Structure-Based Design of Orally Bioavailable 1H-Pyrrolo[3,2-c]pyridine Inhibitors of Mitotic Kinase Monopolar Spindle 1 (MPS1)

The protein kinase MPS1 is a crucial component of the spindle assembly checkpoint signal and is aberrantly overexpressed in many human cancers. MPS1 is one of the top 25 genes overexpressed in tumors with chromosomal instability and aneuploidy. PTEN-deficient breast tumor cells are particularly dependent upon MPS1 for their survival, making it a target of significant interest in oncology. We report the discovery and optimization of potent and selective MPS1 inhibitors based on the 1H-pyrrolo[3,2-c]pyridine scaffold, guided by structure-based design and cellular characterization of MPS1 inhibition, leading to 65 (CCT251455). This potent and selective chemical tool stabilizes an inactive conformation of MPS1 with the activation loop ordered in a manner incompatible with ATP and substrate-peptide binding; it displays a favorable oral pharmacokinetic profile, shows dose-dependent inhibition of MPS1 in an HCT116 human tumor xenograft model, and is an attractive tool compound to elucidate further the therapeutic potential of MPS1 inhibition.


■ INTRODUCTION
The main role of the cell cycle is to enable error-free DNA replication, chromosome segregation, and cytokinesis. Surveillance mechanisms, the checkpoint pathways, monitor passage through the cell cycle at several stages. During mitosis, the spindle assembly checkpoint (SAC) prevents anaphase onset until the appropriate tension and attachment across kinetochores is achieved. 1,2 One of the first components of the SAC signal, identified by a genetic screen in budding yeast, was dubbed MPS1 (monopolar spindle 1, also known as TTK) because of the monopolar spindles produced by MPS1 mutant cells. 3 Subsequently, the MPS1 gene was shown to encode an essential dual-specificity kinase conserved from yeast to humans. 4,5 MPS1 activity peaks at the G2/M transition, is enhanced upon activation of the SAC with nocodazole, 6 and is dependent upon autophosphorylation of a threonine at position 676 in the activation loop. 7 MPS1 is required for normal function of the mitotic spindle checkpoint and subsequent cell division; it is aberrantly overexpressed in a wide range of human tumors including bladder, anaplastic thyroid, breast, lung, esophagus, and prostate cancers. 8−12 In addition, MPS1 has been identified in the signature of the top 25 genes overexpressed in tumors with chromosomal instability 13 and aneuploidy, 14,15 with PTENdeficient breast tumor cells particularly dependent upon MPS1 for their survival such that RNAi-mediated knockdown or chemical inhibition of MPS1 leads to cell death. 14 This body of work has engendered significant interest in the discovery of selective smallmolecule chemical tools to elucidate further the therapeutic potential of MPS1 inhibition for the treatment of cancer.
Here, we describe the discovery of orally bioavailable smallmolecule inhibitors of MPS1 based on the 1H-pyrrolo[3,2-c]pyridine scaffold in a medicinal-chemistry program enabled by structure-based design and cellular characterization of MPS1 inhibition. We show that optimized compounds in this series display potent and selective inhibition of MPS1 in vitro and translate well to cellular assays of MPS1 autophosphorylation and antiproliferative activity when compared to other recently reported MPS1 inhibitors. We also show, by X-ray crystallographic studies, that exemplars of this series stabilize an inactive conformation of MPS1 in which the activation loop is ordered in a manner incompatible with ATP and substrate-peptide binding.
Our starting point was the potent but nonselective and metabolically unstable compound 8 identified in a high-throughput screen of an in-house kinase-focused compound library ( Figure 2).

■ CHEMISTRY
Our synthetic strategy to all desired 1H-pyrrolo [3,2-c]pyridines involved palladium-mediated Sonagashira coupling of an appropriately substituted 4-amino-2-bromo-5-iodopyridine, 9, and an alkyne, 10, to generate key pyrrolopyridine intermediate 11, a domino approach recently exemplified by Schmidt and colleagues. 28 Subsequent palladium-mediated displacement of the 6-bromo substituent of intermediates 11 with the appropriate aniline gave the desired products 12 (Scheme 1). This general route was adapted depending on the identity of the N-1 and C-2 substituents as described below. We consistently observed that when R 1 = H the transformation of 11 to 12 was low-yielding; therefore, we employed a protecting-group strategy whereby a Boc substituent was installed at the N-1 position prior to introduction of the C-6 amino substituent into the 1H-pyrrolo[3,2-c]pyridine scaffold.
Compounds 8 and 21−28 shown in Table 1 were prepared according to the general strategy depicted in Scheme 1 using palladium-mediated substitution of key 6-bromo-pyrrolopyridine intermediate 17, which was itself prepared by sequential Sonagashira cross coupling and base-catalyzed ring closure of sulfonamide 16; introduction of the sulfonamide was necessary to optimize the efficiency of the domino cyclization reaction, presumably by increasing the acidity of the remaining anilinic proton (Scheme 2). Cyclization precursor 16 was prepared from corresponding 4-amino-2-bromopyridine 13; iodination of 13 was unselective, and desired regioisomer 15 was purified from its partner, 14, by chromatographic separation in 38 and 37% yields, respectively; subsequent dimesylation with methanesulfonylchloride and base-mediated removal of one of the two mesyl groups provided intermediate 16 in 54% yield. The required tert-butyl 4-ethynyl-1H-pyrazole-1-carboxylate 20 was prepared  in 56% overall yield from 4-iodopyrazole 18 by Boc protection followed by Sonogashira-mediated coupling with trimethylsilylacetylene and subsequent TBAF-mediated deprotection of alkyne 19 (Scheme 2).
