Proteomic profiling of small-molecule inhibitors reveals dispensability of MTH1 for cancer cell survival

Since recent publications suggested that the survival of cancer cells depends on MTH1 to avoid incorporation of oxidized nucleotides into the cellular DNA, MTH1 has attracted attention as a potential cancer therapeutic target. In this study, we identified new purine-based MTH1 inhibitors by chemical array screening. However, although the MTH1 inhibitors identified in this study targeted cellular MTH1, they exhibited only weak cytotoxicity against cancer cells compared to recently reported first-in-class inhibitors. We performed proteomic profiling to investigate the modes of action by which chemically distinct MTH1 inhibitors induce cancer cell death, and found mechanistic differences among the first-in-class MTH1 inhibitors. In particular, we identified tubulin as the primary target of TH287 and TH588 responsible for the antitumor effects despite the nanomolar MTH1-inhibitory activity in vitro. Furthermore, overexpression of MTH1 did not rescue cells from MTH1 inhibitor–induced cell death, and siRNA-mediated knockdown of MTH1 did not suppress cancer cell growth. Taken together, we conclude that the cytotoxicity of MTH1 inhibitors is attributable to off-target effects and that MTH1 is not essential for cancer cell survival.


Identification of new MTH1 inhibitors.
To identify new MTH1 inhibitors, we used chemical arrays to screen 29,809 compounds from a chemical library of the RIKEN Natural Products Depository (NPDepo) 32,33 . Twenty MTH1 binders were detected (Fig. 1a). Next, we ran an in vitro enzyme assay to determine the inhibitory effects of these screening hits against MTH1, with 8-oxo-dGTP and 2-OH-dATP as substrates. Subsequently, we found that NPD15095, a purine derivative, inhibited MTH1 catalytic activity in a dose-dependent manner (Fig. 1b,c). The IC 50 values of NPD15095 for MTH1 inhibition were 3.3 μ M and 6.7 μ M, for 8-oxo-dGTP and 2-OH-dATP, respectively.
To find more potent MTH1 inhibitors, we explored 131 structurally related compounds with a purine moiety. Two potent MTH1 inhibitors (NPD7155 and NPD9948) and a less active analog (NPD8880) were discovered (Fig. 1b,c). The potencies of NPD7155 and NPD9948 were comparable to those of (S)-crizotinib and SCH51344 (Supplementary Table S1). Therefore, we focused on NPD7155 and NPD9948 for further validation studies.
As NPD7155 and NPD9948 are mimics of the purine moiety of MTH1 substrates such as 8-oxo-dGTP and 2-OH-dATP, we examined their effects on other nucleoside triphosphate pyrophosphatases, such as inosine triphosphate pyrophosphatase (ITPA) and deoxycytidine triphosphate pyrophosphatase 1 (DCTPP1), whose substrates are purine nucleotides and pyrimidine nucleotides, respectively. In our results, NPD7155 and NPD9948 did not affect the enzymatic activities of ITPA and DCTPP1, even at 100 μ M. In contrast, (S)-crizotinib and SCH51344 slightly inhibited DCTPP1 activity at the same concentration, suggesting NPD7155 and NPD9948 have some more target selectivity than the inhibitors (Fig. 2c).
Purine-based MTH1 inhibitors exhibit less potent cytotoxicity. We next examined the effects of NPD7155 and NPD9948 on MTH1 in cancer cells. To determine whether NPD7155 and NPD9948 target MTH1 in intact cells, we performed a target engagement assay based on ligand-increased thermal stabilization 34 . Treatment of human cervical cancer HeLa cells with NPD7155 and NPD9948 considerably stabilized cellular MTH1, but not many other proteins (Fig. 3a), indicating that NPD7155 and NPD9948 specifically bind to MTH1 in HeLa cells.
