-sulfamate groups. 11 Agents possessing such sulfamate esters have reached multiple clinical trials in oncology and elsewhere, primarily for hormone-dependent diseases. 11,12 The sulfamoyl group can serve in diverse roles. 12 STX140 was developed primarily for hormone-independent cancer applications, and the nonsteroidal THIQ core was used as a mimic of the steroidal AB-ring system from which steroidomimetics were constructed. Such compounds have markedly better physicochemical properties than 2ME and even STX140 but can maintain in vivo potency. Substitution of the THIQ nucleus at C6 and C7, with those groups requisite for activity in the steroidal series, was thus desirable.
Attachment at N2 of a group projecting into the area of space occupied by the steroidal D-ring and bearing an H-bond acceptor known to be required for optimal activity completed the prototypical steroidomimetic design. 1,2,11 Further optimization led to THIQs methylated at C3 forcing a more favorable conformation of the steroidal D-ring mimic by steric repulsion (THIQs of type 3 and 4, Figure 1). 3 Related THIQ core chimeric microtubule disruptors possessing the trimethoxy aryl motif common to many tubulin colchicine site binders were designed and then further optimized. 4,5 As a result, the best compounds in both series displayed nanomolar in vitro activities, making them equipotent or better than the steroidal derivatives that underpinned their design. 3−5 Most recently, we designed another steroidal AB-ring mimic by using a dihydroquinazolinone (DHQ) core structure. 13 The most potent DHQs derived from this series were also sulfamates of type 6 ( Figure 1) with the best compounds again displaying nanomolar in vitro activities.
Like the steroids, the most potent heterocyclic derivatives were found to disrupt the polymerization of tubulin and inhibit the binding of [ 3 H]colchicine to tubulin. THIQ sulfamates also inhibit carbonic anhydrases (CA) II, 4 an interaction believed to contribute to the high oral bioavailability observed for the steroid derivatives, 14,15 and CAIX, 16 which affords extra targeting of hypoxic tumors. In addition, the activity of nonsteroidal THIQ derivatives mirrors that of the steroidal series, with compounds capable of inhibiting the growth of taxane-resistant cancer cells 17−19 and human umbilical vein endothelial cell proliferation (a commonly used marker for antiangiogenic activity), 20,21 thus supporting the idea that these small-molecule agents work in a similar manner to 2ME and its bis-O-sulfamate derivative 2a (STX140).
Very recently, the mechanism of these heterocyclic sulfamate esters in the destabilization process of microtubule polymerization was investigated for the first time in atomic detail. The crystal structure of the most potent DHQ sulfamate of type 6, the N-(2′,5′-dimethoxybenzyl) derivative 6a, in complex with tubulin revealed molecular details to support further strategies for structure-based optimization. Interestingly, the sulfamate group was shown to play a role in binding at the colchicine site of tubulin. The mechanism of microtubule destabilization, at least for this particular DHQ 6a, was established as preventing the curved-to-straight tubulin conformational transition required for microtubule formation. 13,22 DHQ derivatives, although highly effective, are not charged at the linking N-atom. THIQ derivatives likely have more developmental potential, their charged nature in particular affording better solubility via salt formation at the N2 position. Moreover, they have been shown to be orally active in vivo when dosed in an aqueous buffer, with some being as potent as STX140. 4 It was thus of interest to explore the potential for further optimization of such derivatives via new substitutions in the pendant aryl ring and, moreover, if this THIQ class could also be cocrystallized with the αβ-tubulin heterodimer, such that atomic level information could be obtained, for further structure−activity relationship (SAR) development. In the present work, we explore the replacement of methoxy groups present in the steroidal D-ring mimic of C3-methylated THIQs with chloro substituents to determine whether such less polar and more lipophilic H-bond acceptors improve microtubule disruptor potency ( Figure 1). Therefore, we introduced 2′,3′-, 2′,4′-, and 2′,5′-dichlorobenzyl groups at N2 to explore SAR in a focused series of phenols and their sulfamate esters and explored their antiproliferative and antitubulin activities.
