Conformational-Design-Driven Discovery of EZM0414: A Selective, Potent SETD2 Inhibitor for Clinical Studies.

SETD2, a lysine N-methyltransferase, is a histone methyltransferase that plays an important role in various cellular processes and was identified as a target of interest in multiple myeloma that features a t(4,14) translocation. We recently reported the discovery of a novel small-molecule SETD2 inhibitor tool compound that is suitable for preclinical studies. Herein we describe the conformational-design-driven evolution of the advanced chemistry lead, which resulted in compounds appropriate for clinical evaluation. Further optimization of this chemical series led to the discovery of EZM0414, which is a potent, selective, and orally bioavailable inhibitor of SETD2 with good pharmacokinetic properties and robust pharmacodynamic activity in a mouse xenograft model.

Lysine N-methyltransferases (KMTs) are a unique class of enzymes that mediate the methylation of lysine residues of histones.[1] Within this class, KMTs can be further divided into two subsets: SET-domain-containing and non-SET-domain. SETD2 is a member of the former subset and is the only known histone methyltransferase (HMT) that can catalyze the deposition of the third methyl group onto the dimethylated state of lysine 36 of histone H3 (H3K36me2). This methylation results in the trimethylated state of lysine 36 of histone H3 (H3K36me3). While H3K36me3 is generally associated with active transcription, SETD2 also plays a role in various other cellular processes, including transcriptional elongation and regulation, alternative RNA splicing, DNA damage repair, and crosstalk with other histone modifications.[2] In addition, SETD2 has been shown to behave as a tumor suppressor, since homozygous loss-of-function mutations have been associated with tumorigenesis and/or chemoresistance in certain solid tumors, most notably in renal cell carcinoma.[2] While SETD2 may function as a classical tumor suppressor in the context of certain solid tumors, this is not true for SETD2’s role in the development of hematological malignancies such as leukemia and lymphoma. In these disease settings, SETD2 loss-of-function mutations are consistently heterozygous, suggesting that a haploinsufficient tumor suppressor role for SETD2 and therapeutic potential of SETD2 inhibition may also exist.[3−9]

We proposed that inhibition of SETD2 could be exploited in B-cell malignancies where perturbation of lysine 36 of histone H3 (H3K36) methylation plays a key role, including histone H1 mutations in DLBCL, and in the setting of t(4;14) chromosomal translocations present in high-risk multiple myeloma.[10,11] In t(4;14) multiple myeloma, MMSET (also known as NSD2) is highly expressed as a result of the translocation and has been shown to play a critical role in tumorigenesis. MMSET is a histone methyltransferase that catalyzes H3K36me1 and H3K36me2 formation, causing increased levels of H3K36me2, the substrate for SETD2.[12,13] We hypothesized that inhibiting SETD2 would impair tumor growth in the setting of a t(4,14) translocation, which would establish the proof of concept in this and other B cell malignancies (e.g., DLBCL) that may demonstrate dysregulated H3K36me3 or a dependence on SETD2.

In our recent report on the discovery of the SETD2 inhibitor tool compound EPZ-719, we focused primarily on exploration of the structure–activity relationship (SAR) with respect to the indole motif and the central core of the molecule, leading to SETD2 inhibitor 2 and the closely related analogues 3 and 3-F (see Figure ).[14] The unique binding mode of these aniline-containing inhibitors was also disclosed in that report. Anilines are prevalent in modern drug discovery screening libraries because of their ease of synthesis, but this structural motif tends to impart less-than-ideal pharmacokinetic properties and potential metabolism-derived toxicities.[15] We reported that saturation of this ring significantly improved the physiochemical properties of the series and avoided the potential Ames toxicity; however, we did not detail the conformational studies leading us to the (1R,3S) stereochemistry seen in EPZ-719. The promising in vivo results with this tool compound led us to continue our work with this prototype in the pursuit of a suitable clinical candidate.[16]

Figure 1

(top) Structures of SETD2 inhibitors 1–3. (bottom) X-ray cocrystal structure of SETD2 and compound 2 (in pale cyan) (PDB ID 7LZD). Hydrogen bonds are shown with dotted yellow lines.

