Understanding inhibitor resistance in Mps1 kinase through novel biophysical assays and structures

Monopolar spindle 1 (Mps1/TTK) is a protein kinase essential in mitotic checkpoint signaling, preventing anaphase until all chromosomes are properly attached to spindle microtubules. Mps1 has emerged as a potential target for cancer therapy, and a variety of compounds have been developed to inhibit its kinase activity. Mutations in the catalytic domain of Mps1 that give rise to inhibitor resistance, but retain catalytic activity and do not display cross-resistance to other Mps1 inhibitors, have been described. Here we characterize the interactions of two such mutants, Mps1 C604Y and C604W, which raise resistance to two closely related compounds, NMS-P715 and its derivative Cpd-5, but not to the well characterized Mps1 inhibitor, reversine. We show that estimates of the IC50 (employing a novel specific and efficient assay that utilizes a fluorescently labeled substrate) and the binding affinity (KD) indicate that, in both mutants, Cpd-5 should be better tolerated than the closely related NMS-P715. To gain further insight, we determined the crystal structure of the Mps1 kinase mutants bound to Cpd-5 and NMS-P715 and compared the binding modes of Cpd-5, NMS-P715, and reversine. The difference in steric hindrance between Tyr/Trp604 and the trifluoromethoxy moiety of NMS-P715, the methoxy moiety of Cpd-5, and complete absence of such a group in reversine, account for differences we observe in vitro. Our analysis enforces the notion that inhibitors targeting Mps1 drug-resistant mutations can emerge as a feasible intervention strategy based on existing scaffolds, if the clinical need arises.

One of the common features of solid human tumors is the presence of an abnormal number of chromosomes, a state often referred as "aneuploidy" (1). Previous studies indicate that a mechanism that sustains aneuploidy in tumor cells is the overexpression of high levels of monopolar spindle 1 (Mps1) 2 kinase (2). Mps1, also known as threonine and tyrosine kinase (TTK), is a dual specificity protein kinase, essential for safeguarding proper chromosome alignment and segregation during mitosis (2). Due to its importance for the viability of tumor cells, Mps1 kinase has become a potential target for cancer treatment (2). Over the past years, several dozens of small compounds have been developed to inhibit the Mps1 kinase activity (2)(3)(4)(5)(6)(7). Recently, NMS-P715 has been described to suppress the growth of medulloblastoma cells, a common malignant brain tumor in children (8). The anti-proliferative activity of NMS-P715 has been also shown in breast, renal, and colon cancer cell lines (9). A derivative of NMS-P715, Compound 5 (Cpd-5) (10), has been reported to display higher potency toward Mps1 than NMS-P715 (11). Although these Mps1 inhibitors have promising results in pre-clinical studies (4,8,9), they also have innate problems as kinase inhibitors (11,12). After the initial drug response, tumor cells frequently acquire resistance and become insensitive to treatment (11,13). The development of drug resistance in cancer cells is often the consequence of mutations of the targeted kinases (14). These mutations are typically found in the ATP-binding pocket, which renders the binding of inhibitors suboptimal, whereas retaining kinase activity (15). A mutation at Cys 604 of the Mps1 kinase has been isolated by raising resistance against a number of inhibitors including NMS-P715 (12) as well as Cpd-5 (11), in cell studies. It is positioned in the "hinge loop" region of the kinase domain, which is part of the ATP-binding pocket (2). Gurden et al. (12) described the isolation of the C604W mutant, and a C604Y mutation has been independently described by Koch et al. (11), raising resistance to Cpd-5. A crystal structure of the Mps1 kinase C604W mutant bound to NMS-P715 (12) showed how this mutation prohibits efficient binding of NMS-P715, suggesting that further development of NMS-P715 derivatives could prove necessary to combat tumor cells conferring drug resistances.
Drug development against kinases can be made more efficient by the availability of automatable direct-readout specific assays, which can be implemented in common laboratory equipment. To detect Mps1 kinase activity in vitro, the myelin basic protein is widely employed as a substrate, using radiola-beled ATP or specific phospho-Mps1 antibodies (16 -18). Alternatively, the KNL1 protein is used as substrate (11), which is a well documented Mps1 substrate of kinetochore components (19,20). These assays are highly specific for Mps1, but hard to quantify. Another more recently demonstrated way of measuring Mps1 kinase activity is a mobility shift assay described by Naud et al. (5): Phosphorylated and non-phosphorylated peptides are separated by electrophoresis based on their charges (14). This mobility shift assay has been compared with a radiometric assay and found to be more robust (21).
Here, we first describe a novel and highly specific assay for Mps1 activity by using fluorescence polarization (FP), which is based on the change of tumbling rate of a fluorescently-labeled KNL1 peptide, monitoring the binding of the 73-kDa Bub1-Bub3 complex as a high-throughput highly specific physiological readout. Using this assay, we determined the IC 50 of NMS-P715 and Cpd-5 for the Mps1 WT and C604Y/C604W mutant, and compare it to the K D against these Mps1 variants. We also present the crystal structures of the Mps1 kinase domain variants bound to Cpd-5 and NMS-P715 and discuss the structural basis of the inhibition mode of Cpd-5 over NMS-P715 and reversine in the Mps1 C604Y/C604W mutants, and how differences between the binding modes of related inhibitors could be exploited in therapy.

