Design, Synthesis, and Pharmacological Evaluation of Second Generation EZH2 Inhibitors with Long Residence Time

Abstract

Histone methyltransferase EZH2, which is the catalytic subunit of the PRC2 complex, catalyzes the methylation of histone H3K27—a transcriptionally repressive post-translational modification (PTM). EZH2 is commonly mutated in hematologic malignancies and frequently overexpressed in solid tumors, where its expression level often correlates with poor prognosis. First generation EZH2 inhibitors are beginning to show clinical benefit, and we believe that a second generation EZH2 inhibitor could further build upon this foundation to fully realize the therapeutic potential of EZH2 inhibition. During our medicinal chemistry campaign, we identified 4-thiomethyl pyridone as a key modification that led to significantly increased potency and prolonged residence time. Leveraging this finding, we optimized a series of EZH2 inhibitors, with enhanced antitumor activity and improved physiochemical properties, which have the potential to expand the clinical use of EZH2 inhibition.

Polycomb repressive complex 2 (PRC2) is responsible for the trimethylation of lysine 27 on histone 3 (H3K27), a modification which serves to repress transcription.¹ Enhancer of zeste homologue 2 (EZH2), a subunit of PRC2, catalyzes the transfer of a methyl group from the cofactor S-adenosyl-l-methionine (SAM) to the ε-NH₂ group of H3K27 culminating in trimethylation of H3K27 (H3K27Me3)² and subsequent silencing of targeted genes.³ The dysregulation of target genes by the “trimethylation marks” has been implicated in a variety of disease processes, particularly oncogenesis.^4,5

EZH2 has been reported to be mutated or overexpressed in a broad spectrum of oncology indications such as prostate cancer,⁶ lung cancer,⁷ myeloma, and lymphoma.⁸ Furthermore, activating mutations in EZH2 have been shown to be a key driver of diffuse large B-cell lymphoma and follicular lymphoma,⁹ while resistance to hormone therapy has been shown to arise from upregulation of EZH2 expression in prostate cancer.⁶ The increased levels of H3K27Me3 that result from EZH2 upregulation reinforce the silencing of target genes, which promote differentiation and restrain proliferation in normal cells.¹⁰ Clinical EZH2 inhibitors (Figure 1) are currently being evaluated as monotherapy and in combination in both hematological malignancies and solid tumors, including specific genomic contexts such as mutations in EZH2 and components of SWItch/Sucrose nonfermentable (SWI/SNF)—a nucleosome remodeling complex (NCT03213665). Initial results in the clinic have further validated the role of EZH2 catalytic activity in the context of follicular lymphoma, where patients with activating mutations respond at a higher rate to EZH2 inhibitors than those with wild-type EZH2.¹¹

Figure 1.

Left: Clinical inhibitors of EZH2 and S-adenosyl-methionine, the endogenous cofactor of EZH2. Right: Overlay of SAH (S-adenosylhomocysteine) in cyan bound to EZH2 (gray) and a CPI-1205 analog (green), with the structures of the ligands depicted above.

EZH2 inhibitors in the clinic have been reported to exhibit favorable safety profiles and display antitumor activity, particularly in hypersensitive contexts, such as lymphoma, and in combination.¹² Initial dose escalation studies have resulted in recommended phase 2 doses of >800 mg BID.¹³ Even though clinical agents have thus far been found to be well-tolerated but significant reduction in exposure is observed after multiple dosing in humans,^12,14 consistent with induction of CYP3A4 leading to faster metabolism of the drug.^11,15 The high dose and CYP induction present potential limitations when including first generation EZH2 inhibitors as part of a combination therapy. Furthermore, our preclinical data showed that prolonged treatment is necessary to trigger a decrease in H3K27Me3, in agreement with the slow turnover kinetics of histone H3K27Me3 modification. To maximize the oncology application potential of EZH2 inhibition in the clinic, we initiated a second generation program that sought to achieve increased and more durable target coverage.