Biology. All six compounds were evaluated at the US National Cancer Institute (NCI) in the full 60-cell-line assay that allows activity across a wide range of cancer types to be assessed (Tables 1 and 2). The in vitro activities of these compounds against DU-145 prostate and MDA MB-231 breast cancer cell line proliferation are shown in Table 1. Data from both cell lines are in strong agreement. Overall, compounds in this small series of N2-dichlorobenzyl C3-methyl-substituted THIQs exhibit activities in the micromolar and nanomolar ranges. Phenols 7b−c are only about 3−5-fold less active than their corresponding sulfamates 8b−c against the proliferation of DU-145 prostate cancer cells and MDA MB-231 breast cancer cells. The exception is phenol 7a that proved to be more than 10-fold less active than its corresponding sulfamate, 8a. GI 50 values of the sulfamates 8a−c range between 90 nM (8c; DU-145) and 551 nM (8a; MDA MB-231). The best

ACS Omega
Article THIQ 8c of this small SAR screening set proved slightly more potent or very similar to previously reported THIQs, 4a−b. 3 However, data from the NCI-60-cell-line panel for 4a−b against DU-145 and MDA-MB 231 cells suggested that 8c is significantly more potent than 4a, but only about as potent as 4b. Overall, the antiproliferative activity was significantly improved when the 2′-methoxy group was replaced with a 2′chloro substituent. However, no further enhancement was observed when the 5′-methoxy group too was replaced with a 5′-chloro substituent. As these (±)-C3 methyl-substituted THIQs adopt a steroid-like conformation with the substituent at C5′ pointing into the same area of space as the steroidal C17 β-hydroxyl group, 1−3 it seems likely that the substituent at C2′ will sterically interact with the two hydrogens at C1. Therefore, only relatively small one-atom substituents such as chlorine seem well tolerated at the C2′ position. Note that the hydrogen and fluoro substituents at C2′ in combination with a C5′ methoxy group were also very well tolerated. 3 It seems that the only marginally larger C2′ methoxy group could already force too much of an angle between the THIQ B-ring and the pendant N2-benzyl ring, resulting in a less favorable overall conformation, leading to a decrease in activity (4a; MDA MB-231: 2.1 μM). However, there seem to be less restrictions for the second H-bond acceptor at C5′ and more space might be available around that binding site. Note, of course, that in all cases here, the activities of racemic mixtures are being compared, with all usual caveats, but vide infra for further discussion on this point.
Additionally, data from six more tumor cell lines are presented along with the mean activity across the whole panel (MGM value; Table 2). The trend that emerged for these THIQ derivatives against the proliferation of DU-145 prostate cancer cells and MDA MB-231 breast cancer cells in vitro was essentially confirmed across the whole NCI 60-cell-line assay. It also confirms the potential of these N-dichlorobenzyl compounds against a broad range of cancer phenotypes with 8c, in particular, proving highly active (e.g., 26 nM in OVCAR-3; MGM = 129 nM). Over the whole 60-cell-line panel, compound 4b was slightly more potent (MGM = 102 nM; Table 2). Additionally, the trend mentioned above with regard to 4a−b and 8c seems to hold well. Overall 4b and 8c were about 4−5 times more potent than 4a.
To confirm the target for these, we established the activity of the sulfamoylated compounds, 8a−c as microtubule disruptors, alongside the established potent clinical microtubule disruptor, combretastatin A-4 (CA-4), initially in tubulin polymerization assays and alongside the previously reported THIQ sulfamate, 10 5 (Table 3). Gratifyingly, the 2′,5′-dichlorobenzyl derivative 8c very effectively inhibited the assembly of tubulin, with an IC 50 of 0.98 μM, being only about 2-fold less potent than CA-4. The IC 50 in these tubulin-based assays, as is typical with antitubulin agents with potent cytotoxic activity, far exceeds

ACS Omega
Article the antiproliferative GI 50 dose, at least in part because the GI 50 values refer to the concentration in the media, not in the cells. Inhibition of colchicine binding to tubulin for these compounds was also determined relative to CA-4 for further proof of targeting, with 8c being the overall best THIQ derivative to date showing 67% inhibition at 5 μM (Table 3). It is reasonable to propose that these novel THIQs can, at least partially, be assumed to disrupt normal dynamic tubulin polymerization by interacting at the colchicine site. Also, this provides further evidence that methoxy groups as polar Hbond acceptors are not an essential requirement in the D-ring mimic. Overall, the tubulin assays showed a striking difference in potency between compounds with methoxy and chloro as the C2′ substituents (4a vs 4b), whereas the difference in potency is much smaller when these substituents are positioned at C5′ (4b vs 8c). But also, here, the C5′ chloro compound 8c displayed an improved IC 50 and inhibition of colchicine binding than the corresponding C5′ methoxy compound, 4b.
