Structure-guided design and development of potent and selective dual BRD4/PLK1 inhibitors

Abstract

The simultaneous inhibition of polo-like kinase 1 (PLK1) and BRD4 bromodomain by a single molecule could lead to the development of an effective therapeutic strategy for a variety of diseases in which PLK1 and BRD4 are implicated. Compound 23 has been found to be a potent dual kinase-bromodomain inhibitor (BRD4-BD1 IC₅₀ = 28 nM, PLK1 IC₅₀ = 40 nM). Compound 6 was found to be the most selective PLK1 inhibitor over BRD4 in our series (BRD4-BD1 IC₅₀ = 2,579 nM, PLK1 IC₅₀ = 9.9 nM). Molecular docking studies with 23 and BRD4-BD1/PLK1 as well as with 6 corroborates the biochemical assay results.

Graphical Abstract

Introduction

The bromodomain and extra-terminal domain (BET) family of proteins is essential for recognizing acetylated lysines (KAc) on chromatin and has emerged as a potential therapeutic target for the treatment of cancer, inflammation and viral infectious diseases.^1–8 BET proteins, including BRD2, BRD3, BRD4 and BRDT, are characterized by dual bromodomains (BD1 and BD2) that bind to KAc on histones to facilitate the recruitment and stabilization of protein complexes, which plays an active role in gene expression.^9–13 With the development of the first small molecule inhibitor of BET bromodomains, (+)-JQ1, we were able to demonstrate that binding to BRD4 displaces BRD4 from chromatin and interrupts the biological function of the protein.^3,14 Since our initial reports, several potent and selective BET inhibitors with different chemotypes have been reported to possess similar binding affinities, selectivity profiles and cellular activity as (+)-JQ1.^15–19 More importantly, more than twelve BET bromodomain inhibitors have progressed into clinical trials, including a derivative of JQ1,²⁰ for the treatment of hematologic malignancies, solid tumors and cardiovascular disease.²¹ With multiple cancer studies demonstrating the development of resistance to single agents, including BET inhibitors, strategies for combination therapies are being intently pursued.^22–24

Recent reports highlight that inhibition of polo-like kinase 1 (PLK1), a crucial cell-cycle regulator, synergizes with BET inhibition in multiple cancer types, including prostate cancer and acute myeloid leukemia (AML).^25,26 While combinatorial inhibition of PLK1 and BET bromodomains with two molecules is appears to be synergistic,²⁶ achieving this with a single agent may increase potency for two disease-relevant targets/pathways while decreasing toxicity. The concept has recently been demonstrated with dual PI3K-BRD4 inhibitors have recently been reported to block expression of the MYC oncogene, while markedly inhibiting cancer cell growth and metastasis in hepatocellular carcinoma and neuroblastoma in vivo.^27,28 The recent discovery that multiple kinase inhibitors, including PLK1 inhibitor, BI-2536, and the janus kinase 2 (JAK2) inhibitor, TG101209 (Fig. 1) exhibit moderate to excellent inhibitory activity against BET family of proteins opens up the possibility of creating polypharmacological compounds that target both kinases and BET proteins equivalently.^7,23,29–32

Figure 1.

Kinase inhibitors with nanomolar inhibitory activity against BRD4

In this study, we utilized the scaffold of BI-2536 to develop inhibitors with varied inhibitory activities against BRD4 and PLK1 in order to perform structure-activity relationship (SAR) studies, as well as understand the effect of inhibiting both proteins (PLK1 and BRD4) with a single agent in cancer, as both the BET family of proteins and PLK1 are intently pursued targets for the treatment of cancer.

First, we inspected the binding modes of BI-2536 with BRD4-BD1 and PLK1, respectively, through co-crystallization studies (Fig. 2). In the binding complex of BI-2536 with BRD4-BD1, the methylated amide functions as an acetylated lysine mimic, where the carbonyl oxygen of the dihydropteridinone ring forms a water-mediated hydrogen bond with conserved acetyl lysine recognition motif, asparagine (N140) in the binding pocket. The methyl group is orientated into a hydrophobic sub-pocket formed by F83, I146, and C136. The NH group on the aniline and one of the pyrimidine nitrogen atoms interact with the Q85 backbone amide through an intermediary water molecule. The ethyl group that branches off the asymmetric dihydropteridinone carbon atom is projected into a small hydrophobic sub-pocket formed by V87, L92, L94, Y97, while the cyclopentyl moiety and the N-methylpiperidine are both solvent exposed.^21,22

Figure 2.

Crystal structure of BI-2536 (dark red) bound to A) BRD4-BD1 (PDB ID: 4O74) and B) PLK1 (PDB ID: 2RKU). The protein residues in contact with ligand showed in yellow sticks; hydrogen bonding showed in red dashed lines; the rest residues near binding pocked showed in gray.

Based on structural information from the co-crystals, we envisioned several strategies to improve the BRD4-BD1 binding affinity while reducing the PLK1 inhibition to achieve balanced dual PLK1/BRD4-BD1 inhibition (Fig. 3). First, we hypothesize that the replacement of the amide methyl group at Site 1 with a slightly larger ethyl or isopropyl might enhance binding affinity to BRD4-BD1. Second, to decrease affinity for PLK1, the elaboration of the 3-methoxy group on the aromatic ring (Site 2) with a bulkier group might modulate binding affinity. This methyl ether group is involved in critical hydrogen bonds with the PLK1 active site (Fig. 2B). The interruption of this interaction is expected to reduce the PLK1 binding affinity of BI-2536 but render its activity to BRD4-BD1 unchanged. Third, modification on ethyl group might have more impact on PLK1 activity as the pocket in PLK1 is narrower compared to the binding pocket in BRD4-BD1. In the end, we will modify the solvent exposed sites (Sites 4 and 5) to fully investigate the effect of each group on the binding affinity of the parent compound toward BRD4-BD1 and PLK1 based on the chemistry we established previously developing fluorescently-tagged kinase inhibitors.²⁰ Thus, we conducted a medicinal chemistry campaign around the BI-2536 core to develop small molecules with fine-tuned activities against PLK1 and BET bromodomains.