We initiated our medicinal chemistry campaign with an analysis of inhibitor bound EZH2 cocrystal structures, which showed that the pyridone moiety occupies the same pocket of the active site as the SAM cofactor (Figure 1).^16,17 In particular, the amino acid moiety of either SAM or S-adenosylhomocysteine (SAH) and the pyridone motif of the EZH2 inhibitors overlap (highlighted in Figure 1). The pyridone competes with the cofactor for the H-bond donor–acceptor interaction with tryptophan 624. Specifically, the N–H of the pyridone serves as an H-bond donor for the peptide-backbone’s carbonyl oxygen. Accordingly, we surmised that favoring the compound’s tautomer population toward the pyridone over the hydroxypyridine would lead to more potent EZH2 inhibitors. This analysis suggested to us that the electronic nature of the pyridone, electron-rich or -poor, would be highly critical for the inhibitor’s potency for EZH2. Therefore, using our CPI-1205 indole scaffold as a starting point we generated a variety of substituted pyridones, including compounds 1–5 (Figure 2). The biochemical potencies of these inhibitors were found to be in the range of 30 nM to 0.2 nM, and their comparison showed that heteroatom substitution was preferred at the 4-position over the 2-position, with the 4-thiomethyl pyridone exhibiting the lowest IC₅₀. The rank ordering of pyridone biochemical potencies also correlated well with their activity in the cellular H3K27Me3 assay. Inhibitors 2 and 3, which feature a 2-chloro and a 2-methoxy pyridone, respectively, were found to have cellular EC₅₀s of >1 μM, whereas the inhibitors that feature heteroatoms at the 4-position, such as 5, 1, and CPI-1205, were found to exhibit cellular EC₅₀’s of <100 nM. In particular, the 4-thiomethyl pyridone 1 was found to be the most potent in both the biochemical and the cellular H3K27Me3 mark reduction assays (EC₅₀ = 2.5 nM). Due to the overperformance of 1 in the cellular assay, we sought to characterize the residence time of the more active analogs, 4, 5, 1, and CPI-1205. We found that, although their biochemical potencies were in a similar range (180 pM to 800 pM), the thiomethyl pyridone 1 exhibited a distinctly longer residence time. More starkly, 1 exhibits nearly a 10-fold greater residence time than its oxygen analog, CPI-1205 (Figure 1), in agreement with a 10-fold improvement in both the H3K27Me3 EC₅₀ and the GI₅₀ (viability in KARPAS-422 lymphoma cells).¹⁸ Note, we also tested the 2-thiomethyl, 4-methyl substituted pyridone analog and found it to be inactive in the biochemical assay, similar to compound 3.

Figure 2.

TR-FRET and MSD based EZH2 inhibition assay and TR-FRET based residence time characterization.

Having identified a novel pyridone motif that engenders increased residence time and cellular potency, we generated a series of 4-thioalkyl substituted pyridones, 6–11 (Table 1). We found that increasing the substituent size (6–9) led to a decrease in potency. Likewise, oxidation of the thioether to a methylsulfinyl or a methylsulfone led to an even greater loss in potency. Given the large drop in the biochemical potency, we did not characterize the residence time for this series. The 4-thiomethyl substituent, exemplified in 1, was found to be the most potent EZH2 inhibitor. We applied this finding to many other clinical EZH2 scaffolds by replacing their pyridone with the 4-thiomethyl pyridone of 1 and found a similar increase in inhibitor residence time (8–23-fold).¹⁹ We hypothesize that the superiority of the 4-thiomethyl substitution is due to a combination of multiple factors, including the following: (i) pyridone/hydroxypyridine tautomeric ratio; (ii) the larger van der Waal’s radius of sulfur which leads to a larger hydrophobic buried surface area; (iii) the tighter methyl contact with the protein due to larger C–S bond lengths; and possibly (iv) a positive S−π interaction with the electron rich side chains of phenylalanines 665 and 686.²⁰

Table 1. Optimization of 4-Thioether Pyridone.

^aAverage with the standard deviation reported in parentheses, where n ≥ 2.

Being conscious of the previously reported CYP3A4 induction, we sought to characterize the propensity of 1 to induce expression of the cytochrome p450 enzyme CYP3A4. Therefore, we treated the immortalized human hepatpocytes, Fa2N-4 cells with 10 μM of 1, and found nearly a 25-fold increase in the expression of CYP3A4.²¹ With the 4-thiomethyl-pyridone motif in place, we sought to optimize the piperidine side chain to further enhance potency and to limit CYP3A4 upregulation. We found that introduction of beta-oxygenation led to a decrease in CYP3A4 upregulation (13); however, this also led to a decrease in biochemical potency (Table 2). The potency was enhanced by replacing the piperidine with an exocyclic amine, as in 14−17. Further characterization of compound mediated CYP3A4 upregulation showed that 17 only upregulated CYP3A4 2-fold, in comparison to 25-fold with 1. SAR of the piperidine side chain showed a correlation between cLogD and CYP3A4 upregulation, where the compounds with lower cLogD exhibited limited CYP3A4 upregulation (Table 2).²² We determined that a 2-fold increase of CYP3A4 expression, under our assay conditions of 10 μM compound treatment, would be sufficient for progressing the program given that we do not anticipate achieving such high concentrations in vivo. As a result, by elaborating from 1 to 17, we improved biochemical potency, retained cellular activity, enhanced residence time and limited CYP3A4 upregulation.