Scheme 5. Preparation of 1H-Pyrrolo[3,2-c]pyridin-2-yl)oxazoles a microsomes, and significant efflux in Caco-2 permeability assays ( Figure 2). An important aim of our initial hit-improvement strategy was to eradicate activity versus cell cycle kinases (e.g., CDK2) and other kinases known to affect mitotic function (e.g., Aurora kinases A and B) to study the profile of a highly selective MPS1 inhibitor on mitotic function in cellular mechanistic assays and in vivo. In addition, we set out to improve the metabolic stability and membrane permeability of compound 8 to discover a chemical tool suitable for in vivo PK/PD studies.
A crystal structure of the kinase domain of MPS1 with compound 8 (Figure 3) indicated binding of 8, albeit with relatively weak electron density. Nevertheless, the ligand could be modeled with partial occupancy along with a molecule of poly(ethylene glycol) wrapped around the active-site Lys553 side chain, a consequence of the presence of a high concentration of PEG300 in the crystallization conditions. The structure revealed the 6-amino-pyrrolopyridine motif interacting with the hinge region of the ATP-binding site by virtue of an H-bond-donor interaction between the backbone amide group of Gly605 and the pyridine nitrogen hydrogenbond acceptor of the pyrrolopyridine scaffold. In addition, the anilinic NH of compound 8 formed a hydrogen bond with the carbonyl group of hinge residue Gly605, thereby positioning the anilinic moiety at the entrance of the MPS1 ATP-binding site, stacked above the post-hinge region (residues 606−611) and pointing toward the solvent. Furthermore, it revealed an H-bond between the C-2 pyrazole and Lys553 as well as a van der Waals interaction between lipophilic C-3 to C-4 atoms and the gatekeeper residue, Met602 ( Figure 3).
A striking difference between the binding of compound 8 and published compound 6 is their respective hydrogen-bond interactions with the hinge. Whereas the backbone functionalities of hinge residue Cys604 were not involved in interactions with compound 8, a peptide flip of Cys604 in the structure of MPS1 complexed with compound 6 (PDB code 3VQU) allowed an H-bond interaction between the backbone carbonyl of Cys604 and the anilinic NH of compound 6. 26 We initially focused our attention on modification of the pyrrolopyridine 6-anilino substituent to replace the electronrich 3,4-dimethoxy aniline, which we regarded as a metabolic liability. 4-Methoxy analogue 21 proved equipotent and replacement with a range of 4-substituents maintained activity (compounds 22−24), consistent with the crystal structure of 8 bound to MPS1, which showed that this vector projects out of the entrance of the ATP-binding site into the solvent ( Figure 3). However, selectivity versus CDK2 remained poor in all of these compounds. Importantly, the 2-methoxy-, 2-ethoxy-, and 2-chloro-substituted aniline derivatives, 25−28, maintained potency while also significantly improving selectivity versus CDK2 and enhancing selectivity versus Aurora A, Aurora B, and GSK3β. This SAR is consistent with previous reports on the use of 2-substituted anilines in other chemical series to enhance selectivity versus MPS1 through the exploitation of a small lipophilic pocket adjacent to Cys604 in the hinge region (see below). 25 Compounds 25−28, although selective for MPS1, remained metabolically unstable, and our attention turned to exploration of the pyrrolopyridine C-2 substituent. This was prompted by our observation that solutions of compound 8 underwent slow air oxidation across the double bond between C-2 and C-3 (Supporting Information Figure S1). We hypothesized that the unsubstituted C-2 pyrazole rendered the pyrrole moiety of the pyrrolopyridine scaffold susceptible to electrophilic attack. Gratifyingly, N-methylation of the C-2 pyrazole was tolerated with only a 5-fold reduction in potency (compound 29, MPS1 IC 50 = 0.12 μM, versus compound 25, MPS1 IC 50 = 0.025 μM; Table 2), resulting in an air-stable compound and a slight improvement in metabolic stability (MLM = 72% for 29 versus 99% for 25) (Table 2). Similarly, electron-withdrawing trifluoroethyl-substituted pyrazole analogue 30 maintained potency (MPS1 IC 50 = 0.079 μM) and also improved metabolic stability (53% turnover in MLM); however, corresponding difluoromethyl analogue 31 proved surprisingly weak, with a 6-fold loss of activity with respect to the trifluoroethyl analogue (IC 50 = 0.46 μM for 31 versus 0.079 μM for 30). 1,3-and 1,5-Disubstituted pyrazole analogues (32 and 33) and the 3,5-disubstituted isoxazole 35 also lost potency despite an improvement in metabolic stability; we rationalized this loss of potency in terms of the sterically encumbered pocket into which the C-2-pyrazole substituent projects. Imidazole analogue 36 also lost potency in comparison with pyrazole 25 despite the presence of a potentially isosteric H-bond-donor or -acceptor interaction with Lys553, depending on the protonation state of the imidazole. Taken together, these results suggested tight SAR along the C-2 vector from the pyrrolopyridine scaffold and were consistent with the crystal structure of MPS1 complexed with compound 8 that showed that the interaction with Lys553 is important ( Figure 3). Unsubstituted oxazole 34 proved to be the only potential C-2 pyrazole replacement that maintained potency with enhanced metabolic stability. The