Identification of (R)-N-((4-Methoxy-6-methyl-2-oxo-1,2-dihydropyridin-3-yl)methyl)-2-methyl-1-(1-(1-(2,2,2-trifluoroethyl)piperidin-4-yl)ethyl)-1H-indole-3-carboxamide (CPI-1205), a Potent and Selective Inhibitor of Histone Methyltransferase EZH2, Suitable for Phase I Clinical Trials for B-Cell Lymphomas

Polycomb repressive complex 2 (PRC2) has been shown to play a major role in transcriptional silencing in part by installing methylation marks on lysine 27 of histone 3. Dysregulation of PRC2 function correlates with certain malignancies and poor prognosis. EZH2 is the catalytic engine of the PRC2 complex and thus represents a key candidate oncology target for pharmacological intervention. Here we report the optimization of our indole-based EZH2 inhibitor series that led to the identification of CPI-1205, a highly potent (biochemical IC50 = 0.002 μM, cellular EC50 = 0.032 μM) and selective inhibitor of EZH2. This compound demonstrates robust antitumor effects in a Karpas-422 xenograft model when dosed at 160 mg/kg BID and is currently in Phase I clinical trials. Additionally, we disclose the co-crystal structure of our inhibitor series bound to the human PRC2 complex.


■ INTRODUCTION
It is well-established that trimethylation of lysine 27 on histone 3 (H3K27) contributes to the modification of chromatin structure, which serves to repress transcription. 1−3 The addition of trimethyl "marks" on H3K27 is generally catalyzed by the multimeric protein complex polycomb repressive complex 2 (PRC2), through its enzymatic subunit enhancer of zeste homologue 2 (EZH2). EZH2 catalyzes the transfer of a methyl group from the cofactor S-adenosyl-L-methionine (SAM) to the ε-NH 2 group of H3K27 culminating in trimethylation of H3K27 (H3K27me3) and subsequent silencing of targeted genes.
Dysregulation of mechanisms that alter chromatin structure has been implicated in a variety of disease processes, particularly oncogenesis. 2 EZH2 is frequently overexpressed in a broad spectrum of solid and hematological cancers such as prostate, breast, kidney, lung, myeloma, and lymphoma. 2,4 Elevated EZH2 transcript and protein levels in these cancers usually correlate with greater levels of H3K27me3, advanced stages of disease, and poor prognosis. 5,6 Additionally, somatic recurrent mutations within the catalytic domain of EZH2 (the suppressor of variegation, enhancer of zeste, trithorax (SET) domain) have been identified in diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, and melanoma. 7 These mutations alter the substrate specificity of EZH2 culminating in an increase in global levels of H3K27me3. 4,8−10 Consequently, the increase in levels of H3K27me3, either by overexpression of EZH2 or its altered function through mutations, in cancer tissues may reinforce the silencing of target genes that promote differentiation and restrain proliferation. 11 Alternatively, EZH2 may serve to silence genes not targeted in normal cells to afford growth and survival advantage in the malignant setting. Collectively, these observations offer a compelling argument for the inhibition of EZH2 as a potential therapeutic approach for the treatment of cancer.
Medicinal Chemistry and Structure−Activity Relationships. We recently disclosed 1, a potent indole based EZH2 inhibitor that showed robust antitumor activity and pharmacodynamic (PD) target engagement in a KARPAS-422 lymphoma xenograft model in mice. 21 This compound, however, suffered from limited oral bioavailability (0.09% F and 0.12% F observed in rats and dogs, respectively).) As part of our ongoing drug discovery and development efforts, we continued to optimize the indole-based scaffold toward clinical candidate selection. Herein we report the optimization of the indole based EZH2 inhibitor series that led to the identification of 13, a potent and selective inhibitor of EZH2 currently under evaluation in Phase I clinical trials. In addition, we report the co-crystal structure of a similar pyridone containing inhibitor (10) bound to human PRC2. This structure has provided a context for the molecular nature of the interaction between our chemical series and its target.