As in previously published work on DHQs, 13 we employed X-ray crystallography to determine the atomic level binding mode of a THIQ derivative within the αβ-tubulin heterodimer ( Figure 2A). To avoid complications that would arise from using the racemic sulfamate, 8c, we chose a highly potent nonracemic THIQ, [2-(3′-bromo-4′,5′-dimethoxybenzyl)-7methoxy-6-sulfamoyloxy-1,2,3,4-THIQSTX3451, 10], without the C3-methyl group chiral center that had been optimized earlier 5 and which is not too dissimilar to 8c in its antitubulin activities (Table 3). STX3451 exerts antiproliferative and antimitotic effects, inducing apoptosis and involving autophagic processes in MDA-MB-231 metastatic breast and A549 epithelial lung carcinoma cells 24 and also induces cell death effectively in NF1 tumor cell lines. 25 It is therefore somewhat more attractive as a potential development candidate than 8c.
Compound 10 was soaked into crystals formed by a protein complex composed of two αβ-tubulin heterodimers, the stathmin-like protein RB3 and tubulin tyrosine ligase (termed T 2 R-TTL), 26 and we solved the T 2 R-TTL-10 complex structure using X-ray crystallography to 2.4 Å resolution (Figures 2 and 3B, 4A; Table 4) (PDB ID 6HX8). STX3451 binds to the tubulin colchicine site 22 at the intradimer interface created by residues from the strands βS8 and βS9 the helix βH8 of β-tubulin, and the loop αT5 of α-tubulin ( Figure 4A). STX3451 is well accommodated within the binding site, displaying an extended "steroid-like" conformation. These data now provide the second cocrystal structure of an antitubulin agent that possesses a sulfamate group and the first one where the steroidal AB-ring system is mimicked by a THIQ core. Figure 4B shows a comparison of the new THIQ sulfamate binding mode within the colchicine site with that of the earlier DHQ sulfamate 6a. 13 The C-ring of STX3451 is buried into a hydrophobic pocket formed by the β-tubulin residues: βVal238, βCys241, βLeu242, βLeu255, βMet259, βAla316, βIle318, βAla354, βThr376, and βIle378 ( Figure 4A). Moreover, the 4′-methoxy oxygen of the C-ring forms a water-mediated hydrogen bond to the main chain amide and carbonyls of βCys241, βGly237, and βVal238. The A-ring of STX3451 is stacked between the side chains of βAsn258 and βLys352, and further stabilization is provided by hydrogen bonds between the sulfamate moiety and the side chains of βLys352, βAsn349, and αSer178, the main chain amide of αVal181, and the main chain carbonyl group of βAsn349. Compared to the DHQ derivative 6a 13 (PDB ID 5OSK), which possesses an amine group in the B-ring, no water-mediated interaction to the carbonyl groups of the side chain and main chain of αAsn101 and αThr179 is observed for STX3451 ( Figure 4B). Moreover, the 3′-bromo substituent of the C-ring in STX3451 forms a water-mediated hydrogen bond to the main chain carbonyl of βVal238, a space that is

ACS Omega
Article otherwise occupied by the 2′-methoxy group in 6a ( Figure  4B).
Recently, THIQ compound (±)-11 ( Figure 3A), which is the same as 10 but with an additional (±)-C3 methyl group as in the present 8c compound series, was synthesized and evaluated. 5 As it was found to be similar to 10 (STX3451) in antitubulin assays (IC 50 2.4 μM, 50% inhibition of colchicine binding 5 vs IC 50 1.3 μM and 49%, respectively) ( Table 3, but note the very slightly different assay conditions), we would suggest that the additional (±)-C3 methyl group of the present series should, in principle, be well accommodated, as presumably also the 2′,5′-dichloroaryl motif. This former feature was not an issue in the DHQ 6a instance (see Figure  4B) because a planar carbonyl group occupies the C3 position. A closer inspection of the C3 environment in the tubulin-STX3451 crystal structure reveals that there may be enough space to accommodate both enantiomers of compound 11 ( Figure 3A). However, we modeled compound 11 in both the (R)-and (S)-configurations into the STX3451 structure without energy minimization and inspected the C3 environment of both the enantiomers ( Figure 5). The model suggests that the (S)-enantiomer likely binds without any structural rearrangements, whereas the (R)-enantiomer would clash into helix βH8, thereby requiring structural adaptations to get accommodated into the pocket or even preventing binding.