The importance of reducing the basicity of the piperazine nitrogen with an electron-withdrawing group became apparent in our previous efforts to optimize the potency, selectivity, and PK properties of this early series.[14] Indolylcarboxamides 2 and 3 had been identified as promising leads with potent in vitro activity; however, translating that to in vivo potency proved to be challenging. In our cellular assay, indolylcarboxamide 3-F (SETD2 biochemical IC50 = 12 nM and H3K36me3 ICW IC50 = 41 nM), a close analogue of 3, was 6-fold more potent than 2 (SETD2 biochemical IC50 = 18 nM and H3K36me3 ICW IC50 = 235 nM), which lacked the electron-withdrawing group. While this optimization strategy continued to improve the cellular potency of analogues like 3, the analogues continued to suffer from suboptimal PK profiles, with moderate to high clearance in mouse. We suspected that this was due to the high lipophilicity of the series. An overall increase in saturation is measured by Fsp, and the number of chiral centers in a molecule can improve its chance for clinical success.[17] This is likely due to a decrease in lipophilicity (highly lipophilic compounds (cLogP > 3; tPSA < 75) tend to have some associated promiscuity),[18,19] an increase in solubility (better permeability and exposure), and better occupancy of the binding pocket (which may provide better target specificity). A suitable value is reported to be Fsp ≥ 0.42; however, compound 3 falls below this value (Fsp = 0.27).[20] Thus, our optimization efforts focused on improving the lipophilic efficiency to mitigate metabolic liabilities while retaining potency and HMT selectivity.

During analogue design, compounds biased to reside in the cis conformation were prioritized because the cocrystal structure of 3 suggested that this would best mimic the planar aromatic ring seen in the early leads 2 and 3 (Figure ).[14] Subsequent conformational analysis showed that cis-cyclohexanes would have a strong energetic bias to reside in conformer I, thereby avoiding 1,3-diaxial strain from the 1,3-disubstitution (Figure ), and this was supported by molecular dynamics simulations. Molecular modeling predicted conformer I to bind well in the ligand pocket compared with the trans-cyclohexane in conformers III and IV, where the conformational changes of the ligand binding site would be less tolerated because of the axial substituent (see the Supporting Information). Analysis of the chair structures for conformer I in the binding site with replacement of compound 2 in the cocrystal structure shown in Figure suggested that there could be a stereochemical preference within the cis-cyclohexane conformations (I-A and I-B). To better understand this, both cis- and trans-(1,3)-diaminocyclohexanes were initially prepared simultaneously from racemic starting materials and isolated via chiral resolution of the diastereomeric product mixture, as shown in Scheme (top). The relative stereochemistry of each stereocenter in 4–6 was determined via 2D NMR experiments after chiral separation. The collection of biochemical data for each (1S/R,3S/R) stereoisomer revealed that there was indeed a preferred cis-chair conformation along with a stereochemical preference for the 1R configuration at the carboxamide junction resulting in greater activity. The absolute stereochemistry of the most active isomer, EZM0414, was confirmed to be (1R,3S) via X-ray crystallography, which was consistent with the stereochemistry of the starting (1R,3R)-3-aminocyclohexan-1-ol 14 upon resynthesis, as shown in Scheme (bottom). This alternative route utilizing chiral starting materials was also employed with an emphasis on piperazine analogues and the corresponding bioisosteres (Scheme , bottom).

Figure 2

Potential chair conformational isomers of cis- and trans-(1,3)-diaminocyclohexyl analogues and stereochemical considerations.

Scheme 1

Synthesis of Stereoisomers of EZM0414 and Synthetic Strategy for Advanced Analogues 7-9

The conformational impact of the cyclohexyl region was confirmed with a cocrystal of EZM0414 and its diastereomer 6 bound to SETD2, as rendered in Figure . In both structures, the indole motif is positioned deep within the lysine channel with two specific hydrogen bonds to the main-chain atoms of Phe1606, consistent with our earlier series exemplified by 3. In the bound conformation, a well-ordered water network suggests the possibility of a water-mediated hydrogen-bonding interaction between the carboxamide and the nearest acceptor residue, Gln1676. With EZM0414, the cyclohexyl ring adopts a rigid cis-chair conformation stabilizing the Tyr1671 residue in an “out” position via a hydrophobic interaction. The (1R,3S) stereochemistry of the ring serves to orient the piperazine substituent in an appropriate direction toward the solvent. With 6, the cyclohexyl ring adopts a trans-chair conformation that allows the Tyr1671 residue to be in an “in” position, and the (1R,3S) stereochemistry of the ring gives preference to the piperazine for the equatorial position (conformer IV in Figure ).

Figure 3

(left) X-ray cocrystal structure of SETD2 and EZM0414 (in pale green) (PDB ID 7TY2), (middle) X-ray cocrystal structure of SETD2 and 6 (in wheat) (PDB ID 7TY3), and (right) overlay showing Tyr1671 movement (90° rotation). SAM is present in both structures but has been omitted for clarity in the renderings. See the Supporting Information for details.