In an attempt to improve upon the physical properties of 1, we began the investigation of our structure−activity relationship (SAR) with different N-substituents on the piperidine ring. We rationalized that the pyridone-indole core was previously optimized with respect to biochemical potencies (against both wild-type and mutant EZH2) as evidenced by N−H piperidine (2) and N−Me piperidine (3) analogues (Table 1). Unfortunately both 2 and 3 suffered from considerable loss in cellular potency in the HeLa H3K27me3 mechanism of action (MOA) assay. Derivatization of the N−H piperidine to a variety of amides, ureas, carbamates, and sulfonamides yielded biochemically potent analogues. However, these analogues displayed less than ideal cellular potencies when examined in
In addition to the disparate trends in potencies (i.e., poor translation into cellular assays), N-acylated derivatives (e.g., amides, ureas, carbamates, and sulfonamides) generally suffered from high microsomal clearance or rapid clearance in vivo ( Figure 2). In contrast, a variety of basic amines derived from piperidine 2 showed acceptable ADME properties (low Cl int and low CYP inhibition). As such, we subsequently sought to improve the physiochemical properties of piperdine 2 to attain the desired in vitro and in vivo potencies.
We found that attenuation of the basicity and hence perturbation of the piperidine pK a had profound effects on the cellular potency, selectivity, toxicity, bioavailability, and PK properties. 24 For example, oxetane 7 (ChemDraw calculated pK a was 7.6) demonstrated a 10-fold improvement in the

Journal of Medicinal Chemistry
Article cellular potency when compared to piperidine 2 (ChemDraw calculated pK a was 9.7). 25 Other electron-withdrawing substituents, such as acetate 8 also afforded improvement in cellular potencies.
We synthesized n-trifluoropropyl piperidine analogue 9 (calculated pK a ∼ 8.2) and observed an over 3-fold increase in potency (EC 50 of 0.29 μM) relative to piperidine 2 when tested in the cellular MOA assay (Table 1). While introduction of the n-trifluoropropyl motif was a step in the right direction, we believed that cellular potency could be further enhanced. To examine the effect of keeping the three-carbon linker fixed, and moving the fluorine atoms to the 2-position of the alkyl chain, we subsequently synthesized the 2,2-difluoropropyl containing analogue (10). The pK a was calculated to decrease by half a log unit relative to analogue 9 (Chemdraw calculated pK a 7.6 versus 8.2, respectively), but gratifyingly, even this modest decrease in basicity resulted in a 10-fold gain in cellular potency (EC 50 of 0.020 μM) over the linear n-trifluoropropyl analogue. The addition of fluorine atoms in the β-carbon (with respect to the piperidine nitrogen) appeared optimal for attenuation of basicity of the piperidine and increase in cellular potency. As such, a series of analogues that embedded the β,β-difluoroethyl
Co-crystal of Ligand Bound to PRC2 Complex. In order to understand the mechanism of action within this lead series, we were able to exploit the crystallization system developed for our recent structure of the human PRC2 catalytic complex. 26 Co-crystals were obtained with the 2,2-difluoropropyl analogue (10) that diffracted to 3.5 Å and the structure determined by molecular replacement (PDB code: 5LS6). Electron density corresponding to compound 10 was identified in a pocket at the interface of the EZH2 SET domain, the SAL region of the EZH2 N-terminus and EED ( Figure 3A). The electron density was improved by 4-fold averaging, so that even at 3.5 Å it is remarkably well-defined and the position and orientation of the inhibitor could be unambiguously determined ( Figure 3B).
Key residues that define the inhibitor pocket include EZH2 SET domain Tyr661, Phe665, Tyr658, and Phe686, the EZH2 SAL region Tyr111 and Met110, and EED residues His213 and Asp237. 27 Although the inhibitor binding site partially overlaps with the pocket for SAH (superimposed from PDB ID: 5HYN), it then extends in the opposite direction and is therefore distinct from both the substrate and cofactor binding sites ( Figure 3B and 3C). The partial overlap between the 2,2difluoropropyl piperidine 10 and the SAH carboxylic acid is consistent with a SAM competitive mechanism of inhibition.
