Preclinical Lead Optimization of a 1,2,4-Triazole Based Tankyrase Inhibitor

Tankyrases 1 and 2 are central biotargets in the WNT/β-catenin signaling and Hippo signaling pathways. We have previously developed tankyrase inhibitors bearing a 1,2,4-triazole moiety and binding predominantly to the adenosine binding site of the tankyrase catalytic domain. Here we describe a systematic structure-guided lead optimization approach of these tankyrase inhibitors. The central 1,2,4-triazole template and trans-cyclobutyl linker of the lead compound 1 were left unchanged, while side-group East, West, and South moieties were altered by introducing different building blocks defined as point mutations. The systematic study provided a novel series of compounds reaching picomolar IC50 inhibition in WNT/β-catenin signaling cellular reporter assay. The novel optimized lead 13 resolves previous atropisomerism, solubility, and Caco-2 efflux liabilities. 13 shows a favorable ADME profile, including improved Caco-2 permeability and oral bioavailability in mice, and exhibits antiproliferative efficacy in the colon cancer cell line COLO 320DM in vitro.


■ INTRODUCTION
Tankyrase 1 and tankyrase 2 (TNKS1/2) are members of the PARP family of enzymes that control protein activities, interactions, and turnover through mono-or poly-ADPribosylation. 1 TNKS1/2 regulate a number of target proteins, including AXIN1 and AXIN2 (AXIN1/2) in the β-catenin destruction complex resulting in WNT/β-catenin signaling pathway inhibition, and AMOT proteins in the Hippo signaling pathway resulting in YAP signaling inhibition. 1−3 Tankyrases, through their ankyrin repeat clusters, bind to AXIN1/2, making AXIN1/2 accessible for ADP-ribosylation by the C-terminal TNKS1/2 catalytic domain. 1 AXIN1/2 is subsequently targeted for proteasomal degradation through polyubiquitination of E3 ubiquitin ligase RNF146, recognizing the poly-ADP-ribose signal. 1,2 Destabilization of AXIN1/2, being a structural protein in the β-catenin destruction complex, leads to increased β-catenin levels which can be counteracted by inhibition of TNKS1/2 catalytic activity. 1,2 Similarly, TNKS1/2 control the stability of AMOT proteins via RNF146. Stabilization of AMOT proteins by inhibiting TNKS1/2 activity sequesters YAP to the cytoplasm and prevents target gene expression driven by YAP in the nucleus. 1,3 TNKS1/2 catalytic activities also interfere with other biological mechanisms and cell signaling pathways such as vesicle transport, energy metabolism, telomere homeostasis, and mitotic spindle formation and affect components in AKT/ PI3K and AMPK signaling pathways. 1,4−6 Several groups of chemical substances have been identified that inhibit TNKS1/2 by binding to the substrate NAD + binding site either by occupying a nicotinamide pocket, adenosine binding pocket or by addressing both of them. 2,7−20 Although the catalytic domains of 17 human ARTD/PARP enzymes are homologous, unique features in the TNKS1/2 catalytic domain allow the development of tankyrase-selective chemical inhibitors. 1 Despite this progress, there is currently no viable selective TNKS1/2 inhibitor in clinical testing or practice for any application including targeting the WNT/βcatenin and YAP signaling pathways in cancer therapy. 17,21−23 It has been shown that TNKS1/2 inhibitors can exhibit anticancer efficacy in mouse models, either as monotherapy against colorectal cancer 8,24 and osteosarcoma 25 or in combination therapies together with PI3K and EGFR inhibitors against colorectal cancer 26 or with PD-1 inhibition against melanoma. 27 Two reports indicate intestinal toxicity 24 and bone loss 28 in mouse models upon treatment with early lead-stage tankyrase inhibitors, while other reports do not document signs of toxicity, intestinal injury, or body weight changes. 8,26,27,29 Hence, there is a continued need for the development of safe drugs directed toward TNKS1/2 and the WNT/β-catenin signaling pathway with improved chemical and biophysical properties. 17,21−23 The compound optimization described here is based on the understanding of the structure−activity relationship, crystallography, and physicochemical properties of our previous 1,2,4triazole analogue series JW74, 30 G007-LK, 9 and OD336 (1). 11 The optimization focused especially on the solubility and atropisomerism liabilities of the former G007-LK and 1 compounds, respectively. In our work we developed a novel series of compounds reaching picomolar IC 50 activity in a cellular WNT/β-catenin signaling reporter assay. Lead compound OM-1700 (13) within the novel series displays high potency and specificity and has overall favorable ADME properties compared to benchmark tankyrase inhibitors.

