Structural investigation of a pyrano-1,3-oxazine derivative and the phenanthridinone core moiety against BRD2 bromodomains

BRD2 is a promising drug target for the treatment of various cancers. Crystal structures of BRD2 bromodomains BD1 and BD2 in complexes with a pyrano-1,3-oxazine derivative and phenanthridinone compounds were determined. Crystal structures and biochemical assays confirm that these ligands are indeed potent inhibitors of BRD2 bromodomains.

Abstract

The BET (bromodomain and extra-terminal) family of proteins recognize the acetylated histone code on chromatin and play important roles in transcriptional co-regulation. BRD2 and BRD4, which belong to the BET family, are promising drug targets for the management of chronic diseases. The discovery of new scaffold molecules, a pyrano-1,3-oxazine derivative (NSC 328111; NS5) and phenanthridinone-based derivatives (L10 and its core moiety L10a), as inhibitors of BRD2 bromodomains BD1 and BD2, respectively, has recently been reported. The compound NS5 has a significant inhibitory effect on BRD2 in glioblastoma. Here, the crystal structure of BRD2 BD2 in complex with NS5, refined to 2.0 Å resolution, is reported. Moreover, as the previously reported crystal structures of the BD1–NS5 complex and the BD2–L10a complex possess moderate electron density corresponding to the respective ligands, the crystal structures of these complexes were re-evaluated using new X-ray data. Together with biochemical studies using wild-type BRD2 BD1 and BD2 and various mutants, it is confirmed that the pyrano-1,3-oxazine and phenanthridinone derivatives are indeed potent inhibitors of BRD2 bromodomains.

1. Introduction

Post-translational modifications of histones are important cellular events in epigenetic regulation of gene expression by altering chromatin structure or recruiting chromatin regu­lators (Hecht et al., 1995 ▸; Edmondson et al., 1996 ▸). For example, the acetylation of lysines on the N-terminal tails of histones can disrupt histone–DNA interaction (Cheung et al., 2000 ▸), change the interactions between the histones of adjacent nucleosomes (Wolffe & Hayes, 1999 ▸) and modify the interaction between the histones and regulatory proteins (Hecht et al., 1995 ▸; Edmondson et al., 1996 ▸). The acetylation of lysine in histone and nonhistone proteins is orchestrated by three classes of proteins, namely the ‘writers’ (histone acetyl­transferases), the ‘readers’ (bromodomains) and the ‘erasers’ (histone deacetylases) (Berenguer-Daizé et al., 2016 ▸). Overall, based on structure-based sequence-comparison studies, 61 bromodomain (BD)-containing proteins have been grouped into eight subgroups (Barbieri et al., 2013 ▸).

The bromodomain and extra-terminal (BET) family of proteins, for example, BRD2, BRD3, BRD4 and the testis-explicit BRDT, bind to acetylated chromatin and recruit transcription regulatory complexes to regulate the gene expression responsible for cell multiplication and cell-cycle progression (Taniguchi, 2016 ▸). The BET family of proteins contain two tandem bromodomains (BD1 and BD2) and an extra-terminal (ET) domain in the C-terminal region (Zeng et al., 2005 ▸; Padmanabhan et al., 2016 ▸; Fig. 1 ▸ a; Supplementary Fig. S1). The BET proteins act as co-activators and co-repressors (Filippakopoulos et al., 2010 ▸; Belkina et al., 2013 ▸), and control the epigenetic regulation of the histone code in a context-dependent manner (Belkina & Denis, 2012 ▸). The BET family members BRD2 and BRD3 recognize monoacetylated and diacetylated histones (Umehara et al., 2010 ▸; Filippa­kopoulos & Knapp, 2012 ▸).

Figure 1.

The crystal structure of BRD2 BD2 in complex with NS5. (a) Schematic domain organization of BRD2. (b) A cartoon representation of the BD2–NS5 complex and the chemical structure of NS5. (c) A polder difference Fourier map of NS5, contoured at 2.7σ, at the binding site. (d) A closer view of the intermolecular interactions between BD2 and NS5. The interacting residues which contribute to hydrophilic interactions are Tyr386, Asn429 and Pro371. Water bridges are formed with Tyr386, Asn429 and Met394. The interacting residues and ligand are shown as sticks. Water molecules are shown as spheres. Hydrogen bonds are shown as broken lines. The figures were generated by PyMOL.