Overall, 10 adopts a zigzag like shape, which is defined by both the constraints of the pocket, as well as the preorganized conformation induced by the chiral methyl proximal to the indole and the torsional angle defined by the C3-indole amide connecting to the pyridone. The indole and piperidine are tightly constrained by the narrow hydrophobic channel and by a putative hydrogen bond from the EED Asp237 side-chain to one of the fluorine atoms. Beyond the piperidine, the pocket widens considerably, which is consistent with the diverse array of substituents tolerated in this region (Table 1). At the opposite end of the molecule the pyridone is surrounded by the side chains of Phe665, Phe686, and Trp624 and has the potential to form two hydrogen bonds with the protein backbone of Trp624 ( Figure 3D). Recently, a co-crystal structure of a human/chameleon hybrid PRC2 construct was reported that also contained a pyridone-based EZH2 inhibitor (PDB codes: 5IJ7 and 5IJ8). 28 Both pyridone orientation and binding is consistent between the two structures.
A major difference between the inhibitor complex structure and that obtained with peptide/SAH occurs at the active site of the EZH2 SET domain ( Figure 4). In the absence of substrate peptide, the C-terminus of the SET domain occupies the histone binding groove with the side chains of Tyr728 and Phe667 located to the target lysine channel. In the inhibitor complex beyond residue Ser729, the C-terminus is disordered. Similarly, in the structure of the isolated apo EZH2, there is a pronounced rearrangement of this region. 29 The alternate conformation of the EZH2 C-terminus observed in these

Journal of Medicinal Chemistry
Article structures are indicative of its inherent flexibility. It is not clear to what extent the conformation observed in the inhibitor complex structure is due to the binding of inhibitor or reflects the absence of cofactor and/or histone substrate.
In Vitro ADME and In Vivo Pharmacokinetics of Fluorinated Analogues. Having established sufficient activity in the MOA assay, we subsequently profiled 9, 10, 12, and 13 in an ADME panel ( Table 2). Relative to 1, these fluorinated analogues generally displayed higher plasma protein binding and in vitro clearance, which is presumably a function of their increased lipophilicity (cLogP). Interestingly, there were significant interspecies differences observed with regards to the in vitro microsomal clearances and PPB. The highest in vitro microsomal clearance (>100 μL/min/mg protein) and plasma protein binding (>97% bound) were uniformly observed in mice for 10, 12, and 13. Additionally, the in vitro clearance measured from rat-derived microsomes closely mirrored those derived from humans for these analogues.

Article
Based on their combination of potency and in vitro profile 12 and 13 were subsequently evaluated for their in vivo metabolic profile and systemic exposure in mice in advance of planned pharmacodynamic (PD) and efficacy studies (Table 3). Both 12 and 13 showed excellent oral bioavailability demonstrating significant improvement over 1. Surprisingly, 12 displayed a poor overall PK profile when compared to 13. When dosed in mice at 1 mg/kg intravenous (iv) and 100 mg/kg per os (po), 12 displayed high clearance of 4.45 L/h/kg (82% liver blood flow), low volume of distribution (1.10 L/kg), short half-life (0.40 h), and excellent bioavailability (∼92% F). In contrast, 13 exhibited moderate clearance of 2.16 L/h/kg (40% liver blood flow), a half-life of ∼1.6 h, similar volume of distribution (1.4 L/kg), and excellent bioavailability (100% F). As part of the comparison, we also examined total exposure (AUC) and unbound exposure (AUC unbound ) because we were interested in maintaining free levels of compound significantly above the measured HeLa EC 50 . Both 12 and 13 achieved unbound exposures well above their respective cellular potencies; however, only the unbound exposure for 13 remained well above the cellular EC 50 up to 4 h ( Figure 5). Overall, the in vivo metabolic profile of 13 provided a compelling argument for further evaluation in a mouse xenograft model.
