Discovery and lead identification of quinazoline-based BRD4 inhibitors

Abstract

A new series of quinazoline-based analogs as potent bromodomain-containing protein 4 (BRD4 inhibitors is described. The structure-activity relationships on 2- and 4-position of quinazoline ring, and the substitution at 6-position that mimic the acetylated lysine are discussed. A co-crystallized structure of 48 (CN750) with BRD4 (BD1) including key inhibitor-protein interactions is also highlighted. Together with preliminary rodent pharmacokinetic results, a new lead (65, CN427) is identified which is suitable for further lead optimization.

Graphical Abstract

The bromodomain-containing proteins are critical components involved in the epigenetic regulation of gene transcription and downstream processes via recognition of specific acetylated lysine residues on N-terminal histone tails.¹ In certain diseases, epigenetic deregulation of gene transcription has resulted in overexpression of growth promoting genes and activation of proinflammatory cytokines that contribute to the progression of cancer and inflammatory diseases.² Therefore, the disruption of acetylated lysine-bromodomain interactions could have potential therapeutic benefits through the modulation of disease-related dysfunctional gene transcription.³

Among known bromodomain-containing proteins, the bromodomain and extraterminal domain (BET) family that comprises four members (BRD2, BRD3, BRD4, and BRDT; each containing two N-terminal bromodomains BD1 and BD2) has been an attractive target for treatment of cancers and other diseases.⁴ Many structurally diverse BET family inhibitors have been reported in recent years.⁵ Some representative molecules highlighted in Figure 1, the I-BET762 (1) and OTX-015 (2, also known as MK-8628) that were structurally similar to the original tool compound JQ1 (3),⁶ have been evaluated clinically for potential treatment of cancers.⁷ Most recently, another clinical candidate, structurally distinct BET inhibitor ABBV-075 (4, mivebresib) with high efficacy in Kasumi-1 xenografts was also reported.⁸ These molecules utilize moieties such as 3,5-dimethyltriazole, 3,5-dimethylisoxazole, and pyrrolo-1-methylpyridin-2-one, which provide critical H-bond interactions with key amino acid residue of bromodomains (e.g. Asn140 in BRD4 (BD1)) and to further inhibit their biological function by interrupting acetylated lysine-bromodomain interactions. Although several compounds have been successfully progressed into the clinic, alternative BET family inhibitors with differentiated core structures, possessing better safety profiles and physicochemical properties remains of high interest.

Figure 1.

Representative BET inhibitors and quinazoline-based inhibitors.

In our efforts to identify novel BET family inhibitors, the 2-quinolinone analog 5 was developed recently that exhibited good affinity against BRD4 (BD1,2) (BROMOScan K_d = 55 nM) with moderate cellular activity (MV4–11 IC₅₀ = 1.8 μM).⁹ By utilizing 5 as a starting point, we envisioned the bicyclic quinazoline could serve as a structurally distinct template that allows facile functionalization (e.g. at 2-, 4-, 6, and 7-position of quinazoline ring) improving potency and drug-like properties. Herein we report on the initial discovery and optimization of a new chemical series of quinazoline-based BRD4 inhibitors focusing on the lead identification.

The synthesis of these quinazoline-based analogs is straightforward and shown in Scheme 1. Replacement of 4-chloro group of commercially available material 7 with the corresponding amine followed by Suzuki coupling of the bromo group with 3,5-dimethylisoxazole-4-boronic ester at lower temperature (e.g. 70 °C) gave intermediate 8. Further utilizing the 2-chloro functionality of 8, the desired analogs were obtained by either replacement with amine (e.g. 14–24) or by Suzuki coupling with corresponding boronic acid or ester at 90–95 °C (e.g. 25–32). This route is particularly suitable for the investigation of SAR at the 2-position of quinazoline ring. Compound 40 was prepared by the same sequence starting from 2,4-dichloro-7-bromoquinazoline instead. For rapid screening of a potential replacement for the 3,5-dimethylisoxazole moiety, intermediate 9 was prepared in a similar manner, which could undergo Suzuki coupling to install various heteroaryl groups at the 6-position (e.g. 41–48). Furthermore, to facilitate the SAR investigation at the 4-position, t-butyl protected intermediate 10 was prepared. The deprotection of t-butyl group followed by phosphonium salt activated replacement of hydroxyl group gave desired products 49–54, 56–58, and 60–64.¹⁰ Finally, intermediate 12 was prepared by a similar phosphonium salt activated protocol from the tautomerizable substrate 11 followed by Suzuki coupling to install the requisite dimethylisoxazole ring. Hydrolysis and subsequent amide formation using HATU as coupling agent afforded analogs 33–39 containing the 2-substituted amide.