Since oxidation of tertiary amines is a common metabolic pathway, we postulated that cyclization at the α-position of the piperazine or the N-acetamide might mitigate this risk and potentially lead to further improvements in the overall PK properties. Some of the most potent examples (7–9) are shown in Chart , including [5,5]-fused and [5,6]-fused cyclic analogues. Ultimately, these advanced analogues 7–9 exhibited excellent potency in both our molecular and cellular assays. These compounds were synthesized as mixtures of diastereomers from chiral intermediate 15, starting from 14 following a previously reported reaction sequence. The HATU-mediated coupling of 13 and 14 followed by oxidation of the intermediate alcohol gave ketone 15. The piperazine analogues could be installed via reductive amination and subsequent chiral resolution, and the most potent diastereomer is reported in Table . While absolute stereochemistry was not routinely determined, 2D NMR techniques were employed to assign relative cis/trans stereochemistry for the final compounds (see the Supporting Information).

Chart 1

Stereoisomers Leading to EZM0414 and Advanced Analogues

Table 1

Conformational Activities and Profiles of Advanced Analogues

		SETD2 activity IC50 (μM)a
compound	diaminocyclohexyl stereochemistryb	biochemicalc	H3K36me3 ICWd	14-day antiproliferation EC50 (μM)a,e	cLogP
EPZ-719	(1R,3S)	0.008 ± 0.003	0.020 ± 0.008	0.025 ± 0.003	1.46
3	achiral	0.027 ± 0.006	0.134 ± 0.066	0.245	2.92
3-F	achiral	0.012 ± 0.017	0.041 ± 0.002	0.060	3.07
4	(1S,3R)	>10	>2	>10	1.85
5	(1S,3S)	>10	>2	>10	1.85
EZM0414	(1R,3S)	0.018 ± 0.009	0.031 ± 0.014	0.057 ± 0.003	1.85
6	(1R,3R)	1.24 ± 0.66	>2	7.79	1.85
7	(1R,3S)	0.015 ± 0.009	0.011 ± 0.003	0.015	2.13
8	(1R,3S)	0.011 ± 0.003	0.014 ± 0.002	0.027	1.97
9	(1R,3S)	0.009 ± 0.002	0.005 ± 0.001	0.017	1.91

Data are reported as geometric means of at least two test results.

Absolute stereochemistries were determined by 2D NMR spectroscopy and X-ray crystallography.

SETD2 biochemical assay.

Cellular potency in the A549 cell line.

KMS-34 cells were incubated with compound for 14 days.

With the improved cyclohexyl series, it became evident that we were approaching the enzymatic assay limits and could no longer reliably rank-order the compound potencies on the basis of this parameter alone. We sought to probe the response to SETD2 inhibition in a subset of cell lines that would be sensitive to SETD2 inhibition, mainly featuring a t(4,14) translocation. As can be seen in Table , we assessed the growth inhibition in KMS-34 cells after 14 days of compound treatment. With this series, cellular potency translated well to our long-term proliferation (LTP) assays, with potent growth inhibition observed with EZM0414 and compounds 7–9.

We next sought to identify a drug candidate with optimized PK and safety profiles to support once-daily dosing in the clinic. Further ADME profiling of EZM0414 and analogues 7–9 (Table ) showed an improvement in hepatocyte stability compared with 3. Further characterization of piperazine bioisosteres 8 and 9 revealed that they suffered from poor apparent permeability in vitro. Despite subsequent mouse PK studies showing compounds 8 and 9 to have reasonable oral absorption, 17–30% of the parent was recovered in feces following iv administration, suggesting moderate direct elimination as a clearance mechanism (data not shown). While both EZM0414 and analogue 7 showed favorable PK profiles following 50 mg/kg po administration in mice, EZM0414 had ∼2-fold higher exposure (AUC) with better oral bioavailability (F) than 7, a trend that was observed in rats as well. Table presents a comparison of the PK parameters for EZM0414, 7, and the starting parent compound 3. Improving the lipophilic efficiency of the earlier series vastly enhanced the overall PK profile, with a 17-fold increase in overall exposure levels (AUC0–∞) achieved in CD-1 mice at 50 mg/kg following oral administration with EZM0414. With regard to the potential for drug–drug interactions as perpetrator of CYP enzymes, EZM0414 exhibited only weak inhibition of CYP isoform 2C8 (4.8 μM), and no inhibition of other tested isoforms was seen (IC50 > 30 μM for isoforms 1A2, 2B6, 2C9, 2C19, 2D6, and 3A4). Analogue 7 showed modest inhibition of isoforms 2B6 (3.2 μM) and 2D6 (10.8 μM), with IC50 > 25 μM for the remaining isoforms tested. Additional characterization of EZM0414 indicated a favorable safety pharmacology profile. In vitro testing of EZM0414 in a safety panel consisting of 47 targets and a diversity panel of 72 kinases showed IC50 > 25 μM for all targets except D2 (IC50 = 13.0 μM, antagonist) and 5-HT1B (IC50 = 3.2 μM, agonist).