Cpd-5 and NMS-P715 inhibit Mps1-mediated phosphorylation of KNL1 peptides
To enable quick spectroscopic quantitation of Mps1 activity, we synthesized a fluorescent KNL1 peptide, TMR-KNL1. As shown in Fig. 1A, phosphorylation of TMR-KNL1 by Mps1 can be detected specifically by the Bub1-Bub3 complex that binds only the phosphorylated form of the peptide. Bub1-Bub3 form a tight complex with TMR-KNL1, with a dissociation constant (K D ) of 12.2 Ϯ 1.4 nM (Fig. 1B). Binding of the 73-kDa Bub1-Bub3 complex reduces the tumbling rate of the ϳ4 kDa TMR-KNL1 peptide, allowing measurement of Bub1-Bub3 binding, and therefore of phosphorylation, by FP. Addition of the Cpd-5 or NMS-P715 Mps1 inhibitors does not notably change the binding constant of the Bub1-Bub3 complex, as expected, but only the end point measurement of the FP signal, which is proportional to the degree of phosphorylation of the peptide.
Having established the conditions for the FP assay, we measured the ability of Cpd-5 and NMS-P715 to inhibit the activity of the Mps1 kinase domain on the KNL1 substrate, by titrating them into our assay mixture ( Fig. 2 and Table 1). From the titration curves, the IC 50 values were estimated to be 9.2 Ϯ 1.6 nM for Cpd-5 and 139 Ϯ 16 nM for NMS-P715. Cpd-5 and NMS-P715 inhibit the C604Y variant with a significantly worse IC 50 of 170 Ϯ 30 nM for Cpd-5 and 3016 Ϯ 534 nM for NMS-P715, respectively. A more moderate reduction was observed for the C604W substitution, with IC 50 of 19 Ϯ 1 nM for Cpd-5 and 900 Ϯ 55 nM for NMS-P715. To validate the result of our new assay, the potency of the inhibitors was assessed also by probing phospho-KNL1 protein with a phospho-specific antibody for the C604Y mutant, which results in similar IC 50 values (supplemental Fig. S1A and Table 1). These results are fully in agreement with previous studies that showed that expression of the C604Y mutant renders cells more resistance to NMS-P715 than to Cpd-5 (11).

Binding affinities of Cpd-5 and NMS-P715 to WT and C604Y/ C604W Mps1
The IC 50 value shows how effective a compound is on a specific assay, but it is not a measurement of the binding of the inhibitor; the latter information is important to understand from a mechanistic perspective the action of inhibitors and the differential effect they assert to mutant proteins. Thus, to evaluate the influence of the C604Y/C604W substitution in the inhibition mode, the binding affinities were determined by microscale thermophoresis (MST; Fig. 2 and Table 1). The MST results show that Cpd-5 binds ϳ200 -300-fold better to the wild-type Mps1 (1.6 Ϯ 0.2 nM) than to the C604Y (471 Ϯ 50 nM) and C604W variants (349 Ϯ 81 nM); a similar trend is observed for NMS-P715, which binds to the wild-type Mps1 (4.7 Ϯ 2.5 nM), to the C604Y (1764 Ϯ 204 nM) and C604W variants (630 Ϯ 115 nM). Notably, however, despite the significant reduction in binding affinity to the C604Y/C604W mutants, Cpd-5 binds better than NMS-P715. This observation is consistent with the IC 50 measurements in vitro and with cellbased assays (11), where Cpd-5 shows 5-20-fold higher potency than NMS-P715 toward inhibiting the C604Y/C604W mutants. This is of potential interest for the discovery of compounds that target resistant variants of the kinase, and is important in light of our crystallographic structure models of Cpd-5 and NMS-P715 bound to the C604Y/C604W mutants.