■ RESULTS AND DISCUSSION
Chemistry. For the synthesis of novel structures in the optimization campaign, we embraced a building block approach (Figure 1). Herein we were able to prepare all compounds following the same synthetic route, simplifying synthesis efforts (Scheme 1). Cyclic amide/urea/carbamate East modifications, however, required a different route (Scheme 2). Since the 1,2,4-triazole as a central scaffold was well established in our previous research, 9 it was left unchanged in the present lead optimization process. In the linker area between the 1,2,4-triazole and the East moieties, we synthesized a series of analogues with a bicyclo[1.1.1]pentane configuration and one compound with a cis-cyclobutane setup. These linker variations were synthesized according to the same scheme as for the default trans linker. For further optimization, the trans-cyclobutyl linker was left unchanged for the majority of the target molecules as it proved to be superior to the tested alternative linker iterations. 11 For compounds having the benzimidazolone West moiety of 1, synthesis was performed as depicted in Scheme 1a and described in our previous work. 11 For East-side variations, a slightly different route enabling East variations in the last step departing from amine G was used (Scheme 1b).
As a first iteration, we replaced the East benzimidazolone group as in our experience this group can result in unwanted solubility, permeability, and efflux properties. When suitable East-amides were identified as a replacement for the benzimidazolone group, a wide range of further options for East-side iterations opened enabling fine-tuning of physicochemical properties. Next, we replaced the West-pyrimidine as this group renders the molecule vulnerable for CYP-mediated oxidation which was confirmed by Med-Id studies of 1.
To synthesize a broad set of targets, we optimized the triazole-forming reaction from E to G (Scheme 1b) from the existing method (TFA, DMA, 120°C, 14 h, 10−21% yield). Here, we found that heating of E and 3a in 1-butanol at 80− 140°C for 5−20 h, depending on the actual substrate, typically resulted in 60−90% yield. Under these conditions, a broader scope of South and West moieties was tolerated in the reaction. All compounds in the present study were prepared accordingly except the benzisoxazolone 18 and lactam 51 (Scheme 2). For compounds with the cis-cyclobutane (75) and bicyclo[1.1.1]pentane (19) moieties, the corresponding hydrazides 3b and 3c respectively were used (Scheme 1b).
Biological Evaluation. All compounds were tested using a TNKS2 biochemical assay and a luciferase-based WNT/βcatenin signaling pathway reporter assay in human HEK293 cells. 9 In the first round of the optimization campaign, we prepared single-point modifications changing any of the four regions in 1 (Figure 1) while leaving the other regions constant. In the following stages of the optimization campaign, additional East, West, and South moieties were also utilized, combining the best structural elements of the first single-point modification round. For the optimization, in vitro ADME properties and solubility of selected compounds were measured. Mouse pharmacokinetics, after oral dosing, was tested for the selected and short-listed compounds 13, 16, and 27 ( Figure 2c and indicated by # in Tables 1 and 2).
Since 1 displayed low solubility and high Caco-2 efflux, 11 we substituted the benzimidazolone NH group by forming the N-Me variant 17; 31 however, this resulted in a 20-fold less efficacious compound ( Figure 2a). Likewise, the oxygencontaining analogue 18 displayed decreased efficacy when compared to 1 (Figure 2a). We then replaced the benzimidazolone moiety, 10,20 as this can inflict high efflux and low solubility by a series of East-positioned amides. From these amides, 20 and 21 turned out to be the most potent resulting in picomolar cellular inhibitory IC 50 efficacies ( Figure  2a). Solubility and Caco-2 cell permeation were completely restored in 21 (Figure 2b). The amide having the N-Me group (44) was inactive, whereas activity was restored in the cyclic version (51) (Supplementary Table 1a). Nonaromatic amides (52, 53, and 54) resulted in inactive compounds (Supplementary Table 1a).