Figure. 3.

Available modification sites on BI-2536

Results and Discussion

Utilizing the structural information from the co-crystal structures of BI-2535 with BRD4-BD1 and PLK1, a series of BI-2536 analogs were synthesized using the general synthetic route shown in Scheme 1 (in Experimental Section). In brief, amino acids with varying alkyl side chains (different R¹ group) were esterified with thionyl chloride in MeOH followed by alkylation using reductive amination with cyclic or acyclic ketones. The resulting secondary amine derivative E1 underwent a regioselective nucleophilic aromatic substitution reaction with 2,6-dichloro-5-nitropyrimidine to produce tertiary anilines. The nitro group was reduced using zinc dust in hot acetic acid generating the dihydropteridinone ring in E2. Methylation of the amide produced the scaffold E3, which was used in a nucleophilic substitution reaction with the aminobenzoic acid derivative and led to the formation of intermediate E4 with an acid motif, which was coupled with different amines to produce the final BI-2536 analogs. As planned in Figure 3, the resulting analogs include derivatives with different substituent groups on the amide bond (Site 1); different ethers at Site 2 with cyclic or acyclic motifs; varying substituent groups, such as hydrogen, methyl, propyl and isobutyl on Site 3; and the solvent exposed Site 4 or 5. All compounds in this focused library were then screened for their inhibitory activity against BRD4-BD1 and BRDT-BD1 using AlphaScreen assay technology to obtain IC₅₀ values (all IC₅₀ values are for BD1 of all of the bromodomains tested unless otherwise noted).³²

Scheme 1.

Synthesis of BI-2536 analogs 1–20

The modification of the cyclopentyl moiety at Site 4 with different cyclic and acyclic groups, cyclohexyl (Compound 1) and isopropyl (Compound 2), created compounds with nanomolar inhibitory activity against BRD4 (Table 1). These analogs, however, were less potent than the parent compound BI-2536, indicating that the cyclopentyl group is crucial for the potent inhibitory activity against BRD4. These compounds, however, maintained low nanomolar inhibitory activity against PLK1 suggesting that modification on Site 4 is tolerated for PLK1 binding. On the other hand, unmethylated amide derivatives of Compounds 1 and 2 (Compound 3 and 4, respectively) resulted in more than 1000-fold decrease in binding affinity for BRD4. Interestingly, the inhibitors with unmethylated amide, Compounds 3 and 4, displayed more than 250-fold selectivity for PLK1 over BRD4, indicating the importance of the methylated amide as an N-acetylated lysine mimic for BRD4 binding. Thus, it is critical to have a methylated amide at Site 1 to serve as N-acetylated lysine mimic for bromodomain inhibition. Taken together, the replacement of cyclopentyl group at site 4 together with an unmethylated amide reduce the BRD4 activity of the inhibitors while maintaining PLK1 selectivity.

Table 1.

Structure Activity Relationships on BI-2536 analogs

				IC50 (uM)
Compound	R4	R3	R1	aBRD4-BD1	bPLK1	aBRDT-BD1
BI-2536	Cp	(R)-Et	Me	0.205±0.008	0.004	ND
1	Cy	(R)-Et	Me	0.65±0.024	0.004	ND
2	i-Pr	(R)-Et	Me	0.186±0.013	0.004	ND
3	i-Pr	(R)-Et	H	>10	2.62	ND
4	Cy	(R)-Et	H	>10	0.0372	ND
5	Cp	(R)-Me	Me	0.084±0.006	0.0011	0.342±0.015
6	Cp	(R)-Pr	Me	2.579±0.180	0.0099	8.468±0.076
7	Cp	H	Me	6.633±0.530	0.0541	25.99±9.70
8	Cp	(R)-iBu	Me	18.8±3.4	0.078	ND
9	Cp	(S)-Me	Me	0.107±0.002	0.0087	0.229±0.010
10	Cp	(S)-Et	Me	0.489±0.017	0.0325	3.513±0.351
11	Cy	(R)-Me	Me	0.31±0.01	0.0086	0.75±0.03
12	i-Pr	(R)-Me	Me	0.421±0.016	0.0252	0.978±0.043
13	tetrahydro(2H)pyran	(R)-iBu	Me	>50	0.322	>50
14	i-Pr	(R)-iBu	Me	41.9	0.218	ND
(S)-JQ1				0.032±0.001	-	0.103±0.004

^aIC₅₀ values were measured by AlphaScreen^TM binding assay and reported as the average of 3 replicates ± SE.