Mps1 C604Y/C604W mutants and inhibitors Crystal structures of Mps1 kinase domain mutants bound to Cpd-5 and NMS-P715
The Mps1 C604Y/C604W kinase domain mutants C604Y and C604W were each co-crystalized with Cpd-5 and NMS-P715. The crystal structures of the Mps1 C604Y variant in complex with Cpd-5 (PDB code 5MRB), NMS-P715 (PDB code 5NTT), and the Mps1 C604W with Cpd-5 (PDB code 5O91) were determined by molecular replacement, at 2.20-, 2.75-, 3.20-Å resolution, respectively (Table 2), whereas the NMS-P715 bound to the Mps1 C604W structure (PDB code 5AP7) was reported in a previous study (12). In all cases, there was clear electron density for Cpd-5 and NMS-P715 in the ATPbinding pocket (Fig. 3). The overall structure of the protein adopts to a conformation very similar to the previously reported structures (2). As in many structures of Mps1, the activation loop encompassing residues 676 -685 had poor density, which only for the NMS-P715 was sufficiently well resolved to make a coarse model of the loop. Thr 686 in the Pϩ1 loop has been shown to be autophosphorylated in the recombinant protein (2); although we tried to model the phosphoryl group of Thr 686 , it is not well resolved in the electron density map and therefore it has not been included in the structural models. Interestingly, the electron density indicated that a polyethylene glycol (PEG) molecule, which sometimes encapsulates the catalytic Lys 553 (2), is not present in our data. Instead, the side chain of Lys 553 forms a hydrogen bond to the O atom adjacent to the diethyl phenyl ring of Cpd-5 and NMS-P715. This shows that ligand binding to Mps1 can be misrepresented if abundant PEG is used during crystallization (36). Both inhibitors are stabilized by two additional hydrogen bond interactions with the amide backbone of the hinge loop residue, Gly 605 . The electron density of the Cys 604 point mutation is clearly resolved, indicating that the side chains of the substituted Tyr and Trp are well ordered in place.

Comparing crystal structures explains inhibitor resistance and selectivity
These structures, allow comparing the differences in the binding of different compounds to the Mps1 kinase domain (Fig. 4). The key difference between Cpd-5 and NMS-P715 is the functional group attached on the phenyl moiety (Fig. 4, A  and B). The trifluoromethoxy group of NMS-P715 and the  Table 1) can be attributed to the increased chances of steric hindrance between Tyr/Trp 604 and Ile 531 and the trifluoromethoxy moiety of NMS-P715.
Structure comparisons also explain the molecular basis of the non-resistant phenotype of the C604Y mutant toward reversine (11) (Fig. 4C); the affinity of reversine to WT and the C604Y Mps1 variant is the same within error (Table 1 and supplemental Fig. S1B). Comparison of the structure of the C604Y mutant bound to reversine (PDB code 5LJJ (23)) clearly shows that there is no steric hindrance with the Tyr 604 side chain, explaining why C604Y, and by deduction C604W, are non-resistant to reversine as previously reported (11,12).
We then compared the binding mode of the NMS-P715 and the Cpd-5 inhibitor to the Mps1 C604Y and C604W mutants (Fig. 5). NMS-P715 binds better to the Trp mutant, than to the Tyr mutant. For NMS-P715 we observe that one of the fluorines when in complex with the C604W mutant makes close contact with the H⑀ 1 of Trp 604 (2.3 Å) and an even closer contact to the H⑀ 21 of Gln 541 (2.0 Å). Particularly the latter close contact can be regarded as a good example of a rare hydrogen bond of the C-F⅐⅐⅐H-N type (37). The 3-torsion angle of Gln 541 changes by ϳ30 degrees to accommodate this interaction, in comparison to the other three structures. In the Tyr 604 mutant, however, these interactions are not formed; instead, a single hydrogen bond between H⑀ 21 of Gln 541 and the side chain oxygen of Tyr 604 is available. This explains why NMS-P715 binds better C604W; the calculated energy difference of ϳ3 kJ/mol based on the measured K D (Table 1) is compatible with the energy of a weak hydrogen bond. Cpd-5 binds with similar affinity to both mutants; which is consistent with the lack of major differences in the binding mode of this inhibitor to both mutants.