To interrogate the West-side of the pharmacophore, pyridine and pyrimidine analogues were prepared. These compounds inhibited the cellular WNT/β-catenin signaling pathway reporter assay to a similar extent as lead 1, except the 2-pyrimidyl substituted compound (56). The ethoxypyridyl derivative 22 was consequently selected as a starting point for further hybrid synthesis (Figure 2a and Supplementary Table   Table 1b). Introduction of aliphatic rings, such as cyclopentane in 63 and cyclopropane in 64, indicated that aromatic ring systems are required in this position for maintaining potency (Supplementary Table 1b).
Next, in a series of synthesized South-aryl products, the 2trifluoromethyl (66) showed comparable activities to the previous lead 1 (Supplementary Table 1b). The thiophenyl moeity as a bioisosteric replacement for the aryl group was less tolerated, while cycloalkyl replacements resulted in activities in the micromolar range (Supplementary  (Figure 2b). Despite the more rigid geometry of the bicyclo[1.1.1]pentane in comparison to trans-cyclobutyl of 1, the cocrystal structure with TNKS2 showed a very similar binding mode at the NAD + binding cleft ( Figure 3). 105 occupied the adenosine subpocket and formed the typical hydrogen bonds to the backbone amides of Tyr1060 and Asp1045 (Figure 3a and Supplementary Figure  1a). A water molecule forms bridging interactions between the pyridine nitrogen and Gly1058 and Tyr1050, and the same applies to all the cocrystal structures described (Figure 3a).
Rotational isomerism (atropisomerism) is a known phenomenon for substituted triazoles and can potentially lead to complexity and challenges for the drug discovery and development processes as atropisomers might have differing biological activities toward a target, different off-target profiles, and different pharmacokinetic properties. 32,33 Since 1 does not contain asymmetric centers, atropisomers are mirror images (enantiomers). Hence, on an achiral HPLC column, as well as in NMR, such atropisomers are indiscernible. In contrast, on a chiral SFC column, lead 1 showed two signals indicating rotational isomerism ( Figure 4). When separated, these isomers did not interconvert at 20°C for 72 h but showed a minor interconversion at 70°C during 72 h ( Figure 4). Interestingly, both isomers, 1-AT-1 and 1-AT-2, differed in potency and efficacy with a factor of 30 to almost 60, respectively ( Figure 4). To investigate whether atropisomerism was induced by the South 2-chlorophenyl substituent, we analyzed all synthesized compounds with chiral SFC. All compounds with a South 2-chlorophenyl group showed atropisomers, while compounds without this group, including the symmetric 2,6-dichlorophenyl 67, did not (Supplementary  Table 1c). In addition, 66, containing a bulky 2-trifluoromethyl group, also showed two signals on a chiral SFC column. No rotamers were observed for the 2-fluorophenyl South group at room temperature. Hence, this group was considered a viable substitution to avoid atropisomerism. As a consequence, in the following optimization campaign, the South 2-chlorophenyl Next, all possible hybrid combinations were synthesized employing the East moieties of compounds 20 and 21, a South 2-fluorophenyl group, and three different West moieties revisiting the 2-pyridyl-4-methylsulfonyl moiety as well (Table 1). 9 From these six molecules, compounds with the 1,5-naphthyridine moiety (10, 11, and 12) showed approximately 30-fold improved cellular inhibitory efficacy compared to their counterparts with the 2-pyridyl moiety (13,14, and 15, respectively, Table 1). The binding mode of compound 13 in cocrystal structures was similar, and the hydrogen bonds to the backbone amides seen in the 1 cocrystal structures were retained ( Figure 3 and Supplementary Figure 1b). The large East moieties 1,5-and 1,6-naphthyridine (e.g., see 87, 88, 106, and 107 Supplementary Figure 4a) appeared to form a more Table 1. Hybrids Derived from 19 and 20, 22, and three West Moieties a a # indicates that these compounds were evaluated in a mouse pharmacokinetics analysis. Table 2. 