crystal structure of MPS1 in complex with oxazole 34 showed unambiguous electron density for the ligand, supporting the initial binding mode for compound 8 and consistent with the SAR observed for the aniline and C-2 heterocycle modifications made to the pyrrolopyridine scaffold ( Figure 4). The hinge-binding motif was the same as that observed with compound 8, and the aniline C-2-methoxy substituent was positioned, as expected, in a small hydrophobic pocket lined by Lys529, Ile531, Gln541, and the gatekeeper +2 residue, Cys604. This pocket is not accessible in many other kinases, including CDK2, GSK3β, Aurora A, and Aurora B, which have bulkier residues at the corresponding gatekeeper +2 position (Phe in CDK2 and Tyr in GSK3β and Aurora A and B). This is consistent with improved selectivity for MPS1 versus related kinases observed for compounds with an aniline 2-substituent (Tables 1 and 2). 25 Like the pyrazole of compound 8, the oxazole of 34 was oriented toward the catalytic Lys553 residue, and an H-bond was observed between the pyrrolopyridine N-1 atom and a water molecule. However, despite the presence of a preferred aniline C-2-substituent, the weak CDK2 activity of 34 coupled with the relative complexity of the synthetic route to C-2 oxazoles (Scheme 5) disfavored this approach. We selected 1-methylpyrazole in preference to the 1-trifluoroethyl pyrazole as the optimal pyrrolopyridine C-2 substituent because of its lower lipophilicity and improved ligand efficiency. Compound 8 is shown with orange carbon atoms and is modeled with partial occupancy along with a PEG molecule, shown with orange and cyan carbon atoms for the two alternate conformers. Selected amino acids that contact the ligand are shown with green carbon atoms. The electron density shown in green is from an F o − F c omit map and is contoured at 3σ. Key H-bond interactions are shown as black dotted lines. The interaction between the C-2 pyrazole and Lys553 has been omitted for clarity. All structural figures were produced with CCP4MG. 31  We then investigated a range of aniline substitutions with the aim of further improving metabolic stability by reduction of both lipophilicity and electron density in the aniline moiety. 2-Methoxy-5-trifluoromethyl analogue 37 (IC 50 = 4.4 μM; Table 2) illustrates poor tolerance of a 2,5-disubstitution pattern on the aniline ring. Analysis of the compound 34-bound MPS1 structure suggested that the addition of a CF 3 substituent to the 5-position of the aniline ring would induce a steric clash with Asp608 ( Figure 4). This observation is consistent with the SAR described for a series of Leucine Rich Repeat Kinase 2 (LRRK2) inhibitors in which a 2,5disubstituted aniline was employed to drive selectivity for LRRK2 over MPS1. 32 Exploitation of the aniline C-4 vector, which extends into the solvent channel ( Figure 3), was more successful and led to the synthesis of compounds 39−44, all of which displayed good potency compared to their unsubstituted parent 38, improved selectivity, and in vitro metabolic stability (Table 2). However, the measured aqueous thermodynamic solubility was low (e.g., 0.01 mg/mL for compound 42).
2-Chloro-4-dimethylcarboxamido-substituted aniline 39 was selected for pharmacokinetic evaluation on the basis of its excellent potency, in vitro selectivity, and improved metabolic stability in mouse and human liver microsomes (25 and 20% turnover after a 30 min incubation, respectively). This compound displayed an improved efflux ratio in Caco-2 (10) compared to original hit compound 8 and demonstrated good in vivo pharmacokinetics in mouse with a low unbound clearance and moderate oral bioavailability, consistent with our strategy of targeting improved in vitro metabolic stability versus compound 8 (Table 3).
A crystal structure of compound 39 bound to MPS1 was obtained by soaking MPS1 crystals for 24 h in a solution containing 1.25 mM of the inhibitor. The crystal structure showed that the overall binding mode of 39 was very similar to those of HTS hit compound 8 and oxazole 34. However, the crystallographic data revealed significant electron density along a vector aligned from the pyrrolopyridine N-1 position ( Figure 5A), which could not be explained by the interacting water molecule observed in the compound 8-bound and compound 34-bound structures. Analysis of the compound sample used for the soaking experiment revealed a 2% impurity of the N-Boc synthetic precursor (48; Table 5), which fitted well with the additional electron density observed in the crystal structure. We reasoned that to preferentially occupy the ATP-binding site in a soaking experiment involving a 50-fold excess of compound 39, compound 48 must be significantly more potent than compound 39. Subsequent elucidation of the crystal structure of MPS1 bound to the N-Boc-containing precursor, compound 48, confirmed this hypothesis because the overall binding mode of 48 was entirely consistent with the other crystal structures and the Boc group was clearly present in the same location as the additional electron density in the compound 39-bound structure ( Figure 5B). The aniline C-2-chloro substituent of compound 48 was located in the same lipophilic pocket as the aniline C-2-methoxy group in compound 34, consistent with the improved selectivity afforded by small lipophilic substituents in the aniline C-2 position (Tables 1 and 2).