Conclusion
Our finding that binding of NMS-P715 is more affected than binding of Cpd-5 is consistent with cell studies showing that the C604Y substitution confers resistance more moderately to Cpd-5 than to NMS-P715 (11). This also indicates that combinations of different Mps1 inhibitors can be used to avoid or combat resistance in the clinic, and molecular understanding of the Mps1 interaction with inhibitors is important. Based on the different favorable interactions in the methoxy group substitutions, we suggest that substitutions of the methoxy group could be used to develop, perhaps less potent, Mps1 inhibitors that could be used to target the Cys 604 mutation, if the need arises in the clinic. For instance, inhibitors for the Tyr 604 mutant may benefit from a single or a double fluorine substitution on the methoxy group rather than the triple substitution in NMS-P715. Concluding, this study contributes in understanding the mechanism of resistance in Mps1 kinase inhibitors, suggests a

Mps1 C604Y/C604W mutants and inhibitors
new efficient and specific assay to aid Mps1 inhibitor discovery, and puts forward novel design principles for the further development of this class of inhibitors.

Protein production
The plasmid containing a construct of the Mps1 kinase domain (residues 519 -808) was a gift from Dr. Nicola Burgess-Brown (Addgene plasmid number 38907) (22). The site-specific mutant of C604W was generated using QuikChange (Stratagene). The Mps1 kinase domain C604Y/C604W variants were produced as previously described (23).
The pFastBac plasmid containing the GST-tagged full-length Mps1 was a gift from Dr. Geert Kops (Hubrecht Institute, Utrecht, The Netherlands). Recombinant baculovirus was produced according to the Bac-to-Bac protocols (Invitrogen). Spodoptera frugiperda (Sf9) insect cells were infected with the baculovirus and allowed to grow for 72 h at 27°C. Cells were harvested by centrifugation and re-suspended in 50 ml of 20 mM potassium phosphate, pH 7.5, 150 mM KCl, and 1 mM tris(2carboxyethyl)phosphine (TCEP) (buffer A) supplemented with one tablet of Pierce Protease TM Inhibitor Tablets EDTA-free (Thermo Fisher Scientific). Samples were stored at Ϫ20°C before proceeding to purification. The re-suspended cells were defrosted at room temperature and lysed by sonication for 1 min at 50% amplitude in a Qsonica Sonicator Q700 (Fisher Scientific). Following centrifugations at 21,000 ϫ g for 20 min at 4°C, the supernatant was loaded on glutathione-Sepharose 4B (GE Healthcare). After extensive washing in buffer A, the protein was eluted in buffer A supplemented with 10 mM GSH. The sample was then loaded on a Superdex G75 16/60 HiLoad (GE Healthcare) pre-equilibrated in 20 mM HEPES/ NaOH, 50 mM KCl, and 3 mM DTT. The protein fractions were pooled together and concentrated to ϳ10 M. The concentration of the GST-Mps1 full-length was determined by absorption spectrophotometry at 280 nm, with calculated ⑀ ϭ 122.8 mM Ϫ1 cm Ϫ1 . The purified protein was aliquoted and stored at Ϫ80°C.
Bub1 (residues 1-280) and Bub3 were produced as previously described (19), with modifications. The baculovirus with the Bub1 and Bub3 constructs was a gift from Dr. Geert Kops. Sf9 insect cells were infected and allowed to grow for 72 h at 27°C. The cell cultures were harvested by centrifugation and re-suspended in 50 mM Tris-HCl, pH 7.7, 1 mM TCEP, and 0.05% Tween 20 (buffer B) supplemented with 300 mM KCl, 10 mM imidazole, and one tablet of Pierce Protease TM Inhibitor Tablets EDTA-free (Thermo Fisher Scientific). Samples were stored at Ϫ20°C before proceeding to the purification. The re-suspended cells were defrosted at room temperature and lysed by sonication for 1 min at 50% amplitude in a Qsonica Sonicator Q700 (Fisher Scientific). Following centrifugation at 21,000 ϫ g for 20 min at 4°C, the supernatant was loaded on a

Mps1 C604Y/C604W mutants and inhibitors
HisTrap HP column (GE Healthcare). After extensive washing in buffer B supplemented with 300 mM KCl and 10 mM imidazole, the protein was eluted with buffer B supplemented with 150 mM imidazole. The sample was then loaded on a Superdex G75 16/60 HiLoad (GE Healthcare) pre-equilibrated in buffer B supplemented with 100 mM KCl. The protein fractions were pooled together and concentrated. The concentration of the Bub1-Bub3 complex was determined by absorption spectrophotometry at 280 nm, with calculated ⑀ ϭ 101.9 mM Ϫ1 cm Ϫ1 . The purified protein was aliquoted and stored at Ϫ80°C.