^bIC₅₀ values were measured using Adapta assay format (ThermoFisher Scientific)

Next, we examined the modification of Site 3 with methyl (Compound 5), propyl (Compound 6), hydrogen (compound 7) and isobutyl (Compound 8) substituents. The analogue with methyl substitution at the asymmetric dihydropteridinone carbon atom, Compound 5, exhibited good inhibitory activity against both BRD4 (IC₅₀ = 84 nM) and PLK1 (IC₅₀ = 11 nM). Selectivity profiling against 32 bromodomains (BROMOscan®) and 468 kinases (KINOMEscan®)³³ confirmed selective activity for the BET subfamily and PLK kinases (PLK1–3) at 1 μM through profiling platform at DiscoverX Specifically, 1 uM Compound 5 exhibited selective activity against BRD4-BD1 (12% of control) and PLK kinases (0.3–0.5% of control) (Sup Inf Fig. 1). No significant difference in binding activity was observed for the (S) and (R) isomers of the methyl substituent, which is similar to the observation made on the (S) and (R) isomers of BI-2536 by the Fletcher group.³¹ However, replacing the asymmetric ethyl group with bulky alkyl groups like propyl (Compound 6) and isobutyl (Compound 8) resulted in a significant decrease in binding affinity towards BRD4. While Compound 6 displayed excellent inhibitory activity against PLK1, with more than 250-fold selectivity over BRD4, further increasing the size of Site 3 to isobutyl, or deleting the methyl group resulted in a reduction of PLK1 inhibitory activity.

With the improvement in BRD4 inhibitory activity with Compound 5, we further examined this compound through the replacement of the cyclopentyl group with cyclohexyl (Compound 11), tetrahydro-2H-pyran (Compound 13) and isopropyl (Compound 12) groups. None of these substitutions improved potency towards BRD4 above that seen for Compound 5, further indicating that substituting the cyclopentyl moiety for the methyl group at the dihydropteridinone ring is optimal for BRD4 inhibition. On the other hand, the inhibitory activity of those analogs against PLK1 remained in the low nanomolar range. Similarly replacing the cyclopentyl group of Compound 8 with tetrahydro-2H-pyran (Compound 13) and isopropyl (Compound 14) groups did not improve potency towards BRD4 but showed modest inhibitory activity against PLK1 suggesting that PLK1 is tolerant of substitutions at the asymmetric carbon and anilide nitrogen (Table 1). Taken together, the adjustment of substituent groups at the Site 3 tunes down BRD4 activity with little impact on PLK1 activity, leading us to discover the most PLK1 selective inhibitor in the series, Compound 6 (> 260 fold).

To increase the BRD4 selectivity of BI-2536 analogs, we utilized Compound 5 to further explore the SAR around the terminal amine (Site 5) and the methoxy group at Site 2 to investigate their impact on potency and selectivity (Table 2). Replacement of the terminal cyclic amine (amino-piperidine, Site 5) with different alkyl and aromatic amines (Compounds 15, 16, 17, 18, 19 and 20) did not grossly impact the potency and selectivity for BRD4, implying that the various amine replacements are not engaged in any significant interaction with the amino acid residues in the binding pocket of BRD4. On the other hand, the modification of methoxy moiety (Site 2) with a cyclopentoxyl group improved potency toward bromodomains while reducing the potency against PLK1. These substitutions created a molecule that has relatively balanced potency toward both PLK1 and BRD4 with less than 2-fold selectivity difference (Table 2). Notably, Compound 23 exhibited improved potency for BRD4 compared to our reference BET inhibitor, (+)-JQ1, but reduced activity against PLK1 compared to BI-2536, making it one of the most potent BET bromodomain inhibitors with reduced PLK1 inhibition in the series. This improvement can be overridden by introducing a larger propyl group (Compound 24) at the Site 1. Also, the (S)-isomer of Compound 23, Compound 21, retains good inhibitory activity against BRD4, but is less potent toward PLK1 suggesting that the chirality change has more impact on PLK1 activity than BRD4 activity, which is consistent with the SAR we established in the previous round. Finally, we examined the three compounds with the most balanced BRD4 and PLK1 activity (Compounds 5, 6, and 23) in a human acute myeloid leukemia (AML) cell line, MOLM13. In general, BRD4 bromodomain inhibition causes a G1 arrest in the cell cycle as demonstrated by JQ1, while dual PLK1/BRD4 inhibitors, such as BI-2536, cause a G1/G2 arrest in the cell cycle. Twenty-four hour treatment with Compound 23 caused cell cycle arrest in the G1/G2 stage of the cell cycle, suggesting that it is functionally inhibiting both BRD4 and PLK1 (Sup Inf Fig. 2).

Table 2.

Structure Activity Relationship on compound 5 analogs

				IC50 (uM)
Compound	R3	R2	R5	aBRD4-BD1	bPLK1	aBRDT-BD1
5	(R)-Me	OMe	4-amine, 1-methylpiperidine	0.084±0.006	0.0011	0.342±0.015
15	(R)-Me	OMe	hexamethylene amine	0.107±0.003	0.16	0.378±0.014
16	(R)-Me	OMe	aminomethylpyridine	0.095±0.002	0.0084	0.289±0.008
17	(R)-Me	OMe	phenylethylamine	0.183±0.009	0.027	0.558±0.015
18	(R)-Me	OMe	N,N-dimethylethylamine	0.104±0.050	0.021	0.392±0.013
19	(R)-Me	OMe	aminopiperidine	0.109±0.003	0.009	0.56±0.025
20	(R)-Me	OMe	piperidine	0.094±0.002	0.095	0.369±0.011
21	(S)-Me	O-Cp	4-amine, 1-methylpiperidine	0.059±0.001	0.127	0.245±0.009
22	(R)-Me	O-Cp	OMe	0.253±0.007	0.457	0.933±0.032
23	(R)-Me	O-Cp	4-amine, 1-methylpiperidine	0.028±0.001	0.04	0.102±0.002
24	(R)-Pr	O-Cp	4-amine, 1-methylpiperidine	0.596±0.014	0.047	1.951±0.080
25	H	O-Cp	4-amine, 1-methylpiperidine	0.69±0.018	-	3.543±0.198

^aIC₅₀ values were measured by AlphaScreen^TM binding assay and reported as the average of 3 replicates ± SE.