Table 2. Optimization of Piperidine Side Chain^a.

^aAverage with the standard deviation reported in parentheses, where n ≥ 2.

Given the improvements in potency and limited CYP3A4 upregulation, we sought to characterize 1 and 17 in the KARPAS-422 lymphoma xenograft mouse model (Figure 3). A 160 mg/kg BID-PO dose of Tazemetostat, which has previously been shown to cause regression in this model,²³ was used as a benchmark compound in the study. With the goal of rank-ordering compounds for our medicinal chemistry campaign, we kept the BID dosing regimen constant during our in vivo studies and evaluated the efficacy and kinetics of achieving regression. We dosed inhibitors at 50 mg/kg BID, which we hypothesized would provide a dynamic range of efficacy between compounds. 1 and 17, which were both dosed SC, began to cause regression at days 7 and 9, respectively, a marked improvement over the benchmark. To ensure enough tumor sample could be obtained for downstream characterization, mice in the treatment arms of 1 and 17 were sacrificed after 17 days (no body weight loss observed, see Supporting Information). Meanwhile, Tazmetostat was dosed out to day 24 to ensure the benchmark compound also achieved regression.

Figure 3.

(a) Evaluation of 1 and 17 in KARPAS-422 lymphoma xenograft mouse model. (b) Evaluation of 17 in 22Rv1 prostate cancer xenograft mouse model.

We then characterized the efficacy of 17 in the 22Rv1 prostate cancer xenograft model in mice. 22Rv1 cells have been shown to express high levels of EZH2 when compared to other prostate cancer cell models, and they are dependent on androgen receptor signaling for growth.²⁴ The 22Rv1 cell-line is not sensitive to androgen-degrader therapies such as Eznalutamide because it expresses the AR-V7 isoform of the androgen-receptor, which lacks the ligand-binding domain.²⁵ We found that a 50 mg/kg BID-SC dose of 17 led to ∼55% tumor growth inhibition (TGI).

Although 17 achieved in vivo efficacy in both models, it was not orally bioavailable; therefore, we next sought to improve upon the inhibitor’s physiochemical properties. We hypothesized that the methoxy substituent was not lipophilic enough to mask to heightened basicity of azetidine 17; this potentially led to decreased absorption.¹⁶ We replaced the methyl group of the azetidinol with substituents that would mitigate the basicity of the exocyclic amine, such as fluoroalkyls 18 and 19 (Table 3). We also replaced the methyl group of the azetidinol with a cyclopropane in 20 to increase the lipophilicity and in-turn improve permeability. Through these modifications we found that the methyl to cyclopropyl replacement resulted in a 3-fold improvement in GI₅₀ (17 vs 20). Although these modifications led to increased potency, particularly in the cellular proliferation assays, as seen with 19 and 20, they also resulted in an increase of CYP3A4 upregulation, potentially due to their higher clogD. To combat CYP3A4 upregulation, we retained the cyclopropoxy side chain of 20 and introduced heteroatoms within the central indole-core to increase polarity. In particular, we introduced a nitrogen within the central core to generate aza-indole 21, which retained many of the desirable properties of 17 such as cellular potency and decreased CYP3A4 upregulation (∼5-fold 20 vs 21). Replacing the aza-indole in 21 with a fluoro-indole in 22 led to diminished activity in the cellular assays, including the cell proliferation assay. We also generated the pyrrolopyrimidine 23 in an effort to decrease the clogD; however, this further led to loss in cellular activity. 5-Halo-substituted aza-indoles 24 and 25 were found to be less active in the cellular assays and, in the case of 25, led to a greater CYP3A4 upregulation than 21. Unlike the clear correlation observed with our earlier side chain modifications (Table 2), we did not observe a strict correlation of clogD with CYP3A4 upregulation in the case of compounds 20–25 (Table 3). Nevertheless, during this survey we identified 21, an aza-indole which provided a balance of favorable properties, such as improved potency and negligible CYP3A4 upregulation, while also exhibiting 15% oral bioavailability (Table 4).

Table 3. Optimization of Piperidine Side Chain and Indole-Core.

^aAverage with the standard deviation reported in parentheses, where n ≥ 2.

Table 4. In Vitro and In Vivo Characterization of 1 and 21.