Synthesis of Indole Piperidine Analogues. Many of the analogues profiled in Table 1 were prepared by similar synthetic routes, with appropriately substituted building blocks. We identified indole 14 as a key scaffold for the construction of Nsubstituted piperidines. Synthetically, the alkylation of 2methyl-1H-indole-3-carboxylate with branched alkyl electrophiles under a variety of basic conditions failed to deliver desired indole 14. The lackluster behavior of 2-methyl-1Hindole-3-carboxylate toward various alkylation conditions was presumably a consequence of deactivation of the indole N-atom by the 3-carboxylate moiety and additional steric constraints

Journal of Medicinal Chemistry
Article imposed by the 2-methyl substitution. As such, we looked to devise an alternative strategy toward a more convergent, robust, and scalable synthesis of indole 14 (Scheme 1). We were drawn to the possibility of utilizing halo-aryl enamines as latent precursors for a palladium-mediated intramolecular C−N bond construction of indoles. 30 We rationalized that generation of enamines from β-keto esters would allow for the introduction of a variety of amines, and this modular approach would serve as a diversity generating element for our drug discovery efforts. To that end, the construction of indole 14, began with the condensation reaction between β-keto ester 30 15 and chiral amine 31,32 16 under mildly acidic conditions to deliver enamine 17, predominantly as the Z-isomer, in 76% yield. Paramount to the success of this intramolecular C−N arylation was a systemic investigation of palladium catalysts and reactions conditions. 30 We were gratified to find that treatment of chiral enamine 17 with Buchwald's RuPhos precatalyst system in the presence of sodium methoxide cleanly induced intramolecular C−N bond arylation to yield chiral indole piperidine 18 while maintaining the stereochemical integrity of the chiral center. Subsequent deprotection of N-Boc piperidine 18 with anhydrous hydrochloric acid produced piperidine 14 in 81% yield over two steps (Scheme 2).
The In Vivo Efficacy Studies. Having sufficient quantities in hand, we evaluated the performance of inhibitor 13 for tumor pharmacodynamic effects and antitumor efficacy in a KARPAS-422 B-cell lymphoma xenograft model in mice. KARPAS-422 xenograft cells harbor a recurrent, monoallelic mutation (Y641N) within the EZH2 catalytic domain. 17 These mutations alter the EZH2 substrate specificity and thus represent a context of constitutive EZH2 pathway activation. Informed by previous in vivo studies, 13 was dosed at 160 mg/kg orally twice daily (po BID) for 25 days in tumor bearing female CB-17 SCID mice ( Figure 6). Upon treatment of tumor-bearing CB-17 SCID mice with 13, tumor regression was observed within 2 weeks. By the end of day 25, significant tumor growth inhibition was recorded (>97% TGI relative to vehicle, see Figure 6A). Inhibitor 13 was well-tolerated for repeat dosing as demonstrated by the absence of significant body weight loss ( Figure 6B). To allow for analysis of tumor tissues at the end of the study, treatment was suspended at day 25. Tumor samples were harvested 1 h post last dose, and their analysis revealed considerable reduction of H3K27me3 (47% reduction in H3K27me3/global H3 ratio relative to vehicle control, see Figure 6C). Analysis of plasma and tumor PK at 1 h post last dose on day 25 shows sufficient plasma and tumor tissue concentrations of 13, 11 388 ng/mL [22 μM] versus 5286 ng/g [10 μM], respectively ( Figure 6D).
Selectivity and Additional Profiling. On the basis of the successful KARPAS-422 efficacy, we selected 13 for further in vivo and in vitro profiling. Additional PK data on 13 in rats and

Journal of Medicinal Chemistry
Article dogs was collected (Table 4). Analogue 13 shows relatively high clearance in both rats and dogs (3.19 L/h/kg and 1.41 L/ h/kg, respectively) but demonstrates good oral bioavailability in both species (44.6% F in rats and 46.2% F in dogs). As part of the characterization of 13, its activity against a number of other targets were evaluated. Inhibitor 13 showed a clean selectivity profile when tested against 30 other histone or DNA methyltransferases. 34 Additionally, compound 13 demonstrated modest selectivity (EZH1 IC 50 of 52 ± 11 nM) when tested against enhancer of zeste homologue 1 (EZH1), a methyltransferase highly related to EZH2. Examination of the sequence similarity between EZH1 and EZH2 in the context of the co-crystal structure reveals that the residue positioned in close proximity to the inhibitor, Cys663, is one of only four residues within the EZH2 SET domain that are not conserved in EZH1. The equivalent EZH1 residue is Ser664. Cys663 makes van der Waals contact with the bound inhibitor and is predicted to be a key selectivity determinant with respect to EZH1 ( Figure 3B). In order to test this hypothesis, we measured the potency of 10 and 13 in the context of an EZH2 version with a single amino acid substitution from cysteine to serine at position 663 (C663S) incorporated into reconstituted PRC2. As expected, the potency of the inhibitors decreased in the mutated EZH2 C663S, which is consistent with the reduction in potency observed in EZH1 enzymatic assays. The fact that this single mutation does not fully recapitulate the EZH2/EZH1 potency differences may reflect a subtle effect of other more distant residue differences between the two proteins.