Since recent publications suggest that MTH1 is essential for cancer cell survival 8,9 , we analyzed the cytotoxic effects of our MTH1 inhibitors against cancer cells. Contrary to our expectations, NPD7155 and NPD9948 exhibited only weak cytotoxicity against HeLa cells (IC 50 = 65 μ M and 35 μ M, respectively; Fig. 3b) despite potent MTH1-inhibitory activity with sub-micromolar ranges of IC 50 values in vitro (Supplementary  Table S1). In addition, our purine-based MTH1 inhibitors exhibited only weak cytotoxicity in other cancer cell lines ( Supplementary Fig. S2). In contrast, (S)-crizotinib and SCH51344 had more potent cytotoxic effects (IC 50 = 13 μ M and 5.5 μ M, respectively; Fig. 3b), even though these compounds displayed the same degree of MTH1-inhibitory activities as NPD7155 and NPD9948 in vitro (Supplementary Table S1). It has been reported that MTH1 plays a crucial role in preventing incorporation of oxidized nucleotides into nuclear and mitochondrial DNA, allaying subsequent DNA damage 35 . Therefore, we examined the effects of our MTH1 inhibitors on 8-oxo-2ʹ -deoxyguanosine (8-oxo-dG) levels in DNA. However, NPD7155 and NPD9948 did not induce the significant accumulation of 8-oxo-dG in HeLa cells, even at cytotoxic concentrations, when compared to that of TH287 ( Fig. 3b and Supplementary Fig. S3a). Next, we investigated whether our MTH1 inhibitors can induce DNA damage, and found that NPD7155 and NPD9948 increased nuclear 53BP1 foci formation, a specific marker for DNA damage, only at the higher concentrations ( Supplementary Fig. S3b). Cell cycle analysis showed that NPD7155 and NPD9948 increased the sub-G1 cell population, an indication of dead cells, at the same higher concentrations (Supplementary Fig. S3c). However, NPD8880 also induced DNA damage ( Supplementary Fig. S3b) and cell growth inhibition (IC 50 = 650 μ M; Fig. 3b) at similar concentrations in spite of  the little MTH1-inhibitory activity (Fig. 1c). These data raised the question whether MTH1 inhibition is responsible for the cytotoxic effects of small-molecule MTH1 inhibitors.

Mechanistic differences among MTH1 inhibitors.
In order to validate whether chemically distinct MTH1 inhibitors have the same modes of action for cytotoxicity, we performed a profiling analysis using ChemProteoBase, a comprehensive database of cellular proteomic variations induced by treatment with well-characterized bioactive compounds 31 . The hierarchical cluster analysis of five MTH1 inhibitors and 41 standard compounds revealed that the proteomic variation induced by NPD7155 is similar to that induced by NPD9948 (Fig. 4). Furthermore, these two purine-based MTH1 inhibitors shared similarity with camptothecin (a topoisomerase I inhibitor) and SCH51344. Both camptothecin and SCH51344, as well as NPD7155 and NPD9948, are known to induce DNA damage 9,36 . On the other hand, the MTH1 inhibitor (S)-crizotinib did not belong to any clusters of the standard compounds in our proteomic profiling system (Fig. 4). More surprisingly, proteomic profiling revealed that TH287, the most potent MTH1 inhibitor we tested (Supplementary Table S1), shared similarity with tubulin-targeting agents such as nocodazole, vinblastine, and paclitaxel, rather than other MTH1 inhibitors (Fig. 4).
TH287 and TH588 are tubulin polymerization inhibitors. Indeed, TH287 and its metabolically more stable analog TH588 8 inhibited tubulin polymerization in vitro in a dose-dependent manner, whereas NPD7155 and NPD9948 did not affect tubulin polymerization, even at 300 μ M (Fig. 5a,b and Supplementary  Fig. S4). Furthermore, TH287 and TH588 disrupted intracellular microtubule networks in HeLa cells at cytotoxic concentrations, while also inducing cell shrinkage ( Fig. 3b and Supplementary Fig. S5). Since tubulin-targeting agents are known to induce phosphorylation of Bcl-2 37,38 , we examined the effects of MTH1 inhibitors on Bcl-2. Our results indicated that TH287 and TH588, but not the other MTH1 inhibitors, induced Bcl-2 phosphorylation at the effective concentrations, as detected by the mobility shift on the upper side of unphosphorylated Bcl-2 ( Fig. 5c). Likewise, cell cycle analysis revealed that TH287 and TH588, but not the other MTH1 inhibitors, increased the population of HeLa cells in G2/M phase in the same manner as vinblastine (Fig. 5d, Supplementary  Figs S3c and S6). These data strongly suggest that tubulin is the primary target of TH287 and TH588, and that it is responsible for their cytotoxic effects. MTH1 inhibition does not affect cancer cell growth. As our proteomic profiling showed mechanistic differences among the MTH1 inhibitors, we investigated whether overexpression of human MTH1 rescues cells from MTH1 inhibitor-induced cell death. Contrary to our expectations, MTH1 overexpression did not alter the sensitivity of cells towards any of the MTH1 inhibitors we tested (Fig. 6a,b). Furthermore, siRNA-induced   knockdown of MTH1 did not suppress cell growth and cell cycle progression in HeLa cells (Fig. 6c-f). These data suggest that small-molecule MTH1 inhibitors exert cytotoxic effects primarily by targeting off-target molecules such as tubulin, rather than MTH1 inhibition.