Western blot-based Mps1 kinase activity assay
100 ng of Mps1 wild-type or Mps1 C604Y was incubated for 10 min at 32°C with the indicated compounds in a buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 20 mM MgCl 2 , 1 mM DTT, 0.2 mM ATP. After this incubation, 200 ng of purified GST-KNL1-M3 was added, and the mixture was incubated for 60 min at 32°C before addition of SDS-sample buffer and heating at 95°C for 10 min. Samples were separated by SDS-PAGE and transferred to nitrocellulose membranes, blocked with 4% (w/v) bovine serum albumin at room temperature for 0.5 h, and incubated with primary antibody (phospho-KNL1 MELT13/ 17), a kind gift from Dr. Geert Kops (24) at 4°C overnight. After incubation with secondary antibody (1:2000 dilution) at room temperature for 1 h, the membranes were developed with chemiluminescence ECL reagent (Amersham Biosciences) and pictures were taken with the ChemiDOC XRSϩ (Bio-Rad). Resulting images were analyzed using ImageJ software.

Fluorescence-based Mps1 kinase activity assay
The Bub1-Bub3 complex was titrated to a reaction mixture containing 20 nM TMR-KNL1 p and 10 nM GST-Mps1 fulllength variants in 20 mM HEPES/NaOH, pH 7.4, 1 mM ATP, 4 mM MgCl 2 , 200 M TCEP, and 0.05% Tween 20 (buffer C) with or without 50 nM inhibitors. The samples were incubated at room temperature in a 384-well Corning assay plate. All measurements were performed in a Pherastar plate reader (BMG LABTECH GmbH, Germany). The excitation and emission wavelengths were 540 and 590 nm, respectively, and FP was calculated. All measurements were carried out in duplicate.
To calculate the IC 50 values, the inhibitors were titrated to a reaction mixture containing 20 nM TMR-KNL1 peptide, 200 nM Bub1-Bub3 complex, and 5 nM GST-Mps1 full-length variants in buffer C. The samples were incubated at room temperature overnight. Measurements were carried in duplicates. The FP values were plotted against the inhibitor concentration and fitted with a standard one-site model (Equation 1) using nonlinear regression in GraphPad Prism 6 (GraphPad Software, Inc.), where FP max is maximum binding in FP; [c] is inhibitor concentration; FP bg is the background FP value.

Microscale thermophoresis
The thermophoresis measurements and data analysis were performed as previously described (23) with a slight modification. The DY-547P1-labeled samples were used at a final concentration of 20 nM in Tris buffer (50 mM Tris-HCl, pH 7.4, 150 mM KCl, 1% DMSO, 0.05% Tween 20). The measurement was performed at 20% LED and 40% MST power.

Mps1 C604Y/C604W mutants and inhibitors
at 18°C within 72 h. Crystals were briefly transferred to a cryoprotectant solution containing the reservoir solution and 20% (w/v) ethylene glycol and vitrified by dipping in liquid nitrogen.

Data collection and structure refinement
X-ray data were collected on beamline ID30A-3, ID29, and ID30A-1 at the European Synchrotron Radiation Facility (ESRF) for 5MRB, 5NTT, and 5O91 (PDB ID codes), respectively. The images were integrated with XDS (26) and merged and scaled with AIMLESS (27). The starting phases were obtained by molecular replacement using PHASER (28) with an available Mps1 structure (PDB code 3HMN) (29) as the search model. Geometric restraints for the compounds were made in AceDRG (30). The models were built using COOT (31) and refined with REFMAC (32) in iterative cycles. Model re-building and refinement parameter adjustments were performed in PDB-REDO (33,34); homology-based hydrogen bond restraints (van Beusekom et al. (39)). were used at some stages of the procedure. The quality of the models was evaluated by MOLPROBITY (35). Data collection and refinement statistics are presented in Table 2.