Further in vitro profiling of 13 showed no time-dependent inhibition (TDI) of the cytochrome P450 enzymes 1A2, 2C9, 2C8, 2D6, and 3A4. Additionally, 13 was evaluated for secondary pharmacology against a panel of fifty-four physiologically relevant receptors, transporters, and ion channels at 10 μM. As such, 13 did not inhibit any target more than 50%. Finally, 13 was examined for any potential cause of cardiac arrhythmias associated with delayed ventricular repolarization (QT interval prolongation). When tested within an in vitro hERG binding assay at concentration ranges of 45 nM to 100 μM, 13 showed an IC 50 of 21.3 μM. This concentration is well above the free concentration predicted to be achieved in patients.
Toxicology Study of 13. To establish a safety window for repeat dosing of 13 in human clinical trials, a preclinical safety study was conducted in two separate species. Compound 13 was orally administered in a GLP compliant toxicity study for 4 weeks to both Sprague−Dawley rats and beagle dogs followed by a 4-week recovery period. The compound was administered by oral gavage at single daily doses (QD) of 100, 300, and 600 mg/kg to rats for 28 days and at twice daily doses (BID) of 50, 150, and 500 mg/kg for 28 days to dogs. In general, 13 was well-tolerated in the 28-day GLP toxicology studies, and any findings were reversible over the recovery period. The GLP toxicology studies of 13 demonstrated an acceptable safety profile and enabled selection of clinical doses.

■ CONCLUSIONS
In summary, we have reported the discovery of 13, a highly potent and selective small molecule inhibitor of EZH2, suitable for introduction into Phase I clinical trials for the treatment of B-cell lymphomas (NCT02395601). Using 1, our previously disclosed EZH2 chemical probe as a starting point, we embarked on a focused campaign to optimize the cellular potency and physiochemical properties of our pyridone-indole scaffold. Removal of the sulfonamide functionality (embedded within 1) yielded analogues that retained biochemical potency but lacked sufficient cellular activity in the HeLa H3K27me3 assay. Observing that attenuation of the pK a of the piperidine N-atom has a profound effect on the correlation between biochemical and cellular potency, we ultimately discovered a series of fluorinated analogues with improved cellular activity and good oral bioavailability.
During the course of our investigations, we successfully determined the co-crystal structure of compound 10 bound to human PRC2. The present crystal structure illustrates several distinct binding features. The overall density reveals a zigzaglike shape of the inhibitor, likely a consequence of conformational preorganization induced by the chiral methyl and C3amide substituents that conforms to the enzyme's pocket. Additionally, the pyridone motif forms two hydrogen bonds with the protein backbone of Trp624 and is constrained in an aromatic environment created by Phe665, Phe686, and Trp624. These crucial interactions with the pyridone motif explain the importance of this functional group for its high affinity binding, its prevalence in a vast majority of other reported EZH2 inhibitors, and the difficulty in finding suitable pyridone replacements.
After triaging these analogues through in vitro and in vivo experiments, N-trifluoroethylpiperidine analogue 13 was evaluated in a KARPAS-422 lymphoma xenograft model. Gratifyingly, 13 was well-tolerated, proved efficacious, and achieved >97% TGI after treatment for 25 days. After further in vitro/in vivo characterization and safety studies, compound 13 was advanced into human clinical trials. The clinical impact of 13 on EZH2 inhibition in oncology will be described in due course.