Discussion
In the present study, we found new purine-based MTH1 inhibitors, NPD7155 and NPD9948, that exhibited potent MTH1-inhibitory activity in vitro comparable to that of (S)-crizotinib and SCH51344 (Supplementary  Table S1). However, NPD7155 and NPD9948, showed only weak cytotoxicity compared to that of (S)-crizotinib and SCH51344 (Fig. 3b and Supplementary Fig. S2), despite their ability to inhibit the MTH1 activity in intact cells (Fig. 3a). This fact raised the questions whether MTH1 inhibition is responsible for the cytotoxic effects of known small-molecule MTH1 inhibitors and whether MTH1 is suitable as a target molecule for cancer therapy. Indeed, proteomic profiling revealed that TH287, (S)-crizotinib, and SCH51344 are mechanistically different (Fig. 4), although these known MTH1 inhibitors showed potent MTH1-inhibitory activity in vitro (Supplementary Table S1). In addition, considering that overexpression of human MTH1 did not rescue cells from the cell death induced by small-molecule MTH1 inhibitors (Fig. 6a,b) and knockdown of MTH1 did not suppress cancer cell growth (Fig. 6c-f) 39 . Together, our data and these reports suggest that tumor growth suppression by genetic or pharmacological inhibition of MTH1 in previous studies was likely due to off-target effects.
The original work showed that TH287 and TH588 did not inhibit other members of the Nudix protein family nor any of the nucleoside triphosphate pyrophosphatases tested 8 . With few exceptions, TH588 had little effect against 87 enzymes, G-protein-coupled receptors (GPCRs), kinases, ion channels, and transporters, even at 10 μ M 8 . However, the effects of TH287 and TH588 on tubulin have not been investigated. We know that in general, the concentrations of compounds required for inhibition of tubulin polymerization in vitro are higher than those required for disruption of microtubule networks in cells 26,30 . In our study, TH287 and TH588 disrupted intracellular microtubule networks and induced Bcl-2 phosphorylation and cell cycle arrest in the M phase at cytotoxic concentrations (Figs 3b and 5c,d and Supplementary Fig. S5), suggesting that tubulin is the primary target responsible for the cytotoxicity of TH287 and TH588 in HeLa cells. In a previous study, TH588 suppressed tumor growth of patient-derived malignant melanoma cells in a mouse xenograft model 8 , but ectopic expression of MutT (a bacterial homolog of MTH1) only partially rescued the cells from TH588-induced cell death 8 . Combined with this report, our findings strongly suggest that TH588 and TH287 exert antitumor effects primarily by inhibiting tubulin.
TH287 increased 8-oxo-dG levels, whereas our MTH1 inhibitors had little effect ( Supplementary Fig. S3a). Accumulation of 8-oxo-dG is an indication of oxidative stress, and tubulin-targeting agents such as vinca alkaloids and taxanes are known to induce ROS during the process of cell death 41 . On the other hand, there are complementary enzymes such as 8-oxoguanine DNA glycosylase (OGG1) that minimize the accumulation of 8-oxo-dG in DNA 42 . Although the cause for the different effects of TH287 and our MTH1 inhibitors is unclear, we speculate that these 8-oxo-dG-increasing and -decreasing factors are the determinants of accumulated 8-oxo-dG level.