The past two decades of research have shown the significance of the BET family BDs as focuses for the treatment of various cancers, neurological disorders, obesity and inflammation (Chung & Tough, 2012 ▸). Dysfunction of BET proteins leads to uncontrolled cell proliferation, cell division and abnormal cell differentiation, which cause several forms of disease (Łukasik et al., 2021 ▸). The development and therapeutic use of BET family inhibitors are well established strategies in the treatment of various cancers. In the past few years, many highly selective and potent BET family inhibitors have been developed by global scientific communities. Among the BET inhibitors, TEN-010, I-BET 762, OTX 015, I-BET 151, CPI-0610, BAY 1238097 and RVX-208 are in clinical trials for various diseases (Ghoshal et al., 2016 ▸; Lu et al., 2020 ▸). Among known BET family inhibitors, only a few compounds, JQ1, OTX015 and I-BET151, have been shown to inhibit the growth of glioblastoma (GBM) cells (Berenguer-Daizé et al., 2016 ▸). The development of highly selective and specific inhibitors for BET family proteins is a vital strategy to hinder their specific functions for the management of disease conditions.

We have recently discovered two scaffold compounds that inhibit BRD2 bromodomains BD1 and BD2 (Fig. 1 ▸ a). A phenanthridinone-based derivative binds to BD2 of BRD2 (Tripathi et al., 2016 ▸; Mathur et al., 2018 ▸). A pyrano-1,3-oxazine derivative binds to BD1 of BRD2 and has been shown to inhibit the growth of GBM cells (Deshmukh et al., 2020 ▸).

In our reported crystal structures of the complex of BD1 with the pyrano-1,3-oxazine derivative NS5 and the complex of BD2 with the phenanthridinone core moiety L10a the electron density corresponding to these ligands was moderate, which may be due to interference of the crystallization reagents with these ligands at the acetyl–lysine binding site. After careful optimization of crystallization screening to obtain co-crystals, the X-ray diffraction data for these two derivatives yielded unambiguous electron density corresponding to these ligands.

Here, we report the crystal structure of BRD2 BD2 in complex with a pyrano-1,3-oxazine derivative (NSC 328111; hereafter referred to as NS5) as well as the redetermination of the crystal structures of BD1 in complex with NS5 and of BD2 in complex with 6(5H)-phenanthridinone (the core moiety of L10; hereafter referred to as L10a). Together with a quantitative binding assay using wild-type BRD2 bromodomains BD1 and BD2 and several mutants, we confirm that the discovered molecules NS5 and L10a are indeed potent inhibitors of BRD2 protein.

2. Materials and methods

2.1. Protein production of BRD2 bromodomains BD1 and BD2 and site-directed mutagenesis

The expression and protein production of BRD2 bromodomains BD1 (amino acids 73–194) and BD2 (amino acids 348–455) were performed as described previously (Mathur et al., 2018 ▸; Deshmukh et al., 2020 ▸). The point mutations for the BD1 and BD2 bromodomains of BRD2 were designed according to their protein–ligand interactions in the crystal structures of BD1 and BD2 complexes. Mutations such as W97A, P98A, N156A and I162A were introduced into BD1, whereas for BD2 an N429A mutation was introduced. The plasmids for BD1 and BD2 were amplified using the mutant primers by Phusion High-Fidelity DNA polymerase (New England Biolabs). The amplified product was subjected to Dpn1 treatment (Takara). The plasmid was isolated (QIAprep Spin Miniprep Kit, Qiagen) and the point mutations were confirmed by sequencing. The primer details for the mutations are shown in Table 1 ▸.

Table 1. Primers used for mutations.

BRD2BD1-W97A-FP	5′-GAAACATCAGTTCGCAGCGCCATTCCGGCAG-3′
BRD2BD1-W97A-RP	5′-CTGCCGGAATGGCGCTGCGAACTGATGTTTC-3′
BRD2BD1-P98A-FP	5′-CATCAGTTCGCATGGGCATTCCGGCAGCCTG-3′
BRD2BD1-P98A-RP	5′-CAGGCTGCCGGAATGCCCATGCGAACTGATG-3′
BRD2BD1-N156A-FP	5′-CCAACTGTTACATTTACGCCAAGCCCACTGATGATATTGTC-3′
BRD2BD1-N156A-RP	5′-GACAATATCATCAGTGGGCTTGGCGTAAATGTAACAG-3′
BRD2BD1-I162A-FP	5′-CAAGCCCACTGATGATGCTGTCCTAATGGCACAAAC-3′
BRD2BD1-I162A-RP	5′-GTTTGTGCCATTAGGACAGCATCATCAGTGGGCTTG-3′
BRD2BD2-N429A-FP	5′-CCAACTGCTATAAGTACGCTCCCCCAGATCACG-3′
BRD2BD2-N429A-RP	5′-CGTGATCTGGGGGAGCGTACTTATAGCAGTTGG-3′