Journal of Medicinal Chemistry
Article ■ EXPERIMENTAL METHODS All commercial reagents and anhydrous solvents were purchased and used without purification, unless specified. Column chromatography was performed using a Biotage chromatography system on Biotage or Silicycle normal phase silica gel columns. NMR spectra were recorded on a Varian Unity Inova (400 MHz) or an Oxford (Varian, 300 MHz) instrument. LC-MS were recorded on an Agilent 1200 series LC connected to an Agilent 6120 MS or Agilent 1100 series LC connected to an Agilent 1956B MS or a Shimadzu LC-MS-2020 system. Preparatory HPLC was performed using a Gilson GX-281 or P230 Gradient System (Elite). Chiral preparatory HPLC were performed using Elite P230 Preparative Gradient System, Thar Prep-80 and Thar SFC X-5 systems. The purity of the final products was >95% as determined by HPLC/MS and 1 H NMR.  (17). 30 A 2-L threeneck round-bottom flask (fitted with a magnetic stir bar, thermocouple, reflux condenser, and rubber septa) was charged with methyl 2-(2-bromophenyl)-3-oxobutanoate (116.25 g, 428.80 mmol), EtOH (850 mL, ∼7 mL/g), (R)-tert-butyl 4-(1-aminoethyl)piperidine-1-carboxylate (121.00 g, 529.93 mmol), and AcOH (29.50 mL, 515.31 mmol). The reaction vessel was heated over a heating mantle to 80°C −85°C for 18 h. After 18 h, the reaction mixture was cooled to ambient temperature, and the tert-butanol was removed in vacuo. The resultant oil was diluted with EtOAc and subsequently poured over saturated aqueous NaHCO 3 while vigorously stirring. Once the evolution of CO 2 (g) ceased, the biphasic solution was transferred to a separatory funnel, and the phases were separated. The aqueous phase was extracted with additional EtOAc (2×). The combined organic phase was dried over MgSO 4 , filtered, and concentrated in vacuo to afford crude product. This material was preabsorbed onto silica gel (∼100 g) and filtered through a fritted funnel with 20% EtOAc to 80% hexanes afford (R)-tert-butyl 4-(1-((3-(2-bromophenyl)-4-methoxy-4oxobut-2-en-2-yl)amino)ethyl)piperidine-1-carboxylate (155.95 g, 76% yield). The material was used without further purification. LC-MS m/z 481 [M + H] + .

Journal of Medicinal Chemistry
Article flask charged with a magnetic stir bar was added 2,2-difluoropropanoic acid (7.5 g, 68.14 mmol) and DCM (250 mL). The mixture was cooled to 0°C, and oxalyl dichloride (5.48 mL, 8.22 g, 64.73 mmol) was added over 1 min. To this solution was added DMF (500 μL, 6.43 mmol), and the solution was warmed to room temperature with stirring until bubbling ceased (about 1 h). The solution is used as is in the subsequent step without further purification.
Step 3. A 1 L 3-necked flask was equipped with magnetic stirrer and was fitted with a reflux condenser and an oil-filled bubbler outlet. The vessel was purged and placed under an atmosphere of nitrogen and methyl (R)-1-(1-(1-(2,2-difluoropropanoyl)piperidin-4-yl)ethyl)-2methyl-1H-indole-3-carboxylate (9.1 g, 23.19 mmol) was dissolved in THF (150 mL) and cannulated into the reaction flask. The reaction was cooled to 0°C in an ice bath and borane (1.0 M THF solution, 55 mL, 55 mmol) was added over 10 min via syringe. When intense bubbling subsided, the reaction mixture was heated to reflux for 2 h. The reaction was then cooled to 0°C followed by the careful addition of MeOH (80 mL) (caution: vigorous H 2 gas evolution observed). The reaction was then stirred at 0°C for 5 min, and then allowed to warm to room temperature. The mixture was then heated to 65°C for 45 min, cooled to room temperature, and was transferred to a 1 L round-bottom flask. The volatiles were removed under reduced pressure. The material was purified by column chromatography (120 g silica column, 10% to 40% EtOAc in hexanes) to afford methyl (R)-1-(1-(1-(2,2-difluoropropyl)piperidin-4-yl)ethyl)-2-methyl-1H-indole-3carboxylate (7.96 g, 90% yield).