The mode of action responsible for the cytotoxic effects of (S)-crizotinib, SCH51344, and our purine compounds could not be explained through proteomic profiling (Fig. 4). (R)-crizotinib is clinically used as a c-MET/ anaplastic lymphoma kinase dual inhibitor 43 , and its (S)-enantiomer has also been reported to inhibit c-MET 44 .
In conclusion, tubulin was identified as another target of TH287 and TH588 to cause the cytotoxicity, and MTH1 inhibition alone may not be sufficient for antitumor therapy.

Methods
Materials. All compounds used for screening were supplied from a chemical library of the RIKEN NPDepo 32,33 . (S)-Crizotinib, SCH51344, TH287 hydrochloride, and TH588 hydrochloride were purchased from ApexBio Technology, Santa Cruz Biotechnology, Axon Medchem, and Sigma-Aldrich, respectively. The purity of the compounds was checked using LC-MS.
Chemical array screening. To prepare the chemical arrays, solutions of the 29,809 compounds (2.5 mg/mL in dimethyl sulfoxide; DMSO) from the RIKEN Natural Products Depository (NPDepo) were arrayed onto 11 separate photoaffinity linker-coated glass slides (using a chemical array-manufacturing apparatus developed by RIKEN) and immobilized via a photo-cross-linking method 14 . The chemical array screening was performed according to our previous reports 17,18 . Briefly, we prepared a lysate of HEK293T cells that were transiently overexpressing DsRed-fused MTH1 or DsRed protein in phosphate-buffered saline (PBS), and incubated the chemical arrays with the cell lysate for 1 h at 4 °C. After washing with PBS, the chemical arrays were scanned using a GenePix microarray scanner (Molecular Devices) with an excitation wavelength of 532 nm. The fluorescence signals were quantified using the GenePix Pro 5.0 software (Molecular Devices) with local background subtraction. We used data from slides treated with a lysate of DsRed-expressing cells as a reference. The images from two slides treated with a cell lysate of MTH1-DsRed-overexpressing cells and that of DsRed-overexpressing cells were colored red and green, respectively, using Photoshop 5.5 software (Adobe Photoshop), and merged into a composite image.
Enzyme assay for MTH1, ITPA, and DCTPP1. An MTH1 enzyme assay was performed as described previously with minor modifications 9 . Briefly, test compounds were dissolved in an assay buffer [100 mM Tris-acetate, 40 mM NaCl, 10 mM Mg(OAc) 2 , 0.005% Tween-20, and 2 mM dithiothreitol (DTT), pH 7.5] containing His-tagged human recombinant MTH1 protein (2 nM; ProSpec). After addition of the mixed reagent from the PPiLight Inorganic Pyrophosphate Assay kit (Lonza Rockland), either 8-oxo-dGTP (13.2 μ M; TriLink BioTechnologies) or 2-OH-dATP (8.3 μ M; Jena Bioscience) was added as a substrate to initiate the enzymatic reaction. Relative MTH1 activity was measured by monitoring the pyrophosphate (PPi) generated through MTH1-catalyzed nucleotide triphosphate hydrolysis. The luciferase-mediated luminescence signal was recorded for 30 min using a luminometer (Molecular Devices). Then, SigmaPlot (Hulinks) was used for kinetic analysis to determine the inhibition patterns of NPD7155 and NPD9948 and calculate their K i values.
To exclude false positive compounds that inhibit PPi detection, the effects of test compounds on 10 μ M Na 4 P 2 O 7 -induced luciferase activity were examined.
Target engagement assay. To analyze the interaction between small-molecule ligands and their target proteins in intact cells, a cellular thermal shift assay (CETSA) was performed as described previously with minor modifications 9,34 . Briefly, HeLa cells cultured in a 100-mm dish to 80% confluency, were treated with NPD7155 (300 μ M), NPD9948 (300 μ M), or (S)-crizotinib (30 μ M) for 1 h. After treatment, cells were collected with trypsin and resuspended in Tris-buffered saline (TBS). The cell suspension was aliquoted into four polymerase chain reaction (PCR) tubes and heated at 45, 50, 55, or 60 °C for 5 min. Subsequently, cells were lysed by three repeated freeze-thaw cycles using liquid nitrogen. Precipitated proteins were separated from the soluble fraction by centrifugation at 17,000 × g for 20 min. Soluble proteins collected in the supernatant were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore) for western blot analysis. The membrane was incubated with a primary antibody against MTH1 (1:500, sc-67291; Santa Cruz Biotechnology) and a horseradish peroxidase-labeled secondary antibody (Vector Laboratories) in sequence. They were then visualized with Fusion Solo S (Vilber Lourmat) using a Super Signal West Pico Chemiluminescence Substrate (Pierce). After detection of MTH1 by western blot, proteins on the membrane were visualized by staining with Coomassie brilliant blue (CBB).