Table 2

In Vitro ADME for Advanced Analogues

	permeability	stability in hepatocytes
compound	Caco-2 Papp (a-b:b-a) (10–6 cm/s)a	scaled h/r/m Clint (mL min–1 kg–1)b
EPZ-719	3.2:18	51/137/721
3	6.3:5.2	32/283/678
EZM0414	3.2:20	9/154/216
7	4.3:17	23/69/425
8	<1.0:32	18/265/889
9	<1.0:22	5/72/103

Parent compound (5 μM) was incubated for 2 h at 37 °C.

Parent compound (0.5 μM) was incubated for 60 min for mouse (m) and rat (r) hepatocytes and 120 min for human (h) hepatocytes. See the Supporting Information for details.

Table 3

Comparison of Pharmacokinetic Parameters of 3, EZM0414, and 7 Following Intravenous and Oral Administration

parameter	3	EZM0414	7
CD-1 IGS Mousea
dose (mg/kg) iv/po	2/50	2/50	1/50
Cl (mL min–1 kg–1)b iv	36	43	35
Vss (L/kg)c iv	1.1	4.7	2.9
Cmax (μg/mL)d po	0.5	8.6	7.6
AUC (μg h/mL)e iv/po	0.9/1.6	0.8/27	0.5/14
t1/2 (h)f iv/po	0.4/0.8	1.8/1.8	1.6/0.8
F (%)g po	7	>100	55
Sprague–Dawley Rata
dose (mg/kg) iv/po	2/10	2/10	1/10
Cl (mL min–1 kg–1)b iv	19	16.2	14
Vss (L/kg)c iv	0.65	4.1	2.7
Cmax (μg/mL)d po	0.2	2.1	1.6
AUC (μg h/mL)e iv/po	2.3/0.7	2.1/10.5	1.3/7.7
t1/2 (h)f iv/po	0.8/2.3	4.0/3.8	4.1/2.7
F (%)g po	6.2	97	60

Data are reported as averages over at least two animals.

In vivo clearance.

Volume of distribution at steady state.

Maximum concentration.

Area under the concentration–time curve from zero to infinity.

Elimination half-life.

Oral bioavailability. Oral dosing solutions were prepared using 0.5% CMC and 0.1% Tween in water at pH 4.

To demonstrate the impact of SETD2 inhibition on H3K36me3 levels and tumor growth in vivo, we conducted an in vivo efficacy study with EZM0414 in a NOD SCID mouse xenograft model implanted with human KMS-11 cells (multiple myeloma cell line; 14-day LTP IC50 = 370 ± 224 nM). Daily dosing of 15 and 30 mg/kg EZM0414 bid in mice was well-tolerated and induced tumor growth reductions of 60 and 91%, respectively (Figure ). EZM0414 exhibited dose-proportional exposures at both doses tested (Table ). Further analysis of intratumoral H3K36me3 levels taken from samples 12 h after administration of the last dose showed a significant modulation of H3K36me3 expression, demonstrating on-target inhibition of SETD2 methyltransferase activity in vivo.

Figure 4

In vivo efficacy with EZM0414, dosed twice daily in a KMS-11 cell line-derived xenograft model. ****, p < 0.00001.

Table 4

PK–PD Relationship of EZM0414 in a KMS-11 Xenograft Tumor Mouse Model

PO dose (mg/kg)a	AUC (μg h/mL)b	PD (%)c
15	9.4	90
30	21.4	93

bid po dosing, prepared using 0.5% CMC and 0.1% Tween in water at pH 4.

Area under the plasma concentration–time curve from zero to 24 h after last dose.

Percent reduction in H3K36me3 measured in the tumor 12 h after last dose compared with vehicle.

In summary, structure-based drug design and conformational analysis of 1,3-disubstituted cyclohexanes were considered in designing a replacement for the aromatic core found in compounds 2 and 3. Evaluation of the different possible cyclohexyl isomers identified an active cis stereoisomer, and the absolute stereochemistry was confirmed to be (1R,3S). The strategic increase in Fsp resulted in a decrease in lipophilicity and led to the identification of EZM0414, a potent and selective inhibitor of SETD2 with good pharmacokinetic properties. For these reasons, EZM0414 was progressed through IND-enabling studies and is undergoing Phase 1 clinical trials.