2.2. Crystallization and soaking of BRD2 BD1–NS5, BRD2 BD2–NS5 and BRD2 BD2–L10a complexes

Crystallization trials for apo BRD2 bromodomains BD1 and BD2 were performed by the vapor-diffusion method at 20°C. Protein buffer components such as glycerol, β-mercaptoethanol and dithiothreitol were completely avoided during protein-purification steps. These buffer components bound to the proteins at the Kac binding site and were responsible for potential ligand-binding hindrance, as shown in Supplementary Fig. S2. Moreover, the solvent DMSO was also observed in the binding pocket of bromodomains. Therefore, soaking studies were performed with NS5 and L10a dissolved in methanol and water, respectively. Plate-like BD1 crystals were obtained in 0.8 M ammonium sulfate, 0.1 M MES pH 6.5, 15% PEG 3350, 5% glycerol in ten days with protein at a concentration of 5.5 mg ml⁻¹. To obtain the BD1–NS5 complex, BD1 crystals were soaked with NS5 in a 1:10 molar ratio for 6 h. On the other hand, the BD2 protein (5.5 mg ml⁻¹) and NS5 were incubated in a 1:12 molar ratio at 4°C overnight to obtain co-crystals. The BD2–NS5 mixture was then used to set up co-crystallization drops with 50 mM Tris pH 7.5, 50 mM NaCl, 25% PEG MME 2000 at 20°C. The BD2–NS5 co-crystals were obtained within two weeks. The BRD2 BD2 crystals were soaked with L10a in a 1:5 molar ratio overnight to obtain the BD2–L10a complex. Both the soaked crystals and the co-crystals were harvested and subjected to X-ray diffraction.

2.3. X-ray data collection, processing and structure determination

X-ray diffraction data were collected using an in-house X-ray diffractometer (Rigaku MicroMax-007, Rigaku Corporation, Tokyo, Japan) integrated with an HyPix6000 detector at NIMHANS. The diffraction data were processed and scaled using CrysAlis^Pro (Rigaku) and HKL-3000 (Minor et al., 2006 ▸). The BD1–NS5 complex belonged to space group C2, with unit-cell parameters a = 115.22, b = 55.89, c = 67.60 Å, β = 94.05°. The crystals of the BD2–NS5 and BD2–L10a complexes belonged to space group P2₁2₁2, with unit-cell parameters a = 52.10, b = 71.12, c = 31.85 Å, and space group P22₁2₁, with unit-cell parameters a = 32.02, b = 52.67, c = 71.66 Å, respectively. The crystal structures of the BD1 and BD2 complexes were solved by the molecular-replacement method using Phaser-MR incorporated in Phenix (Liebschner et al., 2019 ▸). For the structure determination of BD1 and BD2, the structures of PDB entries 4uyf (Gosmini et al., 2014 ▸) and 5xhk (Mathur et al., 2018 ▸), respectively, were used as search models. In the BD1–NS5 complex the asymmetric unit contains three molecules, whereas in the BD2 complexes the asymmetric unit contains one molecule. Clear electron density for the NS5 and L10a ligands was observed in the 2m|F _o| − D|F _c| and polder difference Fourier maps (Liebschner et al., 2017 ▸). The grade web server (http://grade.globalphasing.org) was used to generate the energy-minimized coordinates and crystal information files (CIFs) for the NS5 and L10a ligands.

The crystal structures of the BD1 and BD2 complexes were refined using phenix.refine from the Phenix package (Liebschner et al., 2019 ▸). Subsequently, model building was performed using Coot (Emsley et al., 2010 ▸). The final refinement yielded real-space correlation coefficient (RSCC) values of 0.83 for NS5 in the BD1–NS5 complex, 0.87 for NS5 in the BD2–NS5 complex and 0.94 for L10a in the BD2–L10a complex, confirming the binding of the ligands in the respective complexes. The X-ray data-collection, scaling and refinement statistics are summarized in Table 2 ▸. MolProbity (Chen et al., 2010 ▸) was used to assess the stereochemistry of the crystal structures. The structure factors and structural coordinates of the BD1 and BD2 complexes have been deposited in the PDB (PDB codes 7env for BD1–NS5, 7eo5 for BD2–NS5 and 7enz for BD2–L10a).

Table 2. Crystallographic data-collection and refinement statistics.

Values in parentheses are for the highest resolution shell.