Scheme 1.

Reagents and conditions: (a) R¹R²NH, Et₃N, THF, rt or 60 °C; (b) 3,5-dimethyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)isoxazole, cat. PdCl₂(dppf)-CH₂Cl₂ adduct, K₂CO₃, 1,4-dioxane/H₂O (3/1), 70 °C; (c) corresponding amine, (i-Pr)₂NEt, EtOH, 90 °C, sealed; or corresponding boronic acid or boronic ester, cat. PdCl₂(dppf)-CH₂Cl₂ adduct, K₂CO₃, 1,4-dioxane/H₂O (3/1), 90–95 °C; (d) (3-chlorophenyl)methanamine, Et₃N, THF, rt, 1 h, 97%; (e) N,N-dimethyl-2-(piperazin-1-yl)ethanamine, (i-Pr)₂NEt, EtOH, 90 °C, sealed, 5 h, 95%; (f) corresponding boronic acid or boronic ester, cat. PdCl₂(dppf)-CH₂Cl₂ adduct, K₂CO₃, 1,4-dioxane/H₂O (3/1), 90–95 °C; (g) t-BuOK (1 M in THF), THF, 0 °C, 1.5 h; (h) N,N-dimethyl-2-(piperazin-1-yl)ethanamine, (i-Pr)₂NEt, EtOH, 90 °C, sealed, 5 h, 76%; (i) HCl (4 M in dioxane), CH₂Cl₂, rt, 5 h; (j) corresponding amine, bromotri(pyrrolidin-1-yl)phosphonium hexafluorophosphate (PyBrop), Et₃N, 1,4-dioxane, rt or 70 °C; (k) (3-chlorophenyl)methanamine, bromotri(pyrrolidin-1-yl)phosphonium hexafluorophosphate (PyBrop), Et₃N, 1,4-dioxane, rt, 5 h, >96%; (l) 1 N NaOH_(aq), THF/MeOH (10/1), 50 °C, 2 h, 70%; (m) corresponding amine, HATU, (i-Pr)₂NEt, DMF, rt.

The initial goal was focused on quick screening of substitutions at 2-, 4-, and 6-position of quinazoline ring, and to collect the structure-activity relationships (SAR) information that could support the identification of suitable leads. The competitive binding affinity (K_d) with BRD4 (BD1,2) as the primary assay was measured using BROMOScan platform.¹¹ Additionally, the viability of MV4–11 cells (a leukemia cell line) was used to assess cellular activity (i.e. viability, IC₅₀) of the synthesized analogs. The results compiled in Table 1 focused on SAR at the 2-substituted group (R¹). The prior art compound 1 that demonstrated potent binding affinity (K_d = 37 nM) and MV4–11 viability (IC₅₀ = 0.8 μM) was included in assays as a reference for direct comparison. For the smaller substitutions such as chloro (13) and methylamine (14), the binding affinity was diminished with K_d values of 980–2900 nM. However, larger substitutions particularly with a basic group, e.g. 15–16, had ~20-fold improvement in K_d values. Further extended side chain (e.g. 17–19) seemed much favored with K_d below 100 nM range. Furthermore, these basic side chains (e.g. 15–19) also provided better cellular activity in the low micromolar range. The terminal polar group appears to be involved in an important H-bonding interaction, as the lack of the nitrogen atom (20) significantly decreased the potency in both K_d and MV4–11 viability. Replacement of the piperazine ring with homopiperazine analog 21 is tolerated, albeit slightly less potent in K_d (150 nM). Some extended amide side chains (22–24) were also well tolerated with good potency maintained. The preliminary SAR information indicated that the 2-substitution is pointing to a polar region, perhaps the solvent area or the ZA channel that is surrounded by polar amino acid residues such as Gln85 and Asp88. Replacing piperazine ring with other heteroaryl groups, e.g. pyridine (25) or pyrazole (26), slightly decreased potency (K_d = 140–250 nM). Interestingly, attaching a pyrazole ring from its 4-position (27) regained potency back to original range (i.e. compared to 17). Consistent with SAR observed above, the side chains containing a terminal group with the capability of forming H-bonds, e.g. OH, OMe, N(Me)₂, or amide (27–32), were mostly preferred.

Table 1.

SAR of R¹ substitutions.

^aValues are the average of two distinct runs.

^bValues are the average of three distinct runs.

^cInternal re-screened data.