Compound 48 displayed potency at the low end of the dynamic range of our in vitro MPS1 assay (IC 50 = 0.006 μM; Table 5; MPS1 enzyme concentration = 3−12.5 nM, see the Supporting Information). We therefore set up a highthroughput cell-based assay that measures the inhibition of ectopic MPS1 autophosphorylation at Thr33 and Ser37 using   Table 5). We hypothesized that a combination of increased in vitro potency and lipophilicity-driven cell penetration was responsible for the increase in cellular potency. Gratifyingly, compound 48 not only retained metabolic stability in mouse and human liver microsomes (48 and 34% turnover after a 30 min incubation, respectively; Table 5) despite increased lipophilicity (48 AlogP = 5.1, 39 AlogP = 3.0) but also displayed reasonable chemical stability in aqueous acid and base (38% cleavage of the Boc group was observed after a 75 min incubation of 48 in a simulated gastric acid fluid at 37°C, and no cleavage was observed after a 5 h incubation of 48 in a simulated duodenum solution at 37°C), indicating that the N-1-Boc substituent of compound 48 may survive gut media on oral administration. Moreover, the increased lipophilicity of 48 improved passive permeability (PAMPA >100 × 10 −6 cm/s at pH 7.4) and abrogated Caco-2 efflux (A to B = 17 × 10 −6 cm/s, efflux ratio = 1). In vivo pharmacokinetic profiling in mouse revealed increased clearance, consistent with higher lipophilicity and MLM metabolic turnover (48% for 48 at 30 min versus 25% for 39), increased volume of distribution, and higher bioavailability compared to compound 39, consistent with increased lipophilicity and passive permeability (Table 4).

Journal of Medicinal Chemistry
Finally, we were concerned that increased lipophilicity imparted by our serendipitous discovery of the influential N-Boc substituent might erode the in vitro selectivity profile; however, selectivity versus CDK2 and Aurora A was maintained (Table 5), and we also observed complete selectivity over other mitotic kinases, for example, NIMA-related kinase 2 (NEK2) and Polo-Like Kinase 1 (PLK1) (IC 50 > 100 μM). Thus, compound 48 resolved many of our issues with original hit compound 8, and we elected to maintain the N-1-Boc substituent in further analogues. We next turned our attention to improvement of the cell-based GI 50 in HCT116 cells, which remained relatively weak for compound 48 (GI 50 = 2.20 μM; Table 5).
As expected, further exploration of the aniline C-4 vector in the N-Boc-substituted pyrrolopyridine series revealed broad tolerance for a variety of substituents, with optimal translation to cell-based potency observed for azetidine amide 51, piperidine amides (52 and 53), and thiomorpholine 1,1-dioxide amide 54. Consistent with previous SAR, we were pleased to note that C-2-oxazole 55 was also tolerated in this series (Table 5), and the crystal structure of 55 bound to MPS1 confirmed that the oxazole maintains an interaction with Lys553 ( Figure 6), consistent with the structure of MPS1 with compound 34. However, neither the C-2-oxazole nor the C-2-pyrazole compounds with variations at the aniline C-4 vector provided a significant improvement in cellbased antiproliferative activity ( Table 5).
Analysis of the crystal structures of compounds 48 ( Figure 5B) and 55 ( Figure 6) showed that the aniline C-4-dimethylamido substituent projected toward the solvent above the Asp608− Ser611 helix-capping motif in the post-hinge region of the kinase.   This suggested that replacement of the aniline 4-amido substituent with an appropriate heterocycle may be tolerated, and we were keen to explore the effect of this modification on cellular potency (Table 6). Gratifyingly, C-4-pyrazolo analogues 61, 62, and 63 proved to be potent inhibitors of MPS1 in the in vitro biochemical assay, with acceptable metabolic stability in mouse and human liver microsomes. Analogous to our observations with compounds 39 and 48, we observed a significant (43-fold) increase in inhibition of MPS1 autophosphorylation in cells for N-1-Bocsubstituted compound 61 versus its N-1-H analogue 62 (P-MPS1 IC 50 = 0.16 μM versus 6.90 μM), and this improvement was also observed in an assay of cell proliferation (61, HCT116 GI 50 = 0.50 μM; 62, HCT116 GI 50 = 4.60 μM). Although oxazolesubstituted analogue 64 and those substituted with six-membered heterocycles, 66−68, all maintained potent inhibition of MPS1 in vitro, translation to cell-based activity was not improved compared to pyrazole 61 (Table 6). However, the 1-methyl-imidazol-5-yl moiety at the 4-position of the aniline (compound 65) conferred improved translation to cell-based potency (P-MPS1 IC 50 = 0.04 μM and HCT116 GI 50 = 0.16 μM), which is comparable to or better than the cell-based potency of reported MPS1-selective inhibitors tested in our assays (Table 7). Compound 65 displayed in vitro potency versus MPS1 at the low end of the dynamic range of our in vitro assay, which together with an excellent translation to cell-based assays prompted further analysis of the binding mode of 65 by X-ray crystallography ( Figure 7A). The structure was determined by cocrystallization of the kinase domain of MPS1 with 65 using PEG3350 as the precipitant instead of PEG300 in an attempt to remove the artifactual PEG molecule bound in the ATP-binding site. This resulted in a more physiologically relevant structure without a PEG molecule wrapped around the active-site Lys553 residue, which allowed for the formation of the conserved Lys553−Glu571 ion-pair. Although the binding mode of compound 65 was entirely consistent with our previous compound-bound crystal structures, MPS1 activation-loop residues Ala668−Thr675 adopt an ordered conformation, and the ordered loop forms an antiparallel β-sheet interaction with the P-loop ( Figure 7B). In addition, activationloop residues