Assay	1	21
EZH2 IC50 (nM)	0.18	0.057
H3K27Me3 EC50 (nM)	2.5	1.8
cLogD (StarDrop)	3.8	3.1
CYP3A4 UpReg @ 10 μM	25X	2.2X
KARPAS-422 GI50 (8 days) (nM)	N/A	12.2
τ (h)	46.3	57.5
Mouse IV Clearance (L/h/kg)	0.56	3.35
Mouse oral bioavailability - F (%)	40	15
AUClast/D (h·mg/mL)	414.4	46.5
Plasma Protein Binding - m, h	99.8, 97	87, 97
KARPAS-422%TGI	77 @ 50 MPK BID-SC	91 @ 25 MPK BID-PO

Next, we profiled 21 in the KARPAS-422 lymphoma mouse model at 25 mg/kg BID-PO, which is nearly a 6-fold lower dose than the benchmark in Figure 3. In this study, we found that the tumors began to regress at day 14 and resulted in TGI of 75% by day 21, with no impact on mouse body weight (see Supporting Information). Unbound plasma levels of 21 were found to be above H3K27Me3 mark reduction assay EC₅₀ in the tumors for 4 h. Tumor PD experiments showed that H3K27Me3 mark was significantly decreased at day 14 (Figure 4). The oxidative metabolic stability for 21 was not characterized in metabolic liability identification studies. However, we did not observe 10 or 11 during mouse PK experiments of 1.

Figure 4.

(a) Evaluation of 21 in KARPAS-422. (b) Pharmacokinetics of 21 at day 1 of treatment. (c) Modulation of H3K27Me3 pharmacodynamic marker upon treatment for 14 days.

In summary, the goal of our second generation EZH2 medicinal chemistry campaign was to improve target coverage and to limit CYP3A4 upregulation. Starting with a reassessment of the binding interaction of the inhibitor and EZH2, we identified the 4-thiomethyl substituent as a key modification for the enhancement of potency and residence time in 1. We then tackled the challenge of retaining the improved cellular potency and limiting CYP3A4 upregulation through the identification of an exocyclic azetidine side chain in 17. As a result, we were able to expand the utility of an EZH2 inhibitor to an enzalutamide-resistant prostate cancer mouse model (22Rv1). Although 17 provided impressive efficacy, it was not orally bioavailable. Therefore, we replaced the indole-core with a set of heterocycles, which led to the discovery of 21.

Aza-indole 21 was found to be a highly efficacious EZH2 inhibitor, achieving >91% TGI in 25 days with the oral dose of 25 mg/kg BID-PO. Pharmacokinetic (PK) and pharmacodynamic (PD) profiling showed that the exposure levels of 21 were sustained above cellular EC₅₀ for 4 h and the level of H3K27Me3 mark reduction was also maintained. The discovery of the 4-thiomethyl pyridone modification and its ability to bestow durable target coverage was a key stepping-stone to developing second generation EZH2 inhibitors.

Chemistry

We took advantage of the well-precedented peptide coupling of the aminopyridone and the indole acid to synthesize most of the compounds included in this report.¹⁶ The 4-thioether amino-pyridones were synthesized by adapting the route from Scheme 1, while the 6-methoxy and 6-chloropyridones were synthesized via the 6-hydroxypyridone carbonitrile (see Supporting Information). The synthesis of the 4-thiomethyl aminopyridone intermediate 1 commenced by condensing carbon disulfide onto acetone to give sodium bisthiolate Int1a. The thiolates were then capped with methyl iodide to provide thioacetal Int1b, which was subjected to an aldol condensation reaction with 2-cyanoacetamide, that results in eventual cyclization under thermal conditions to provide cyanopyridone Int1c. The resulting nitrile was reduced to the primary amine, which was protected with Boc anhydride to simplify its purification and isolation (24% yield from acetone). Subsequent protecting group removal with hydrochloric acid resulted in a benchtop stable salt (>6 months) of 4-thiomethyl aminopyridone Int1.

Scheme 1. Synthesis of 4-Thiomethyl Aminopyridone Int1.

Reagents and conditions: (a) t-BuONa (2 equiv), CS₂ (1.0 equiv), toluene, 0 °C, 4 h; (b) MeI (2.0 equiv), MeOH, 70 °C, 1 h (42% yield over two steps); (c) t-BuONa, t-BuOH, 80 °C, 12 h (75% yield); (d) BH₃-Me₂S (4.0 equiv), THF, 70 °C, 2 h; (e) Boc₂O (3.0 equiv), Et₃N (1.5 equiv), THF, 25 °C, 12 h (75% yield over two steps); (f) HCl, MeOH, 70 °C, 2 h.