^bIC₅₀ values were measured using Adapta assay format (ThermoFisher Scientific)

To gain a better understanding of the potency and predict the binding modes of our BI-2536 analogs, we utilized molecular modeling with PLK1 and BRD4-BD1(Table 3). As described above, Compound 23 shows equivalent potency for both BRD4 and PLK1, while Compound 6 demonstrates the most selectivity for PLK1 over BRD4 in the series. To gain a better understanding of how Compound 23 interacts with BRD4, we further modeled the docking of 23 with BRD4-BD1. As shown in Fig. 4B, all of the interactions between residues on BRD4 and BI-2536 are conserved when 23 resides in the active site of BRD4. The conserved phenyl ring in the center of Compound 23 develops π−π stacking interactions with the aromatic rings in the WPF shelf (W81, P82, F83) of BRD4 as seen with the parental compound BI-2536. Furthermore, almost all of the original binding patterns established in our co-crystallography study of BI-2536 and BRD4 (Figure 2) are strengthened as indicated by the van der Waals (vdw) interaction energy decomposition, where stronger interactions formed with P82, Q85, V87, C136, N140, I146 and Compound 23 (Fig. 4A). Although the L92, L94 and Y97 hydrophobic sub-pocket of BRD4 is moved further from the dihydropteridinone, the deep pocket formed by F83, I146 and M149 made closer contact with the newly embedded cyclopentane to compensate or even gain back the hydrophobic interaction loss (Fig. 4B). Overall, the docking analysis of Compound 23 with BRD4 is consistent with the binding affinity defined in our biochemical assays.

Table 3.

Docking results of some BI-2536 analogs on BRD4 and PLK1 (Details in SI)

Molecule	Docking score (Kcal/Mol)		IC50 (uM)		Ligand-receptor interactions (vdw energy < −4 kcal/mol)
	BRD4	PLK1	BRD4	PLK1	BRD4	PLK1
23	−7.34	−8.46	0.028	0.0400	Pro82, Gln85, Val87, Ile146	Leu59, Cys133, Arg136, Phe183
6	−6.54	−8.68	2.579	0.0099	Trp81, Pro82, Gln85, Val87, Leu92	Leu59, Arg136, Phe183
8	−6.63	−7.62	18.80	0.0780	Gln85, Val87, Asp88	Gly63, Asp194
5	−6.50	−8.23	0.084	0.0011	Val87, Asp88, Ile146	Leu59, Arg136, Phe183
BI-2536	−7.02	−8.60	0.025	0.00083	Leu92, Ile146	Arg57, Leu59, Arg136, Phe183

Figure 4.

A. Comparison of the per-residue decomposition of two BRD4-BD1-ligand 23 in blue, BI-2536 in red) intermolecular vdw interactions as a function of receptor sequence. B. Docking of 23 (cyan) with BRD4-BD1 (PDBID: 4O74), protein residues in contact with ligand shown in yellow sticks, the rest in gray.

In modeling the docking between Compound 23 and PLK1, we found that the C133 backbone amide nitrogen of PLK1 conserves its hydrogen bonding to the nitrogen of dihydropteridinone in 23, as indicated by unchanged dips at residue C133 from the vdw energy decomposition. The π-π stacking interaction between F83 and the dihydropteridinone in 23 is also conserved (Fig. 5A). The resulting docking pose of 23 with PLK1 retains similar interactions as BI-2536 with PLK1 (Fig. 5B). In addition,the methyl group in 23, which replaced the original ethyl group in BI-2536, does not reduce interactions with A80 and K82 significantly. However, a notable difference upon the substitution of cyclopentane for the methyl group in 23, is that the cyclopentane extends deeper into the hydrophobic sub-pocket between R57, F58 and L59 and creates additional interaction with R136, E140 that could further stabilize the overall binding.

Figure 5.

A. Comparison of the per-residue decomposition of two PLK1-ligand 23 in blue, BI-2536 in red) intermolecular vdw interactions as a function of receptor sequence. B. Docking of 23 (cyan) in PLK1 (PDBID: 2RKU), protein residues in contact with ligand shown in yellow sticks, the rest in gray, hydrogen bonding in red dashed line.

Molecular modeling of Compound 6 with BRD4-BD1 demonstrates that its docking pose demonstrates drastic variation from the co-crystal structure of BRD4-BD1 bound to BI-2536. When BI-2536 binds to BRD4-BD1, the dihydropteridinone and cyclopentane rings rest in between the WPF shelf (W81, P82, F83) and the BC loop (particularly N140 through bridging water) of BRD4-BD1, with the ethyl group of BI-2536 pointing away to make additional contacts with V87, L92, L94 and Y97 (sharing the bridging water with N140) (Figs. 6A & 6B). With BI-2536, Q85 interacts with the inhibitor through water-mediated hydrogen bonding effect. However, the longer alkyl chain of Compound 6 breaks this interaction due to steric clashes with L94 and Y97 (Fig. 6C). Therefore, Compound 6 has to adapt a novel orientation in the BRD4-BD1 active site, which demonstrates much weakened interactions by a striking positive vdw energy contribution from Y97. Although the phenyl ring of Compound 6 shifts into the pocket thereby enhancing binding with the WPF shelf, it still cannot overcome the repulsive effect of the positive vdw effect. We believe that these adjustments contributed significantly toward the reduced activity of 6 against BRD4-BD1.

Figure 6.