Profiling by ChemProteoBase. ChemProteoBase profiling analysis was performed as described previously 31 . Briefly, HeLa cells were treated with compounds for 18 h. Proteome analysis of the cell lysate was performed using a 2D difference gel electrophoresis (2-D DIGE) system (GE Healthcare). Images of the gels were analyzed with Progenesis SameSpots (Nonlinear Dynamics). Of the more than 1000 spots detected in each 2D gel, 296 variational spots found to be common between the gels of the reference and compound-treated cells were selected as described 31 . Next, the volume of each spot was normalized using the average of the corresponding control values from DMSO-treated HeLa cells, and the log-fold ratio of the normalized volumes and the average value in triplicate was calculated. The ChemProteoBase dataset was constructed from the data acquired from HeLa cells treated with RIKEN NPDepo library standard compounds. Data for the 296 spots, which were acquired from the test compounds, and the ChemProteoBase dataset were combined, and hierarchical clustering analysis was performed with Cluster 3.0 (clustering method; centroid linkage with the means of uncentered correlation). The predictive dendrogram was visualized with Java Tree View 1.1.3.

Tubulin polymerization assay.
In vitro tubulin polymerization assay was performed using a tubulin polymerization assay kit (Cytoskeleton) according to the manufacturer's instructions. Briefly, lyophilized porcine tubulin was solubilized to a final concentration of 2 mg/mL in reaction buffer containing 80 mM piperazine-N,N′bis(2-ethanesulfonic acid) (PIPES; pH 6.9), 2 mM MgCl 2 , 0.5 mM ethylene glycol tetraacetic acid (EGTA), 1 mM guanosine-5′ -triphosphate (GTP), 10 μ M fluorescent reporter, and 20% glycerol, and kept at 4 °C. Compounds (100 × DMSO stock solutions) were added to prewarmed half-area 96-well black plates. Cold tubulin solution was added to the wells. The plate contents were mixed by shaking, and fluorescence (Ex 350 nm, Em 435 nm) was read every minute for 1 h using a fluorescence microplate reader (Molecular Devices).

Overexpression of MTH1.
For construction of the pcDNA3.1/MTH1 plasmid, the MTH1 cDNA was cloned from HeLa cells by PCR using the forward primer 5′ -AGTGTGGTGGAATTCATGAGTGGAATTAGCCCTCA-3′ and the reverse primer 5′ -ATATCTGCAGAATTCGGACCGTGTCCACCTCGCGG-3′ , and inserted into pcDNA3.1/myc-His vector (Invitrogen) with the EcoRI restriction site. The destination plasmid was verified by sequencing analysis. For preparation of MTH1-overexpressing cells, HeLa cells were seeded in 6-well plate, cultured overnight, and transfected with the pcDNA3.1/MTH1 plasmid or the control vector using Effectene Transfection Reagent (Qiagen) according to the manufacturer's instruction. After 24 h of the transfection, the cells were reseeded in 96-well plate, incubated for 12 h, and then treated with MTH1 inhibitors for 84 h. The cell viability was examined by WST-8 assay (see Supplementary Methods "Cell death assay"). After 24 h of the transfection, the expression level of MTH1 protein was analyzed by western blot (see "Target engagement assay").

Knockdown of MTH1.
Control siRNA (sc-37007, Santa Cruz Biotechnology) and human MTH1 siRNA (sc-62647, Santa Cruz Biotechnology) were obtained. HeLa cells were transfected with 20 nM siRNA using Lipofectamine RNAiMAX according to the manufacturer's instructions (Invitrogen). MTH1 knockdown after 72-h siRNA treatment was analyzed by western blot. After transfection, cells were counted using Automated Cell Counter (Bio-Rad).