Compounds containing the trifluoroehtylpiperidine side chain were synthesized using the general scheme previously described¹⁶ (see Supporting Information for synthesis of pyridones). The synthesis of 17 (Scheme 2) began with demethylation of the commercial (R)-p-methoxybenzylamine, followed by its reduction, using PtO₂ at 55 psi of H₂, to aminoalcohol 27. The aminoalcohol was then condensed with the bromoketoester to generate enamine 28 and subsequently cyclized to furnish the indole with the RuPhos precatalyst (Gen II) in 88% yield. Cyclohexanol 29 was oxidized to ketone 30, which also served as a key intermediate for 14–15 and 18–20.

Scheme 2. Synthesis of 17.

Reagents and conditions: (a) HBr (35% in AcOH), 100 °C, 12 h; (b) PtO₂ (0.05 equiv), H₂ (55 psi), AcOH (1.5 equiv), MeOH, 50 °C, 30 h; (c) MeONa (1.1 equiv), MeOH, 25 °C, 1 h then ketoester (1.1 equiv), AcOH (1.3 equiv), tBuOH, 85 °C, 12 h (7.6% yield over 3 steps); (d) RuPhos precatalyst (generation 2) (10 mol %), dioxane, 100 °C, 13 h (88% yield); (e) Dess–Martin reagent (2.0 equiv), DCM, 10 °C, 16 h (77% yield); (f) iPr₂NEt (8.3 equiv), MeOH, 25 °C, 1 h then 30 (1.0 equiv), THF, 25 °C, 2 h then LiBH₄ (1.4 equiv), −70 °C, 15 min (45% yield for desired diastereomer); (g) NaOH (10.0 equiv), MeOH, H₂O, 80 °C, 12 h; (h) perfluorophenyl 2,2,2-trifluoroacetate (1.5 equiv), pyridine (2 equiv), DCM, 25 °C; (i) Int1 (3.0 equiv), iPr₂NEt (3.0 equiv), DMF, 25 °C, 12 h (31% yield over 2 steps).

The methoxyazetidine was appended via reductive amination using a two-step procedure of imine formation and reduction with lithium borohydride, which provided the desired diastereomer in an 8:1 ratio (use of sodium cyanoborohydride led to a 1:1 mixture of both diastereomers). The two diastereomers were readily separated by column chromatography. The desired diastereomer, 32, was hydrolyzed and activated as a perfluorophenyl (PFP) ester and then directly treated with the aminopyridone Int1 to give 17.

Glossary

Abbreviations

PTM

post-translational modification

PRC2

polycomb repressive complex 2

H3K27

histone 3 lysine 27

EZH2

enhancer of zeste homologue 2

SWI/SNF

SWItch/Sucrose nonfermentable

SAM

S-adenosyl methionine

PD

pharmacodynamics

cLogD

calculated LogD

SAR

structure−activity relationship

PK

pharmacokinetics

SC

subcutaneous

IV

intravenous

PO

per os

F

bioavailability

DMA

dimethylacetamide

PEG

polyethylene glycol

iPrNEt2

Hunig’s base

CDI

carbonyldiimadazole

THF

tetrahydrofuran

TGI

tumor-growth inhibition

BID

twice a day

QD

once a day

m

mouse

h

human

PtO₂

platinum oxide

PFP

perfluorophenyl

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00045.

• Methods and materials, biochemical and cellular assay procedures, experimental details for in vivo experiments, synthetic experimental details, and compound purity assessment (PDF)

Author Present Address

^† (Alexandre Côté) Schrödinger Inc., New York, NY.

Author Present Address

^§ (Shilpi Arora and Shruti Apte) X-Chem Inc., Waltham, MA.

Author Present Address

^∥ (Ludivine Moine) Blueprint Medicines, Cambridge, MA.

Author Present Address

^⊥ (Jehrod Brenneman) KSQ Therapeutics, Cambridge, MA.

Author Present Address

^# (Jacob I. Stuckey, James E. Audia, and Robert J. Sims, III) Third Rock Ventures, Boston, MA.

Author Present Address

^∇ (Ashwin Ramakrishnan) BMS, Cambridge, MA.

Author Present Address

^○ (Kamil Bruderek) Relay Therapeutics, Cambridge, MA.

Author Present Address

^@ (William D. Bradley) Cyteir Therapeutics, Lexington, MA.

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Author Contributions

^‡ A.K. and A.C. contributed equally.

The authors declare no competing financial interest.