2D East (Green) and West Library (Red) with TNKS2 IC 50 (nM), HEK293 IC 50 (nM), and Solubility (μM) Values Depicted a efficient π−π-stacking interaction with His1048 and a hydrophobic interaction with Phe1035, causing the side chain to gradually rotate according to the form of the East moiety ( Figure 3c Table 2). However, compounds with this naphthyridine moiety were less stable in microsomes relative to the corresponding East-pyridines, and consequently, 13 was shortlisted for mouse pharmacokinetics studies. In addition, the quinoline-containing 16 displayed similar inhibition in the cellular WNT reporter assay compared to 12 and also possessed improved calculated physical−chemical properties (ChemAxon, cLogP = 3.0/3.7, tPSA = 128/141 for 16 and 12,    respectively) (Figure 2a and Table 1). On the basis of these results, 16 was short-listed for mouse pharmacokinetics analysis. Further optimization focused on pyridine-type Westside variations and discarded the West 2-thiazole (11) because of adverse solubility and low efficacy (14) ( Table 1). In an East and West 2D library, fluorinated analogues of the 2pyridyl and the 1,5-naphthyridine were introduced (Table 2  and Supplementary Table 2). From these compounds, 27 was short-listed for mouse pharmacokinetics analysis. Compared to the initial benchmark lead compound 1, the peroral mouse pharmacokinetics data of the selected and shortlisted compounds showed significantly improved profiles exhibiting lower clearance and volume of distribution and 15−35 times higher exposure (Figure 2c). In due course of the study, 13 had been further characterized including selectivity toward other members of the PARP family, structural analysis of its binding mode, kinetic solubility, Caco-2 permeability and efflux, CYP3A4 inhibition, mouse plasma stability, mouse plasma protein binding, hERG inhibition, Ames test, and offtarget safety panel, exhibiting overall favorable parameters ( Figure 3, Table 3, Supplementary Table 3 and Supplementary Figure 2).
Tankyrase inhibition can context-dependently antagonize proliferation and viability in cancer cell lines in vitro and in vivo, including in the colorectal adenocarcinoma cell line COLO 320DM harboring WNT/β-catenin signaling-inducing APC mutations. 5,22 Hence, cultured COLO 320DM cells were treated with various doses of 13 to evaluate the efficacy in reducing canonical WNT/β-catenin signaling and the potential as an antiproliferative agent. As expected for a potent tankyrase inhibitor, treatment with 13 reduced TNKS1/2 protein levels, stabilized AXIN1 and AXIN2 proteins, and reduced the level of transcriptionally active β-catenin (nonphoshorylated) in both the cytoplasmic and nuclear fraction (Figure 5a and Supplementary Figure 3a,b). Administration of 13 also decreased transcription of the WNT/β-catenin signaling target genes AXIN2, DKK1, NKD1, and APCDD1 (Figure 5b). Moreover, 13 exposure decreased proliferation and viability in COLO 320DM cells (GI 50 = 650 nM and GI 25 = 94 nM), while control APC wild-type RKO colorectal cancer cells were only modestly affected by the treatment at a 10 μM compound concentration (Figure 5c). In conclusion, these results authenticate that 13 can both potently and specifically inhibit WNT/β-catenin signaling activity and act as an antiproliferative agent in COLO 320DM cells.

■ CONCLUSION
In summary, through a systematic building-block-based and crystallography-guided structure−activity-relationship analysis, we identified novel 1,2,4-triazole-based optimized lead tankyrase inhibitors with low nanomolar and picomolar IC 50 activities in a WNT/β-catenin signaling cellular reporter assay. The adverse chemical properties of the preceding lead compound 1, 11 displaying atropisomerism and solubility liabilities, were excluded in the here identified advanced lead compound 13. Compound 13 shows high selectivity toward TNKS1/2, an overall favorable ADME, including highly improved Caco-2 permeability and microsomal stability, a clean off-target safety profile, a >15-fold increased exposure in a mouse pharmacokinetics analysis, and a robust inhibition of WNT/β-catenin signaling and proliferation in the colon cancer cell line COLO 320DM. Our work provides a considerably optimized compound for targeting TNKS1/2 and WNT/βcatenin signaling in cancer and other disease models.