Insertion of an amide functionality between the basic side chain and quinazoline core, e.g. 33, decreased potency dramatically. While the aniline type amide substitutions (e.g. 37, 39) were disfavored, the amide substitutions generated from heterocyclic or heteroaryl-methyl amines (e.g. 34–36, and 38) were well tolerated with appreciable potency in both binding affinity and in MV4–11 cells.

We next screened the substitution that mimics the acetylated lysine, e.g. dimethyl isoxazole group, maintaining dimethylaminoethylpiperazine as 2-substituted group (Table 2). Moving the dimethylisoxazole ring to 7-position caused ~12 fold potency drop in binding affinity (K_d = 56 and 610 for analog 17 and 40, respectively). Surprisingly, simply removing one of the methyl groups from dimethylisoxazole ring resulted in a significant decrease in potency (41–42). Other heteroaryl groups, such as dimethylpyrazole (43) or hydroxyl pyridines (44–46), were also not well tolerated. Although the N-methylpyridone (47) was much less potent (K_d = 7100 nM), switching the carbonyl position (48, CN750) did show marked improvement of potency of K_d and cellular activity at 44 nM and 0.54 μM, respectively.¹² These results indicated the substitution in this region is critical to provide key interaction with Asn140 and for maintaining good binding affinity.

Table 2.

SAR of R² substitutions.

^aValues are the average of two distinct runs.

^bValues are the average of three distinct runs.

To allow direct comparison with 17, we maintained the basic side chain and dimethylisoxazole group at 2- and 6-position, respectively, and explored the SAR of substituents at 4-position of quinazoline ring (Table 3). The results compiled in Table 3 revealed that the fluoro substitution is comparable but slightly less potent (49, K_d = 170 nM), While the ortho-chloro substitution (50) is well tolerated, the para-chloro substitution was disfavored causing ~9-fold decrease in of potency (51, K_d = 520 nM). Heteroaryl groups, such as pyridine, was found to be a suitable replacement for phenyl, with the 3-pyridine seemed better than 2-pyridine (52 vs. 53) and the thiophene bioisostere (54) exhibited similar potency. However, the aniline substitution (55) or chlorophenyl side chain (56) resulted in significantly decreased potency. Meanwhile, adding small substitution group on the nitrogen (57) or at benzylic position (58–60) slightly decreased the binding affinity ~2–3 fold.

Table 3.

SAR of R³ substitutions.

^aValues are the average of two distinct runs.

^bValues are the average of three distinct runs.

To gain structural insight into the binding mode and guide further lead optimization, analog 48 was co-crystallized with BRD4 protein.¹³ The screening of melting temperature of both BRD4 BD1 and BD2 domains with 48 indicates the BD1 domain is better stabilized (ΔT_m = +15.0 °C and +11.0 °C for BD1 and BD2 domain, respectively, with 48 (at 250 μM)). Therefore, 48 was soaked with BRD4 (BD1) protein to obtain the crystal which diffracted to a resolution of 1.26 Å (PDB code: 6E4A). The crystal structure of the BRD4 (BD1) complex with 48 contained two protein monomers in the asymmetric unit, termed monomer A and monomer B. It is worth noting that the electron density map for 48 in the monomer B binding site is better defined than that of monomer A. The largest difference of 48 in both monomers is apparent in the conformation of terminal dimethylamino side chain. The alignment of 48, surface area of protein and key interactions of 48 with BRD4 (BD1) of monomer B are highlighted in Figure 2. The binding pocket consists mainly of hydrophobic residues which provide most of the interactions with the compound 48. The N-methyl-pyridone is found deep in the BRD4 (BD1) binding pocket between Ile146 and Val87 where it forms van der Waals interactions with these residues. The oxygen of N-methyl-pyridone forms H-bonding both with key Asn140 side chain, the only compound-protein direct H-bonding interaction, and with a water molecule that forms H-bonding with Tyr97 and Cys136. The quinazoline core lies on the WPF shelf and forms π-edge interactions with Trp81. The nitrogen (1-position) of the quinazoline also forms H-bonding with Gln85 through the water molecule. Additionally, the 3-chlorophenyl forms π-edge interactions with Trp81 and directs its chlorine atom in a protein pocket consisting of Met149, Pro82, Asp144 and Asp145. The basic dimethylaminoethylpiperazine side chain did not point to ZA channel. Instead, it forms stacking interactions with the plane of Trp81 side chain and is highly solvent exposed that also provides interactions of the tertiary amine group with a neighboring protein molecule in the crystal.

Figure 2.

Co-crystallized structure of 48 with BRD4 (BD1) at 1.26 Å resolution (PDB code: 6E4A). (A) Analog 48 with surface of BRD4 (BD1); (B) Key interactions of 48 with BRD4 (BD1).