A. Comparison of the per-residue decomposition of two BRD4-BD1-ligand 6 in gray, BI-2536 in dark red intermolecular vdw interactions as a function of receptor sequence. B. Docking of 6 (gray) in BRD4 (1) (PDBID: 4O74), protein residues in contact with ligand shown in yellow sticks, the rest in gray). C. Steric clashes caused by the propyl group if 6 docked to BRD4(1) in the same pose as BI-2536 does (bumps shown in red dices).

Alternatively, the molecular docking of Compound 6 with PLK1 shows great similarity with BI-2536 co-crystalized with PLK1. The residue C133 conserves as hydrogen bonding receptor to the nitrogen in the dihydropteridinone in 6, and the π-π stacking interaction between F83 and the same aromatic ring (Figs. A & 7B). Furthermore, the propyl group in 6, which replaced the original ethyl group in BI-2536 at site 3, does not reduce interactions with A80 and K82 much. The docking pose suggests that the carbonyl oxygen moved away from the active site, but may form hydrogen bonds with either R136, R57 or crystal water, as was observed in the BI-2536-PLK1 complex structure, to reduce the entropy penalty when the new complex is forming. Again, the molecular docking analysis is in agreement with the binding affinity assays that illustrate a potent interaction between PLK1 and 6. The detailed ligand interaction diagrams (LID) are listed in Supplement Figs 3 & 4.

Figure 7.

A. Comparison of the per-residue decomposition of two PLK1-ligand 6 in gray, BI-2536 in dark red) intermolecular vdw interactions as a function of receptor sequence. B. Docking of 6 (gray) in PLK1 (PDBID: 2RKU), protein residues in contact with ligand shown in yellow sticks, the rest in gray, hydrogen bonding in red dashed line.

CONCLUSION

Our focused SAR study of the dual kinase-bromodomain inhibitor BI-2536 has led to the identification of potent, balanced inhibitors of BRD4 and PLK1, as well as a more PLK1 selective inhibitor. Through these studies, we determined the requirements for both PLK1 and BET protein binding. We determined that the methylated amide and the cyclopentyl group of the BI-2536 scaffold is important for maintaining bromodomain binding activity but less important for PLK1 inhibitory activity. Moreover, substituting the asymmetric ethyl moiety at the dihydropteridine ring with a methyl group improved the inhibitory activity of resulting compounds against both BRD4 and PLK1, as seen with Compound 5. Installing bulky alkyl groups at the asymmetric carbon resulted in inhibitors with potent and selective PLK1 activity but largely limited BRD4 activity, as seen with Compound 6. Further modification on the methoxy group of Compound 5 led to the identification of Compound 23, a compound with increased potency for BRD4 with much reduced PLK1 activity. Our molecular docking studies suggested possible binding orientations with BRD4-BD1 and PLK1 and the binding energies were in agreement with the observed SAR established through our medicinal chemistry exercise.

In summary, we have demonstrated that fine-tuning the selectivity for inhibitors demonstrating polypharmacology can be achieved through structural guidance. We initiated an SAR campaign of BI-2536 in order to alter the PLK1 and BRD4 potency/selectivity of the compound and confirmed our biochemical findings through molecular modeling. In this study, we discovered a set of compounds with balanced selectivity against BRD4 and PLK1 that will entitle us to further explore the biological outcome of dual BRD4/PLK1 inhibition. In particular, Compound 23 is a well-balanced BRD4 and PLK1 dual inhibitor when assayed biochemically and in cells. 23 demonstrates increased potency for BRD4 and decreased potency for PLK1 as compared to the parent molecule BI-2536. We envision that this set of molecules will generate sufficient PLK1 and BRD4 inhibition in order to study simultaneous targeting of PLK1 and BRD4 in cancer.

EXPERIMENTAL SECTION

General Synthetic Information:

All chemicals and solvents were purchased from commercial suppliers and used as received. ¹H NMR (300 MHz) and ¹³C NMR (75 MHz) spectra were recorded on a Varian NMR spectrometer, and ¹H NMR (400 MHz) and ¹³C NMR spectra (101 MHz) were recorded on Agilent NMR spectrometers. LC-MS were performed on an Agilent 2100 LC with a 6130 quadrupole MS spectrometers. Accurate mass measurements were performed on a Fusion Lumos Orbitrap mass spectrometer (Thermo Fisher Scientific) at a mass resolution setting of 500000 immediately after calibration or under lock mass calibration using solvent ions. A C₁₈ column (5.0 μm, 6.0 × 50 mm) was used for separation. The mobile phases were MeOH and H₂O both containing 0.05% CF₃CO₂H. A linear gradient from 25:75 (v/v) MeOH/ H₂O to 100% MeOH over 7.0 min at a flow rate of 0.7 mL/min was used as a mobile phase. UV detections were conducted at 210 nm, 254 nm and 365 nm. Low resolution mass spectra were recorded by APCI (atmospheric pressure chemical ionization) unless otherwise specified. Flash chromatography separation was performed on YAMAZEN AI-580 system with Agela silica gel (12 or 20 g, 230–400 mesh) cartridges. Final products were purified on Angela HP-100 preparative-LC system with a Venusil PrepG-C₁₈ column (10 μm, 120 Å, 21.2 mm x 250 mm).

Synthesis of BI-2536 analogs.