For these reasons, we expect that administration of the racemate would preferentially select for the (S)-enantiomer binding to tubulin. It is tempting to view the respective antitubulin IC 50 values of (±)-11 and 10 in the light of this (2.4 and 1.3 μM) and speculate that (R)-11 might be a very weak inhibitor or even totally inactive. The antitubulin activity of racemic 3-methyl-substituted THIQs, such as 8c, may therefore actually be an underestimate if the (R)-enantiomer is not accommodated or only binds weakly in the colchicine site because of the structural adaptations necessary. Exploration of this possibility must however await the resolution of the isomers but, given the greater development potential of the achiral THIQ sulfamate 10, this has not been pursued. Further development of these THIQs through medicinal chemistry might include the introduction of a (±)-C4 hydroxyl group as it could attract a water-mediated interaction to the carbonyl groups of the side chain and main chain of αAsn101 and αThr179, as was observed for DHIQ, 6a. 13 We further compared the binding mode of STX3451 10 with that of colchicine by superimposition of both β-tubulin chains of their tubulin-ligand complexes ( Figure 6; PDB ID 4O2B; rmsd of 0.428 Å over 393 C α -atoms of chains B). Tubulin undergoes a "curved-to-straight" conformational change when polymerized into microtubules, 22,27 involving an overall compaction of the colchicine binding site through mainly the βT7 loop and βS8 strand of the β-tubulin chain. 22 As already observed for the DHQ derivative 6a, 13 both compound poses show similarity and reveal that the respective A rings and associated methoxy groups of STX3451 and colchicine are very well superimposed, whereas the more flexible C ring of STX3451 is more extended. Both compounds, STX3451 and 6a, share a unique and common polar interaction with the sulfamoyl group of the main chain of αVal181. As a result of the different shapes of the molecules (STX3451 vs colchicine), major conformational changes of both the αT5 and βT7 loops are observed ( Figure 6). In summary, the results have established STX3451 10 as a tubulin-binding ligand at the colchicine site in atomic detail. They also imply that 10, similar to other colchicine-site ligands, 22,28 achieves microtubule destabilization mechanistically by preventing the curved-to-straight conformational transition.

■ CONCLUSIONS
THIQ sulfamate-based microtubule disruptors have been further explored. Racemic N2-dichlorobenzyl C3-methylsubstituted THIQs showed excellent in vitro activities in the NCI-60-cell-line assays and antitubulin properties. Chlorine substitution for methoxy was found to be effective, and 8c in particular exhibits antiproliferative activity in the 90 nM range, inhibiting tubulin assembly and interfering effectively with the colchicine site. To explore tubulin binding at the atomic level without the complication of chirality, a parent THIQ STX3451 10 was cocrystallized successfully with the αβ-tubulin heterodimer. STX3451 binds more deeply in the colchicine site than colchicine itself, and the sulfamate group is involved in binding through specific interactions with β-tubulin residues beyond those accessed by colchicine. This is the first example of a THIQ derivative bearing a sulfamate ester bound to tubulin to be explored in such detail. With subtle differences, 10 adopts a broadly similar binding pose to a related DHQ derivative. Structural implications derived from the ligand are that C3 methyl group substitutions, as in the dichlorobenzyl Highest shell statistics are in parentheses. b CC 1/2 = percentage of correlation between intensities from random half-datasets. 29 c As defined by MolProbity. 30

ACS Omega
Article derivatives, can be accommodated in the colchicine site. Modeling of (R)-11 and (S)-11 into the STX3451-tubulin structure further revealed that the (S)-enantiomer might be mainly responsible for the tubulin activity of such racemic C3 methyl-THIQ sulfamates as the C3 methyl group of the (R)-enantiomer would sterically clash with tubulin residues. Most likely, mechanistically microtubule destabilization by STX3451 is via prevention of the curved-to-straight conformational transition of tubulin, as shown for colchicine and the STX3451-related quinazolinone DHQ derivative. With the usual caveats with regard to chirality, such compounds of this THIQ class are worthy development candidates for oncology.  ) with the final 2mFo-DFc (blue) and mFo-DFc (green and red) electron density maps contoured at 1.0σ and ±3.0σ, respectively. The green electron density blob marked with an arrow suggests an alternate conformation of the 5′-methoxy group of the C-ring. (D) Sigma A-weighted mFo-DFc (light green mesh) simulated annealing electron-density omit map contoured at 3.0σ. The map was calculated excluding the atoms corresponding to STX3451 only. Both the STX3451 molecule (aquamarine sticks) and the two water molecules (red spheres, HOH) of the final refined structure are superimposed to highlight the quality of the map.   Figure 4A. The structures were superimposed onto their β1-tubulin chains.