■ EXPERIMENTAL SECTION General Methods. All starting materials and dry solvents were commercially obtained. The synthesis of hydrazide 2 was scaled up employing a similar procedure reported in our earlier publication. 11 Reactions were performed under an inert atmosphere of nitrogen when necessary. Microwave reactions were carried out in sealed vials. Column chromatography was carried out on silica gel cartridges (40 μm irregular), and TLC analysis was performed on silica gel 60 F 254 plates.
NMR. NMR spectra were recorded in chloroform-d, unless otherwise stated, on a 400 MHz spectrometer with tetramethylsilane as internal standards. Coupling constants are given in Hz. Peaks are reported as singlet (s), doublet (d), triplet (t), quartet (q), quintet (p), sextet (h), septet (hept), multiplet (m), or a combination thereof. br stands for broad. All test compounds were found to be >95% pure by LCMS and H NMR analysis. Intermediates in the synthesis were >95% pure unless stated otherwise. (10). The title compound was prepared according to general procedure F as a white solid (21.4 mg, 38% yield). LC/MS (ESI) m/z for C 28  3-(trans-3-(4-(2-Fluorophenyl)-5-(thiazol-2-yl)-4H-1,2,4-triazol-3-yl)cyclobutyl)-1,5-naphthyridine-4-carboxamide (11). The title compound was prepared according to general procedure F as a white solid (6.6 mg, 23% yield). LC/MS (ESI) m/z for C 24  3-(trans-3-(4-(2-Chlorophenyl)-5-(pyrimidin-4-yl)-4H-1,2,4triazol-3-yl)cyclobutyl)-2-oxo-2,3-dihydrobenzo[d]oxazole-6carbonitrile (18). Crude hydroxybenzonitrile (8) (10 mg, 0.014 mmol, ∼60% pure) was dissolved in DCM (dried, 3.0 mL) under a nitrogen atmosphere. Acetonitrile (anhydrous, 1.0 mL) was also added followed by CDI (11 mg, 0.070 mmol), and the mixture was stirred for 60 h at ambient temperature. The solvent was evaporated, and the residue was purified by preparative SFC. After freeze-drying the purified fractions from acetonitrile/water, 2.8 mg (42% yield) of a white powder of the target compound (18)  were added and the mixture was stirred for 3 h at ambient temperature. Extra methanol was added (extra dry, 1.0 mL) for better dissolution. The mixture was then heated at 50°C overnight for 40 h while two extra portions of sodium triacetoxyborohydride were added (each of 85 mg, 0.40 mmol). The mixture was evaporated to dryness, and the residue was quenched with water and a few mL of 1 N aqueous HCl. DCM was added, the aqueous phase was basified with aqueous sodium bicarbonate, and the crude product was extracted with DCM (three times). The extracts were dried over sodium sulfate, filtered, and evaporated to dryness. The crude material was purified on a 12 g silica gel cartridge eluted with a gradient of methanol (0% to 5%) in DCM. The product fraction was lyophilized from acetonitrile/ water providing a white powder ( Separation of Atropisomers of 1 and Investigation of Their Interconversion. An amount of 40 mg of racemic 1 11 (40 mg, 0.085 mmol) was separated into pure atropisomers by preparative chiral SFC (see Supporting Information) providing, respectively, 16.1 and 11.8 mg of off-white solids. Two sets of two samples dissolved in an acetonitrile/methanol mixture were prepared for each atropisomers. One set was heated in a reaction block at 70°C, while the other was kept at ambient temperature. Samples were analyzed after 24, 48, and 72 h to check for the interconversion of atropisomers. After 24 h no interconversion was observed at ambient temperature, while at 70°C 2.5% of the opposite isomers were detected in both samples. After 48 h of heating about 5−6% interconversion was observed at 70°C. After 72 h, the isomers showed no interconversion at ambient temperature, whereas at 70°C about 8−9% of the other atropisomer was observed by chiral SFC.