The synthesis of Compounds 1–20 shown in Tables 1 and 2 was carried out following the reported procedures (Scheme 1).³³ The synthesis of analogs of Compound 5 including Compounds 21, 23–25 listed in Table 2 was performed following the route shown in Scheme 2. Bromocyclopentane (1.3 equiv) was added slowly to a stirred suspension of potassium carbonate (1.5 equiv) in DMF containing methyl- 3-hydroxy-4-nitrobenzoate (1 equiv) at 65 ^oC for 4 h, then the product was treated by SnCl₂·2H₂O in EtOAc at 50 ^oC overnight. A mixture of Ar-Cl (1 equiv), X-Phos (28 mol %), Pd₂(dba)₃ (7 mol %), methyl 4-amino-3-(cyclopentyloxy)benzoate (1.3 equiv), and K₂CO₃ (5 equiv) in t-BuOH was heated at 100 ^oC in a sealed-tube for 8 h. The mixture was then diluted with EtOAc and washed three times with aqueous NaHCO₃. The organic layer was dried over anhydrous Na₂SO₄, filtered and condensed. The crude product was purified by preparative-LC system to get the final product.

Scheme 2.

Synthesis of compounds 21 and 23–25

(R)-4-((8-cyclopentyl-5,7-dimethyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 5). ¹H NMR (300 MHz, CD₃OD) δ 8.41 (1H, d, J = 9.0Hz, 7.72 (1H, s), 7.41 – 7.31 (3H, m), 4.50 – 4.45 (1H, m), 4.31 – 4.24 (1H, q, J = 6.7 Hz), 4.02 – 3.82 (6H, s, N-CH₃ and O-CH₃), 3.27 – 3.26 (6H, m), 3.22 (3H, s), 2.57 – 2.47 (6H, m), 2.06 – 1.86 (4H, m), 1.25 (3H, d, J = 6.6 Hz); ¹³C NMR (75 MHz, CD₃OD) δ 170.9, 166.6, 156.3, 153.0, 139.44, 138.7, 127.3, 121.2, 117.3, 117.3, 109.9, 59.5, 59.3, 56.4, 55.6, 55.02, 47.1, 44.9, 31.2, 30.5, 29.9, 28.5, 24.3, 23.8, 19.0; HRMS (ESI) m/z: (M+H)⁺ calcd for C₂₇H₃₇N₇O₃ 536.3349, found 508.3031 (M⁺ +1).

(R)-4-((8-cyclopentyl-5-methyl-6-oxo-7-propyl-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 6). ¹H NMR (400 MHz, CDCl₃) δ 8.54 (1H, d, J = 8.4 Hz), 7.68 (1H, s), 7.59 (1H, s), 7.51 (1H, s), 7.42 (1H, s), 7.42 (1H, d, J = 1.8 Hz), 4.56 – 4.46 (1H, m), 4.24 (1H, dd, J = 8.0, 3.7Hz), 3.97 (3H, s), 3.31 (3H, s), 2.86 – 2.84 (2H, m), 2.31 (3H, s), 2.25 – 2.10 (4H, m), 2.07 – 2.04 (4H, m), 1.81 – 1.71 (14H, m), 1.37 – 1.21(4H, m), 0.87 (3H, t, J = 7.3 Hz); ¹³C NMR (101 MHz, CDCl₃) δ 163.9, 155.0, 152.1, 147.2, 137.9, 133.2, 126.2, 126.0, 118.8, 116.4, 115.8, 109.0, 58.8, 58.1, 54.4, 35.8, 32.4, 29.6, 29.1, 28.2, 23.5, 23.0, 17.8, 13.9; HRMS (ESI) m/z: (M+H)⁺ calcd for C₂₉H₄₁N₇O₃ 536.3349, found 536.3342 (M⁺ +1).

(R)-4-((8-cyclopentyl-7-isobutyl-5-methyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 8). ¹H NMR (300 MHz, CD₃OD) δ 8.40 (1H, d, J = 9.0 Hz), 7.74 (1H, s), 7.42 – 7.38 (3H, m), 4.21 – 4.17 (1H, m), 3.91 (3H, s), 3.87 – 3.83 (1H, m), 3.23 (3H, s), 2.91 – 2.88 (1H, m), 2.27 (3H, s), 2.21 – 2.13 (2H, m), 2.02 – 1. 97 (2H, m), 1.92 – 1.83 (2H, m), 1.74 – 1.54 (8H, m), 1.41 – 1.36 (1H, m), 0.92 (3H, d, J = 6.3 Hz), 0.78 (3H, d, J = 6.6 Hz); ¹³C NMR (75 MHz, CD₃OD) δ 169.1, 165.2, 156.2, 153.6, 148.4, 139.6, 134.0, 127.4, 121.2, 117.4, 117.3, 109.9, 59.6, 58.4, 56.4, 55.5, 45.8, 42.5, 32.0, 30.49, 30.4, 28.6, 25.7, 24.2, 23.9, 23.7, 21.7; HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₀H₄₃N₇O₃ 550.3506, found 550.3501 (M⁺ +1).

(R)-4-((8-isopropyl-5,7-dimethyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 12). ¹H NMR (300 MHz, CD₃OD) δ 8.42 (1H, d, J = 9.3 Hz), 7.69 (1H, s), 7.38 (3H, m), 3.99 – 3.90 (5H, m), 3.90 (3H, s), 2.72 – 2.65 (4H, m), 2.55 (3H, s), 1.86–1.76 (4H, m), 1.30 – 1.23 (9H, m); ¹³C NMR (75 MHz, CD₃OD) δ 169.2, 166.4, 156.3, 152.4, 148.3, 139.5, 134.1, 127.1, 121.3, 117.2, 109.8, 56.4, 54.7, 53.6, 46.6, 44.4, 30.7, 28.5, 21.5, 19.9; HRMS (ESI) m/z: (M+H)⁺ calcd for C₂₅H₃₅N₇O₃ 482.2880, found 482.2884 (M⁺ +1).