ACS Omega
Article ■ EXPERIMENTAL SECTION Tubulin Assays. Bovine brain tubulin was prepared as previously described 31 and used in the studies presented. Assembly IC 50 's were determined as fully described elsewhere. 32 In brief, 1.0 mg/mL (10 μM) tubulin was preincubated with varying compound concentrations without GTP for 15 min at 30°C. Reaction mixtures were placed on ice, and GTP (final concentration, 0.4 mM) was added. Mixtures were transferred to cuvettes held at 0°C in a recording spectrophotometer. Baselines were established at 0°C and increase in turbidity was followed for 20 min following a rapid (<30 s) jump to 30°C. Concentrations of compound required to reduce the turbidity increase by 50% were determined. Inhibition of the binding of [ 3 H]colchicine to tubulin was described fully previously. 33 Reaction mixtures contained 0.1 mg/mL (1.0 μM) tubulin, 5.0 μM [ 3 H]colchicine, and a potential inhibitor at 5.0 μM. Compounds were compared to CA-4, a particularly potent inhibitor of colchicine binding to tubulin. 34 Mixtures were incubated for 10 min at 37°C, a time point at which colchicine binding in control reaction mixtures is generally 40−60% complete.
Crystallization, Data Collection, and Structure Solution. Crystals of T 2 R-TTL were generated as previously described. 25,35 Crystals were soaked for 3 h at 20°C in a reservoir solution (10% PEG 4K, 16% glycerol, 30 mM MgCl 2 , 30 mM CaCl 2 , and 0.1 M 2-(N-morpholino)ethanesulfonic acid/imidazole pH 6.7) containing 5 mM of compound 10 (STX3451) and subsequently transferred into a reservoir supplemented with 20% glycerol before cryo-cooling in liquid nitrogen. All data were collected at the Swiss Light Source (beamline X06DA, Paul Scherrer Institut, Villingen PSI, Switzerland) with images indexed and processed using XDS. 36 Structure solution using the difference Fourier method and refinement were performed using PHENIX 37 and model building was carried out iteratively using Coot software. 38 The atomic coordinates and structure factors have been deposited in the Protein Data Bank (www.rcsb.org) under accession number 6HX8 (T 2 R-TTL-STX3451).
Structural Analysis and Figure Preparation. Molecular graphics and analyses were carried out using PyMol (The PyMOL Molecular Graphics System, Version 1.8.6.2. Schrodinger, LLC).
Chemistry. All chemicals were either purchased from Aldrich Chemical Co. (Gillingham, UK) or Alfa Aesar (Heysham, UK). Organic solvents of A.R. grade (PE, EtOAc, CHCl 3 , acetone, and CH 2 Cl 2 ) were supplied by Fisher Scientific (Loughborough, UK) and were used as supplied. Petroleum ether (PE) used for crystallization was of fractions 40−60°C. DMA and DMF were purchased from Aldrich and stored under a positive pressure of N 2 after use. Sulfamoyl chloride was prepared using an adaptation of Appel and Berger methodology 39 and was stored in the refrigerator as a solution in toluene under a positive pressure of N 2 , as described by Woo et al. 40 An appropriate volume was freshly concentrated in vacuo immediately before use. Compound 9 was synthesized according to a literature procedure. 3 Reactions were performed at room temperature unless stated otherwise. Flash column chromatography was carried out on silica gel (MatrexC60). 1 H NMR spectra were recorded with a Varian Mercury VX 400 NMR spectrometer and a Bruker AVIII HD 400 spectrometer at 400 MHz. 13 C NMR spectra were recorded with a Bruker AVIII HD 400 spectrometer at 100 MHz. Chemical shifts are reported in parts per million (ppm) relative to the solvent residual peaks as internal standards. 1 H NMR: 7.26 ppm (CDCl 3 ); 2.05 ppm (acetone-d 6 ). 13 C NMR: 29.84 ppm (acetone-d 6 ). High resolution mass spectrometry was performed using a Bruker microTOF electrospray ionization mass spectrometer. Compounds were ≥96% pure by reversed-phase HPLC run with CH 3 CN/H 2 O or MeOH/ H 2 O (Sunfire C18 column, 4.6 × 150 mm, 3.5 μm pore size).