In view of the co-crystallized structure, the region near to the nitrogen atom and benzylic carbon of the 4-benzylic amine substitution seemed quite open to allow further modification. Based on this information, along with the fact that only a slight loss of potency was observed for compounds 57–60, we decided to revisit this region by tethering both substitutions to form a heterocyclic ring, such as 61 (Table 4). While no improvement in potency was obtained for 61, a clear potency trend was observed by increasing the ring size from 4- to 6-membered ring (e.g. 61–63). Furthermore, instead of piperidine ring, the morpholine ring analog 64 had improved binding activity (K_d = 28 nM) and comparable cellular activity (MV4–11 IC₅₀ = 1.83 μM). Finally, the combination of phenylmorpholine moiety and hydroxyethylpyrazole substitution at 2-position gave analog 65 (CN427) with binding affinities and cellular activities comparable to the clinical compound 1.

Table 4.

SAR of cyclic substitutions at 4-position.

^aValues are the average of two distinct runs.

^bValues are the average of three distinct runs.

After completing our initial SAR explorations around R¹–R³ substitutions, a few analogs with suitable potency in both binding affinity and MV4–11 cellular activity, such as 34 and 48, were selected and evaluated to assess their pharmacokinetics (PK) in CD-1 mice. Although analogs 34 and 48 exhibited reasonable mouse liver microsomal stability (MLM, t_1/2 in multi-point format, is 28 and 53 min for 34 and 48, respectively), extremely low drug exposure after oral administration (po, 10 mg/kg) was observed (Table 5, AUC_0–∞ = 54 h●ng/mL and no exposure for 34 and 48, respectively). However, the analog 65 bearing phenylmorpholine and pyrazole substitutions exhibited a more suitable PK profile. The high systemic exposure (e.g. oral AUC_0−∞ = 4494 h●ng/mL) with 51% of bioavailability indicates that 65 is well absorbed following oral administration. The steady-state volume of distribution (V_ss) of 1.7 L/kg also suggests that the compound penetrates to tissues well. Compound 65 also exhibited lower clearance (CL_p= 19.8 mL/min/kg) and long half-life of drug exposure (t_1/2 = 3.1 h and 2.1 h for iv and po administration, respectively). In addition, 65 possessed good permeability (2204 ×10⁻⁶ cm/s) in parallel artificial membrane permeability assay (PAMPA). Taken together, this ADME profile supported that 65 is an appropriate new lead worthy of further optimization.

Table 5.

The PK profiles of selected analogs in CD-1 mice.

Cpd.	administrationa	t1/2 (h)	Cmax (ng/mL)b	AUC0−∞ (h●ng/mL)	Vss (L/kg)	CLp (mL/min/kg)	F%
34	po (10 mg/kg)c	4.0	9	54
48	po (10 mg/kg)d			BLOQe
65	iv (2 mg/kg)f	3.1	1690	1757	1.7	19.8
65	po (10 mg/kg)f	2.1	934	4494			51

^an= 3, iv and po represents intravenous and oral administration, respectively.

^bThe maximum drug concentration (Cmax) was observed at t = 5 min, the first sampling time point after iv administration.

^cThe compound was formulated as a solution in PEG400/water (1/3).

^dThe compound was formulated as a solution in 20% HP-β-CD in water.

^eBLOQ: below limit of quantitation.

^fThe compound was formulated as a solution in 30% solutol in water and adjusted pH with 1.05 equiv of HCl using 0.1 N HCl_(aq).

In summary, a new series of quinazolined-based analogs as potent BRD4 inhibitors has been discovered and optimized. The SAR indicates that pyrazole and phenylmorpholine substitutions are suitable replacements for basic piperazine side chain and benzyl amine substitutions at 2- and 4-position of quinazoline ring, respectively. In addition, our systematic SAR efforts led to analog 65 (CN427) with a markedly improved PK profile. Although the substituted group at 6-position of quinazoline ring is critical, the N-methyl-2-pyridone was found to be a good replacement for dimethylisoxazole ring. The co-crystallized structure of 48 (CN750) with BRD4 (BD1) domain provided an insight into key interactions that could be used for further optimization. With compound 65 in-hand, additional optimization to identify analogs with improved potency and ADME properties, including PK, for in vivo efficacy studies are in progress and will be reported in due course.

Highlights.

• Quinazoline can serve as a novel core structure in the design of potent BRD4 inhibitors

• N-Methyl 2-pyridone is a viable alternative to 3,5-dimethylisoxazole

• Combination of pyrazole and phenylmorpholine substitutions improved pharmacokinetics