(R)-4-((7-isobutyl-5-methyl-6-oxo-8-(tetrahydro-2H-pyran-4-yl)-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 13). ¹H NMR (300 MHz, CD₃OD) δ 8.42 (1H, d, J = 9.0 Hz), 7.78 (1H, s), 7.45 – 7.42 (3H, m), 3.92 (3H, s), 3.56 – 3. 53 (2H, m), 3.22 (3H, s), 3.03 – 2.99 (2H, m), 2.37 (3H, m), 1.96 – 1.92 (3H, m), 1.70 – 1.65 (4H, m), 1.55 – 1. 38 (5H, m), 0.94 (3H, d, J = 6.3 Hz), 0.79 (3H, d, J = 6.6 Hz); ¹³C NMR (75 MHz, CD₃OD) δ 171.5, 171.4, 170.9, 169.1, 164.86, 153.0, 150.0, 140.1, 127.5, 121.2, 117.4, 110.0, 68.6, 68.5, 57.3, 56.4, 55.3, 55.2, 45.4, 43.4, 32.6, 31.6, 28.7, 25.8, 23.9, 21.7; HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₀H₄₃N₇O₄ 566.3455, found 566.3448 (M⁺ +1).

(S)-4-((8-cyclopentyl-5,7-dimethyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-methoxy-N-(1-methylpiperidin-4-yl)benzamide (Compound 21). ¹H NMR (300 MHz, CD₃OD) δ 8.43 (1H, d, J = 8.7 Hz), 7.75 (1H, s), 7.43 – 7.41 (3H, m), 4.31 – 4.29 (1H, q, J = 6.0 Hz), 3.94 (3H, s), 3.24 (3H, s), 3.12–3.07 (3H, m), 2.90 – 2.87 (3H, m), 2.12 – 2.09 (3H, m), 1.91 – 1.89 (4H, m), 1.79 – 1.67 (6H, m), 1.26 (3H, d, J = 6.9 Hz); ¹³C NMR (75 MHz, CD₃OD) δ 171.1, 169.1, 166.6, 156.2, 153.0, 148.3, 139.4, 134.0, 127.4, 121.2, 117.2, 109.9, 59.3, 56.4, 55.6, 55.4, 47.8, 45.7, 31.9, 30.5, 29.9, 28.5, 24.2, 23.8, 19.0; HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₁H₄₃N₇O₃ 562.3506, found 562.3501 (M⁺ +1).

(R)-4-((8-cyclopentyl-5,7-dimethyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-(cyclopentyloxy)-N-(1-methylpiperidin-4-yl)benzamide (Compound 23). ¹H NMR (400 MHz, CDCl₃) δ 8.53 (1H, d, J = 8.0 Hz), 7.71 (1H, s), 7.60 (1H, s), 7.41 (1H, d, J = 1.9 Hz), 4.62 – 4.51 (1H, m), 4.32 – 4.30 (1H, m), 4.14 – 4.06 (1H, m), 3.97(3H, s), 3.31 (3H, s), 2.83 – 2.81 (3H, m), 2. 29 ( 3H, s), 2.20 – 2.10 (3H, m), 2.08 – 1.93 (4H, m), 1.32 (3H, d, J = 6.8 Hz); ¹³C NMR (101 MHz, CDCl₃) δ 166.6, 165.0, 155.2, 151.8, 147.2, 138.3, 133.1, 126.6, 118.8, 116.3, 116.0, 109.1, 57.6, 55.9, 54.4, 46.2, 32.5, 30.0, 29.5, 28.3, 23.7, 23.2, 19.0. HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₁H₄₃N₇O₃ 562.3506, found 562.3508 (M⁺ +1).

(R)-4-((8-cyclopentyl-5-methyl-6-oxo-7-propyl-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-(cyclopentyloxy)-N-(1-methylpiperidin-4-yl)benzamide (Compound 24). ¹H NMR (300 MHz, CD₃OD) δ 8.44 (1H, d, J= 8.1 Hz), 7.72 (1H, s), 7.42 – 7.39 (3H, m), 4.39 – 4.34 (1H, m), 4.27– 4.24 (1H, m), 3.85 – 3.80 (1H, m), 3.25 (3H, s), 2.86 – 2.60 (2H, m), 2.26 (3H, s), 2.16 – 2.09 (3H, m), 2.01 – 1.80 (14H, m), 1.76 – 1.59 ( 8H, m), 1.26 – 1.17 (2H, m), 0.83 (3H, t, J = 7.2 Hz); ¹³C NMR (75 MHz, CD₃OD) δ 169.12, 165.4, 156.0, 153.4, 146.5, 139.0, 134.8, 127.3, 120.8, 117.2, 112.4, 81.8, 60.67 60.5, 55.6, 49.7, 46.0, 36.8, 33.6, 32.2, 30.2, 29.9, 28.4, 24.8, 24.4, 24.2, 18.7, 14.0; HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₃H₄₇N₇O₃ 590.3819, found 590.3815 (M⁺ +1).