Met671−Pro673 form a complementary hydrophobic pocket wrapped around the N-1-Boc substituent of 65, completely enclosing the inhibitor in the ATP-binding site. Intriguingly, this activation-loop conformation has previously been observed in the crystal structure of MPS1 with a pyrimidodiazepine ligand (PDB code 3H9F) 20 and is incompatible with binding of a PEG molecule around Lys553 because of steric hindrance. In this structure, a triply phosphorylated Thr−Thr− Ser motif at residues 675−677 formed magnesium-mediated crystal contacts, which may have influenced the activation-loop conformation. A careful analysis of two recently reported crystal structures of MPS1, one in complex with a diamino-pyridine inhibitor (PDB code 3VQU) 26 and one with an early indazolebased inhibitor (PDB code 3W1F), 27 also showed a similarly ordered activation loop in both structures but with the Thr− Thr−Ser motif disordered. In our compound 65-bound MPS1 structure, Thr676 and Ser677 are also disordered, and we did not observe electron density for a phosphate group on Thr675 or for the mediating magnesium atoms. This suggests that in our 65bound MPS1 structure these residues are not involved in meaningful crystal contacts that could have an effect on the conformation of the activation loop. Further analysis of the crystal packing showed that the only residue of a symmetryrelated molecule that is in the vicinity of the activation loop is Ile738. However, a comparison of several compound-bound structures (Supporting Information Figure S2A) showed that Ile738 is located in a region of the protein with only minor conformational flexibility, whereas the activation loop in the respective structures shows a wide range of conformations and a varying degree of order. The absence of any concerted conformational changes between the activation loop and the symmetryrelated Ile738 region in these structures led us to conclude that the ordering of the activation loop is not influenced by crystal contacts. Nevertheless, the ordering of the activation loop is clearly prevented by the binding of a PEG molecule in MPS1 structures resulting from the PEG300-containing crystallization conditions. This is supported by soaking MPS1 crystals grown in PEG300 with compound 65, which resulted in a 65-bound MPS1 structure with a disordered activation loop (data not shown).
The structures of MPS1 in complex with the precursors of the diamino-pyridine and indazole-based inhibitors 6 26 and 7 27 show an ordering of the activation loop through interactions with the P-loop and an ethoxy-group, which in both inhibitors is located in a similar position as the N-1-Boc substituent in 65. Moreover, in the pyrimidodiazepine-bound MPS1 structure, the inhibitor interacts with the ordered activation loop via a cyclopentyl moiety in a similar position as the N-1 Boc in compound 65. Taken together, these findings support the hypothesis that the ordering of the activation loop might have a compound-dependent component for inhibitors with substituents similar to the N-1 Boc in compound 65.
It is important to note that the overall structure of the MPS1 kinase domain in the compound 65-bound structure has many features of an active kinase conformation. These include the positioning of the αC-helix and the conserved DFG motif in an "in" conformation as well as the presence of the canonical active Lys553−Glu571 ion pair. Further analysis of the conformation in light of the "spine concept" 33 clearly shows the presence of the catalytic C-spine and shows only minor distortions in the regulatory R-spine (Supporting Figure 2B), also indicating that the conformation of MPS1 in this structure is close to an active kinase conformation. However, the compound-induced conformation of the activation loop is incompatible with ATP binding and substrate-peptide binding to the kinase because it blocks the phosphate-binding region and the peptide binding site is not formed, a situation that would certainly render MPS1 inactive ( Figure 7D).
Introduction of the methylimidazole into compound 65 resulted in compromised passive permeability (PAMPA = 18 × 10 −6 cm/s at pH 7.4) and increased efflux (Caco-2 A to B = 5 × 10 −6 cm/s, B to A = 12 × 10 −6 cm/s, ER = 2.5) compared to compound 48. Although the thermodynamic aqueous solubility of compound 65 was very low (<0.001 mg/mL), solubility in fasted-and fed-state-simulated intestinal fluid (FaSSIF and FeSSIF) was 0.01 and ∼0.55 mg/mL, respectively, consistent with the presence of a weakly basic center (the methylimidazole ring). All other in vitro properties were maintained, and in view of its favorable in vitro profile, compound 65 was selected for more extensive in vitro profiling versus a panel of 121 kinases (Supporting Information  Table S1) and in vivo pharmacokinetic evaluation. Of the 121 kinases in the panel, MPS1 showed the greatest inhibition by 65, and only three other kinases showed inhibition greater than 80%. Mouse and rat blood pharmacokinetics revealed a favorable profile with moderate clearance and good to moderate oral bioavailability (Table 8). This compound was progressed to a human tumor xenograft model to test whether pharmacodynamic biomarker modulation could be achieved in vivo. Oral administration of two doses of compound 65 at 50, 75, and 100 mg/kg b.i.d. to mice bearing HCT116 human colon carcinoma xenografts demonstrated dose-dependent modulation of MPS1-driven phospho-histone H3 levels versus control animals at 2 and 10 h but not at 72 h after the last dose, consistent with engagement of MPS1 in vivo ( Figure 8A). The compound was well-tolerated at all doses, and the observed decrease in phospho-histone H3 inhibition over time by compound 65 tracks with a decrease in total plasma and tumor tissue exposure measured in the same experiment ( Figure 8B,C).