Microsomal Stability Assay. Test compounds in DMSO (10 mM) were further diluted to 100 μM in acetonitrile. Human or mouse liver microsomes (BioIVT) from selected species are incubated in duplicate with the test compound at a final concentration of 1 μM in 0.1 M potassium phosphate buffer (pH 7.4) containing 3.3 mM MgCl 2 , 0.5 mg/mL microsomal protein, in the presence or absence of NADPH (1 mM). Incubations were performed at 37°C in a total volume of 500 μL. Control incubations with reference substances were included for each experiment. At t = 0, 5, 15, 30, 45 min, an amount of 50 μL of the incubation mixture was transferred to a quench plate containing acetonitrile and internal standard (200 nM labetalol) cooled at 4°C. After the last time point, the quench plates are mixed thoroughly and centrifuged for 15 min at 3700 rpm and 10°C (Eppendorf 5804R). The supernatant was transferred to a 96-well plate and analyzed by LCMS (Vanquish Horizon UHPLC equipped with a diode array detector coupled to a Q Exactive focus hybrid quadrupole-Orbitrap mass spectrometer). The metabolic stability is evaluated by plotting the natural logarithm of the percentage test compound remaining versus time and performing linear regression.
Mouse Pharmacokinetical Analysis. The pharmacokinetical analyses in mice were performed according to the standard protocols of Medicilon and, as previously described, 9 using 3 animals per treatment group using 5% DMSO, 50% PEG400 (both Sigma-Aldrich) and 45% saline as vehicle.
Crystallography. Compounds 105, 13, 10, 106, 107, 87, 88 (Supplementary Table 4) were cocrystallized with the catalytic domain of human TNKS2 (residues 952−1161) in the presence of chymotrypsin (1:100) based on crystallization efforts previously described. 11,20 Protein (0.2 mM, 5.6 mg/mL) was mixed with 0.4 mM compound from a 10 mM DMSO stock solution. The crystallization was set up at 22°C using sitting-drop vapor diffusion method by mixing 200 nL of protein with 100 nL of precipitant solution (0.1 M Bicine, pH 8.5−9.0, 7.5−25% PEG6000). Rod-shaped crystals appeared within 24 h and were cryoprotected using the precipitant solution containing 25% PEG6000 and 20% glycerol. Data were collected at ESRF Grenoble on beamlines ID30B, ID23-1, and ID30A-1 or at Diamond Light Source on beamline I04. Diffraction data were processed using the XDS package. 38 All structures were solved using molecular replacement with PHASER 39 using the structure of TNKS2 (PDB code 5NOB) as a starting model. Coot 40 and Refmac5 41 were used for model building and refinement, respectively. The images of the structures were prepared using PyMOL (The PyMOL Molecular Graphics System, version 1.8.4.0, Schrodinger, LLC.).
Proliferation Assay. 1000 COLO 320DM cells per well were seeded in 96-well plates. The day after, cell culture medium was changed to contain various doses of 13 (n = 4) or vehicle (DMSO, Sigma-Aldrich) and the plates were incubated at 37°C for 5 days. The cells were incubated for 1 h at 37°C with CellTiter 96 AQueous nonradioactive cell proliferation assay (MTS, Promega) according to the supplier's recommendations. Abs 490 was measured (Wallac 1420 Victor2 microplate reader, PerkinElmer) and compared to Abs 490 (t 0 ) using the following formula to determine single well values relative to the DMSO vehicle control: (sample A 490 − average A 490 t 0 ) × 100/ (average A 490 [for 0.01% DMSO controls] − average A 490 t 0 ).
General and specific synthetic procedures and spectra (PDF) Additional figures and tables for inhibition data and crystallography (PDF) Molecular formula strings and some data (CSV)

Accession Codes
Atomic coordinates and structure factors have been deposited to Protein Data Bank with accession codes 6TG4, 6TKM, 6TKN, 6TKP, 6TKQ, 6TKR, and 6TKS.