4-((8-cyclopentyl-5-methyl-6-oxo-5,6,7,8-tetrahydropteridin-2-yl)amino)-3-(cyclopentyloxy)-N-(1-methylpiperidin-4-yl)benzamide (Compound 25).¹H NMR (300 MHz, CD₃OD) δ 8.43 (1H, d, J = 8.4 Hz), 7.64 (1H, s), 7.42 –7.39 (3H, m), 4.06 (3H, s), 3.28 – 3.25 (4H, m), 3.23 (3H, s), 3.02 – 2.97 (3H, m), 2.37 (3H, s), 2.30 – 2.26 (3H, m), 1.98 – 1.93 (7H, m), 1.74 – 1.68 (9H, m); ¹³C NMR (75 MHz, CD₃OD) δ 169.1, 163.5, 156.1, 153.1, 146.4, 138.5, 134.8, 127.1, 120.8, 117.0, 116.8, 112.3, 81.8, 56.9, 55.37, 47.6, 45.9, 45.5, 33.6, 31.7, 28.0, 27.9, 24.8, 24.8; HRMS (ESI) m/z: (M+H)⁺ calcd for C₃₀H₄₁N₇O₃ 548.3350, found 548.3346 (M⁺ +1).

Molecular Docking Software:

For protein preparation: UCSF Chimera Dock Prep tool,³⁴ UCSF DOCK6.³⁵ For compound structures and charge assignment method: ChemDraw 15.0, Chem3D 14.0, GAFF, AM1-BCC. For docking and energy decomposition analysis: UCSF DOCK6 suite. Dock poses, visualization and illustration: UCSF Chimera ViewDock, PyMOL.

Compounds and protein preparation:

The synthesized compounds were built in ChemDraw, and then their Cartesian coordinates’ construction and energy minimization using MM2 force field^36,37 were generated in Chem3D. The crystal structures of BI-2536 binding with BRD4-BD1 (PDBID: 4O74) and PLK1 (PDBID: 2RKU) were downloaded from Protein Data Bank (PDB). Protein targets were prepared using Chimera DockPrep tool, in which AMBER force field ff14SB³⁸ was employed to provide accurate descriptions for proteins. GAFF³⁹ and AM1-BCC⁴⁰ were used to generate parameters and assign charges for all inhibitor molecules.

Docking and energy decomposition analysis:

DOCK6 suite was used for docking compounds to their protein targets. Flexible ligand docking was performed under the guidance of the anchor-and-grow algorithm. Dock Score was used as the criterion for ranking the binding poses. We visually examined the top 10 ranked poses, and the highest ranked poses are reported in Table 3. In the case of PLK1 and BRD4-BD1, DOCK6 successfully sampled and top-scored the BI-2536 docking poses to reproduce the corresponding crystal binding poses with rmsd < 2 Å. DOCK Score is composed of Van der Waals (vdw) and electrostatic interactions. The vdw composition of Dock score was further decomposed onto every residue of its binder protein by using DOCK write_footprint function, which generates both the vdw and electrostatic parts. Only the vdw composition was noticeably diversified and thus referred to represent the contribution of the protein target per residue in protein-ligand interaction energy profiles as shown in Figures 4–7.

Specific favorable and unfavorable interactions recognition:

As listed in Table 3, the specific residual information of the protein receptors interacting with BI-2536 analogs was recognized by referring to vdw energy and by visualization of binding poses. A threshold of < −4 kcal/mol in decomposed vdw energy was used to include the first shell of residues around the molecules (see also Sup Inf Fig S-1). The positive peak observed in per-residue decomposition of BRD4-DB1 Compound 6, Figure 6A, was ascribed to the steric clashes of residue Y97, L94 with the propyl group on 6. The disc was scripted in PyMOL software to show vdw overlaps or steric clashing (See also Sup Inf Fig S-2).

BRD4-BD1 and BRDT-BD1 activity assays:

Compound potency was assessed by relative IC₅₀ potency determined by an AlphaScreen biotin-JQ1 competition assay as reported previously by our group.⁴¹ All reagents were diluted in 50 mM HEPES, 150 mM NaCl, 0.1% w/v BSA, 0.01% w/v Tween20, pH 7.5 and allowed to equilibrate to room temperature prior to the addition to plates. After addition of Alpha beads to master solutions, all subsequent steps were performed in low light conditions. A 2x solution of components with final concentrations of BRD at 40 nM, Ni-coated Acceptor Bead at 25 μg/mL, and 20 nM biotin-JQ1 was added in 10 μL to 384-well plates (AlphaPlate − 384, PerkinElmer, USA). After a 1 min 1000 rpm spin-down, 100 nL of compounds in DMSO from stock plates were added by pin transfer using a Janus Workstation (PerkinElmer, USA). The 2x, 10 μL streptavidin-coated donor beads (25 μg/ml) were added to the previous solution. Following this addition, the plates were sealed with foil to block light exposure and to prevent evaporation. The plates were spun down again at 1000 rpm for 1 min. Next, the plates were incubated at room temperature with the plate reader (for temperature equilibration) for 1 h prior to reading the assay. Signal is stable for up to 3 h after donor bead addition. AlphaScreen measurements were performed on an Envision 2104 (PerkinElmer, USA) utilizing the manufacturer’s protocol.

PLK1 activity assay:

The enzymatic activities against PLK1 were tested in Z-Lyte assays with ATP concentration of K_M for each kinase per manufacturer’s protocol (Thermo Fisher).

All the compounds were tested in parallel with cross-linker assay to rule out the assay binder/interrupter.

Selectivity profiling:

Compound 5 was assayed at 1 uM in DiscoverX’s BROMOscan® (bromoMAX^SM) and KINOMEscan® (scanMAX^SM) assay platforms. The results for single concentration binding interactions for Compound 5 are reported as % of control (DMSO).

Cell cycle study:

MOLM13 cells were treated with 1 μM of JQ1, BI2536, TAE684, Compound 5, Compound 6 or Compound 23 for 24 h. Cells were subsequently harvested and stained with propidium iodide (Calbiochem), followed by measurement of propidium iodide-mediated fluorescence with a Guava H6T (Millipore Sigma).