■ CONCLUSIONS
We describe the structure-based optimization of a potent but nonselective and metabolically unstable 1H-pyrrolo[3,2-c]pyridine HTS hit 8 to compound 65, a highly potent inhibitor of MPS1 that demonstrates high selectivity versus kinases tested in a broad kinome profiling panel. We observed excellent translation of in vitro biochemical potency versus isolated MPS1 enzyme to cell-based potency (P-MPS1 IC 50 = 0.04 μM and HCT116 GI 50 = 0.16 μM), which is comparable to or better than the cell-based potency of other literature-reported MPS1-selective inhibitors tested side-by-side in our assays. Medicinal-chemistry optimization to 65 was educated by structure-based design; in particular, incorporation of an N-1-carbamate substituent was inspired by our observation that the activation loop of MPS1 became ordered in the presence of this substituent. The crystal structure of 65 in MPS1 confirmed this activation-loop stabilization, which results in an occluded ATP-binding site and is incompatible with ATP and substrate binding. Despite the increased lipophilicity imparted by the carbamate moiety, 65 demonstrates a good oral pharmacokinetic profile in mouse and rat as well as inhibition of MPS1 activity in vivo following oral administration; 65 is a suitable chemical probe 35 for cell-based assays and in vivo evaluation of the effect of MPS1 inhibition in human tumor xenograft models.

■ EXPERIMENTAL SECTION
Chemistry. Commercially available starting materials, reagents and dry solvents were used as supplied. Flash column chromatography was Table 6. Effect of the Aniline Substituent on Cell-Based Potency a MLM/HLM: percentage of parent compound metabolized after a 30 min incubation in mouse and human liver microsomes. b n = 1.       (15). 4-Amino-2-bromopyridine 13 (22.8 g, 131.8 mmol) and sodium acetate (20.8 g, 254 mmol) were stirred in AcOH (82 mL), and a solution of iodine monochloride (1 M in AcOH, 134 mL, 134 mmol) was added. The mixture was stirred and heated at 75°C for 3 h. Most of the AcOH was evaporated, and the residue was partitioned between water and EtOAc. The aqueous fraction was again extracted with EtOAc. The combined extracts were washed twice with 10% sodium carbonate solution, 10% sodium thiosulfate solution, water, and brine, dried, and evaporated. This gave 40.3 g of a crude product that was combined with the product from a reaction on 7.5 g of 4-amino-2bromopyridine. Purification by chromatography on a silica column (9 cm internal diameter with 28 cm bed of silica) eluting with 5% EtOAc in CH 2 Cl 2 , then 10% EtOAc in CH 2 Cl 2 , and then 20% EtOAc in CH 2 Cl 2 gave desired isomer 15 (20.2 g, 38%). 1  N-(2-Bromo-5-iodopyridin-4-yl)methanesulfonamide (16). 4-Amino-2-bromo-5-iodopyridine 15 (3.055 g, 10.2 mmol) was stirred in CH 2 Cl 2 (34 mL), and Et 3 N (6.9 mL, 49.1 mmol) was added. The mixture was cooled in ice. To the cold solution was added dropwise a solution of methanesulfonyl chloride (3.2 mL, 40.6 mmol) in CH 2 Cl 2 (11.5 mL) over a period of 14 min. The cold bath was removed, and the reaction was stirred at rt for 1.5 h. The reaction was diluted with CH 2 Cl 2 and washed twice with water. The solution was dried and evaporated. Trituration with ether gave a solid (5.01 g). The crude product was passed in 5% EtOAc in CH 2 Cl 2 through a 2.5 cm pad of silica in a 10 cm diameter sinter to give N-(2-bromo-5-iodopyridin-4yl)-N-(methylsulfonyl)methanesulfonamide (3.01 g, 64%). 1  tert-Butyl 4-((Trimethylsilyl)ethynyl)-1H-pyrazole-1-carboxylate (19). 4-Iodopyrazole 18 (7.85 g, 40.4 mmol) was dissolved in THF (120 mL), and Et 3 N (8.5 mL, 60.5 mmol) and di-tert-butyl dicarbonate (9.7 g, 44.5 mmol) were added. The reaction was stirred at rt for 3 h. The THF was evaporated, and EtOAc was added. The solution was washed with water and brine, dried, and evaporated to leave an oil (14.2 g). The crude product was purified by chromatography on a pad of silica in a sinter (10 cm diameter, 6 cm thick) eluted with 10% EtOAc in cyclohexane and then 20% EtOAc in cyclohexane to give tert-butyl 4-iodo-1H-pyrazole-1carboxylate (11.66 g, 98%). 1 1H). tert-Butyl 4-iodo-1H-pyrazole-1carboxylate (4.67 g, 15.9 mmol) and trimethylsilylacetylene (2.18 g, 22.2 mmol) were dissolved in DMF (22 mL) and placed under argon. Diisopropylamine (2.9 mL, 20.7 mmol), copper(I) iodide (197 mg, 1.03 mmol), triphenylphosphine (832 mg, 3.18 mmol), and palladium acetate (239 mg, 1.06 mmol) were added, and the flask was flushed again with argon. The reaction was heated at 60°C for 1.25 h. The reaction was cooled and added to water. The product was extracted with ether. The combined extracts were washed with water and brine, dried, and evaporated. The crude product was purified by flash chromatography (silica, eluting with 10% EtOAc in cyclohexane) to give 19 (3.88 g, 92%). 1  tert-Butyl-4-(6-bromo-1-(methylsulfonyl)-1H-pyrrolo[3,2-c]pyridin-2yl)-1H-pyrazole-1-carboxylate (1.48 g, 3.35 mmol) was stirred in THF (20 mL), and DBU (0.51 mL, 3.4 mmol) was added. The reaction was warmed at 40°C for 1 h. The reaction was cooled, and THF was evaporated. The residue was dissolved in EtOAc (50 mL), washed with water and brine, dried, and evaporated. 1 H NMR of the residue revealed incomplete conversion. The material was redissolved in THF (20 mL), and DBU (0.3 mL) was added. The reaction was heated at 40°C for 1.5 h. MeOH (1 mL) was added, and heating was continued for 0.5 h. The solution was evaporated. and EtOAc was added. The solution was washed with water. The organic solution was washed again with water and brine, dried, and evaporated. 1 H NMR of the residue revealed both demesylated and completely deprotected products. To this material were added EtOAc and di-tert-butyl dicarbonate (1.11 g, 5.1 mmol) followed by Et 3 N (0.72 mL, 5.1 mmol) and a crystal of DMAP. The reaction was stirred at rt for 1 h, more ditert-butyl dicarbonate (414 mg, 1.9 mmol) was added, and stirring was continued for a further 2 h. The solution was evaporated, and the residue was kept at ambient temperature overnight. It was adsorbed from CH 2 Cl 2 onto flash silica, packed onto a flash column made in 20% EtOAc in cyclohexane, and eluted with this solvent and then with 40% EtOAc in cyclohexane to give 17 (1.2 g, 77%). 1
MPS1 Kinase Assay. The enzyme reaction (10 μL total volume) was carried out in black 384-well low-volume plates containing full-length MPS1 (LifeTechnologies or in-house, in a range from 3 to 12.5 nM to obtain 10% total conversion during the assay), fluorescentlabeled peptide [H236, sequence: 5FAM-DHTGFLTEYVATR-CONH 2 , Pepceuticals Ltd., Enderby, UK] (5 μM), ATP (10 μM), either 1% (v/v) DMSO or the test compound (in the range from 0.25 nM to 100 μM in 1% (v/v) DMSO), and assay buffer (50 mM HEPES, pH 7.0, 0.02% (w/v) NaN 3 , 0.01% (w/v) BSA, 0.1 mM orthovanadate, 10 μM MgCl 2 , 1 μM DTT, and Roche protease inhibitor). The reaction was carried out for either 60 or 90 min at room temperature and stopped by the addition of buffer (10 μL) containing 20 mM EDTA and 0.05% (v/v) Brij-35, in 0.1 M HEPESbuffered saline (Free acid, Sigma, UK). The plate was read on a Caliper EZ reader II (PerkinElmer Life Sciences, Waltham, MA, USA). The reader provides a Software package ('Reviewer') that converts the peak heights into percent conversion by measuring both the product and substrate peaks, and it also allows selection of control wells that represent 0 and 100% inhibition, respectively. The percent inhibition of the compounds was calculated relative to the mean values of selected control wells. IC 50 values were determined by testing the compounds at a range of concentrations from 0.25 nM to 100 μM. The percent inhibition at each concentration was then fitted to a fourparameter logistic fit using the Studies package (Dotmatics, Bishops Stortford, UK): y = (a + ((b − a)/(1 + ((c/x d )))), where a = asym min, b = asym max, c = IC 50 , and d = Hill coefficient. The reader provides a Software package ('Reviewer') that converts the peak heights into percent conversion by measuring both the product and substrate peaks. The percentage inhibition was calculated relative to blank wells (containing no enzyme and 1% (v/v) DMSO). IC 50 values were determined by testing the compounds at a range of concentrations from 0.25 nM to 100 μM. The percent inhibition at each concentration was then fitted to a four-parameter logistic fit using the Studies package (Dotmatics, Bishops Stortford, UK): y = (a + ((b − a)/(1 + ((c/x d )))), where a = asym min, b = asym max, c = IC 50 , and d = Hill coefficient.
GSK3β Kinase Assay. All GSK3β percentage inhibitions at 1 μM were performed in duplicates by Invitrogen in a Z'LYTE activity assay using their SelectScreen biochemical kinase profiling service.
Cell Viability Assay. Cell proliferation assays were carried out by colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Sigma). Briefly, cells were plated in 96-well plates at 1500 cells per well in 100 μL of culture medium in triplicate. On the next day, 2-fold dilutions of the compounds to be tested were made in culture medium so that when diluted 5× the final concentration in the wells ranged from 0 to 20 μM. Twenty-five microliters of compounds dilutions in the medium was added to 100 μL of cells and incubated at 37°C and 5% CO 2 for 3 more days (72 h). Cells were then incubated with 40 μL of 5 mg/mL solution of MTT reagent at 37°C for 3 h. Media was carefully removed, and crystals were dissolved in 100 μL of DMSO. The absorbance was measured at 570 nm with the Wallac VICTOR2 1420 multilabel counter (PerkinElmer), and analysis was performed to calculate the GI 50 using GraphPad PRISM.
MSD Assay for MPS1 Autophosphorylation. Cellular IC 50 values of MPS1 autophosphorylation inhibition were measured by an in-house electrochemiluminescence (Meso Scale Discovery, MSD) assay that measured phosphorylation of ectopic MPS1 at the Thr33 and Ser37 sites. 36,37 On day 1, 3 × 10 4 HCT116 cells per well in a 96-well plate were reverse-transfected with 100 ng of wild-type Myc-MPS1 using Lipofectamine LTX (Invitrogen). On the next day, cells were treated with 50 ng/mL of nocodazole. On the following day, cells were treated with 2-fold dilutions of test compounds ranging from 0 to 10 μM for 2 h in the presence of 10 μM MG132. After treatment, cells were washed with PBS and lysed with 60 μL per well of complete lysis buffer (50 mM NaCl, 20 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1% (v/v) Triton X100, 10 mM NaF, protease inhibitor tablet, and phosphatase inhibitor cocktails) on ice for 30 min with shaking. Cell lysates were mixed thoroughly by pipetting up and down, and 25 μL of lysate was loaded onto MSD plates that were precoated with total MPS1 antibody (mouse monoclonal, Invitrogen, cat. no. 35-9100) and blocked with 3% (w/v) BSA. After a 1 h incubation at room temperature on a shaker, plates were washed three times with MSD wash buffer, and 25 μL of pThr33/pSer37 antibody (Invitrogen, cat. no. 44-1325G) diluted in 1% (w/v) BSA was added followed by incubation for a further 1 h at room temperature. Plates were washed again three times with MSD wash buffer and incubated with 25 μL of anti-rabbit sulfo-TAG antibody (Meso Scale Discovery, cat. no. R32AB) diluted in 1% (w/v) BSA) for 1 h. After the final incubation, plates were washed three times with MSD wash buffer and read in the presence of 1× MSD read buffer. IC 50 values were determined using GraphPad PRISM.
Crystal Structure Determination of MPS1 with Ligands. The kinase domain (residues 519−808) of MPS1 was produced in E. coli and purified as described previously. 38 Purified MPS1 was crystallized in the apo form at 18°C using the sitting-drop vapor-diffusion method. The crystallization drops were composed of 2 μL of protein X-ray data were collected at Diamond Light Source, Oxfordshire, UK, at beamlines I04 and I04-1. Crystals belonged to the space group I222 and diffracted to a resolution between 2.36 and 2.80 Å. Data were integrated with MOSFLM 39,40 or XDS 41 and scaled and merged with AIMLESS. 42 The structures were solved by molecular replacement using PHASER, 40,43 and a publicly available MPS1 structure (PDB code 4BI1) 38 with ligand and water molecules removed was used as the molecular replacement model. The protein−ligand structures were manually rebuilt in COOT 44 and refined with BUSTER 45 in iterative cycles. Ligand restraints were generated with grade 46 and Mogul. 47 The quality of the structures was assessed with MOLPROBITY. 48 The data collection and refinement statistics are presented in Supporting Information Table S2.
Mouse Liver Microsomal Stability. Compounds (10 μM) were incubated with male CD1 mouse liver microsomes (1 mg/mL) protein in the presence of 1 mM NADPH, 2.5 mM UDP-glucuronic acid (UDPGA), and 3 mM MgCl 2 in 10 mM PBS at 37°C. Incubations were conducted for 0 and 30 min. Control incubations were generated by the omission of NADPH and UDPGA from the incubation reaction. The percentage of compound remaining was determined after analysis by LC/MS.
Human Liver Microsomal Stability. Compounds (10 μM) were incubated with mixed-gender pooled human liver microsomes (1 mg/mL) protein in the presence of 1 mM NADPH, 2.5 mM UDPGA, and 3 mM MgCl 2 in 10 mM PBS at 37°C. Incubations were conducted for 0 and 30 min. Control incubations were generated by the omission of NADPH and UDPGA from the incubation reaction. The percentage of compound remaining was determined after analysis by LC/MS.
In Vivo Mouse PK. All procedures involving animals were carried out within The ICR's Animal Ethics Committee and national guidelines. 49 Mice (female Balb/C) were dosed po or iv (5 or 10 mg/kg) in 10% (v/v) DMSO and 5% (v/v) Tween 20 in saline. After administration, mice were culled at 5, 15, and 30 min and 1, 2, 4, 6, and 24 h. Blood was removed by cardiac puncture and centrifuged to obtain plasma samples. Plasma samples (100 μL) were added to the analytical internal standard (olomoucine; IS) followed by protein precipitation with 300 μL of methanol. Following centrifugation (1200g, 30 min, 4°C), the resulting supernatants were analyzed for compound levels by LC/MS. For blood pharmacokinetics, 20 μL was spotted on a Whatman B card and allowed to dry for at least 12 h at room temperature and 6 mm diameter disks were punched and extracted with 200 μL of methanol containing 500 nM olomoucine used as internal standard. Sample extracts were analyzed by LC/MSMS against calibration curves (six levels) and six quality controls (three levels in duplicate). Separation was carried out by an Acquity UPLC binary system (Waters) on a reverse-phase Kinetex C18 (Phenomenex 50 × 2.1 mm, 1.7 μm particles) analytical column. Elution was achieved with a 4.5 min gradient of 0.1% formic acid/ methanol (95% formic acid to 0%) following a 0.5 min isocratic period. Detection was performed in positive ion mode ESI multiplereaction monitoring (MRM) on a QTRAP 4000 (AB-SCIEX).
In Vivo Mouse PK/PD Study. Human HCT116 colon carcinoma cells (3 × 10 6 ) were sc injected bilaterally in the flanks of female CrTac:NCr-Fox1(nu) athymic mice. Once tumors reached a mean diameter of ∼8 mm (day 15), mice were dosed twice at a 12 h intervals with 50, 75, or 100 mg/kg of compound 65 in 10% (v/v) DMSO and 5% (v/v) Tween 20 in saline. Mice were culled (n = 3 per group) at 2, 10, and 72 h after the second dose. Tumors were snap-frozen and stored at −80°C until analysis. Tumor samples were homogenized in PBS (3 vol/tumor weight).