	BRD2 BD1–NS5	BRD2 BD2–NS5	BRD2 BD2–L10a
Data-collection parameters
Space group	C2	P21212	P22121
a, b, c (Å)	115.22, 55.89, 67.60	52.10, 71.12, 31.85	32.02, 52.67, 71.66
α, β, γ (°)	90, 94.85, 90	90, 90, 90	90, 90, 90
Wavelength (Å)	1.54056	1.54056	1.54056
Resolution (Å)	50.0–2.45 (2.49–2.45)	50.0–1.72 (1.75–1.72)	14.14–1.70 (1.73–1.70)
Unique reflections	15805 (744)	13233 (610)	13875 (700)
R merge † (%)	13.5 (53.0)	7.8 (26.4)	13.1 (35.0)
〈I/σ(I)〉	10.82 (1.45)	23.02 (2.72)	7.6 (3.2)
Completeness (%)	99.1 (92.8)	98.5 (92.4)	99.7 (100)
Multiplicity	4.4 (2.3)	3.3 (2.2)	5.8 (5.4)
CC1/2	0.993 (0.593)	0.981 (0.86)	0.985 (0.874)
Refinement
Resolution (Å)	31.6–2.45	29.07–2.00	14.14–1.70
No. of reflections	15798	8314	13834
R work ‡/R free § (%)	18.7/25.9	17.6/22.4	18.9/22.6
No. of atoms
Total	2841	1138	1217
Protein atoms	2699	974	972
Water molecules	140	133	218
Ligand molecules	1	1	1
PEG molecules	—	1	1
Sulfate ions	1	—	1
Real-space CC (RSCC) for the ligand	0.83	0.87	0.94
Average B factor (Å2)	37.47	19.47	13.15
R.m.s.d., bonds (Å)	0.008	0.007	0.007
R.m.s.d., angles (°)	0.937	0.877	0.813

^† R _merge = , where I _i (hkl) is the intensity of the ith measurement and 〈I(hkl)〉 is the mean intensity for that reflection.

^‡ R _work = , where |F _obs| and |F _calc| are the observed and calculated structure-factor amplitudes, respectively.

^§ R _free was calculated with 5.0% of reflections in the test set.

2.4. Microscale thermophoresis (MST) assay

The wild-type BRD2 BD1 and BD2 bromodomains and their mutants were labeled with NT-647 reactive lysine dye using N-hydroxysuccinimide (NHS) ester chemistry (NT-647 labeling kit, NanoTemper Technologies, Germany) as per the manufacturer’s protocol. The BD1 and BD2 proteins were prepared (100 µl each) in labeling buffer (NanoTemper Technologies, Germany) with a final concentration of 10 µM. The fluorescent dye was prepared at a final concentration of 600 µM in 100% DMSO. In the labeling buffer, the concentration of the fluorescent dye was adjusted to 20 µM. The protein–dye mixture (1:1) was incubated for 30 min in the dark at room temperature, and unreacted dye was then removed using the supplied dye-removal columns equilibrated with MST buffer (50 mM Tris–HCl pH 7.5, 150 mM NaCl, 10 mM MgCl₂). Labeled proteins at a final concentrations of 25 nM were used in the assay. The NS5 and L10a stocks were prepared in 100% DMSO. A highest concentration of 250 µM NS5 was used for BD1 and BD2, with 16-step serial dilution in 1× PBST buffer. A highest concentration of 500 µM L10a was used in the assay with BD2. A Monolith NT.115 (Nanotemper Technologies, Germany) was loaded with Monolith standard capillaries containing samples for thermophoresis measurements. The data were acquired and analyzed using the MO Affinity Analysis 2.2.7 software (NanoTemper Technologies GmbH).

3. Results and discussion

3.1. Crystal structure of the BD2–NS5 complex

The crystal structure of the second bromodomain of BRD2 (BRD2 BD2) in complex with a pyrano-1,3-oxazine derivative (NS5) was solved by molecular replacement and refined to 2.00 Å resolution. The tertiary structure of the BD2 protein is a left-handed α-helical bundle of four helices (αZ, αA, αB and αC; Fig. 1 ▸ b). The loop regions, comprising the ZA loop (Trp370–Gly382) and the BC loop (Asn429–His433), are responsible for forming a deep hydrophobic pocket. The pocket recognizes the acetylated lysine (Kac) residue of acetylated histone tails. The polder difference Fourier map and |2F _o| − |F _c| map clearly showed electron density for the ligand NS5 [NSC 328111; 7-chloro-2-(3-chloroanilino)­pyrano[3,4-e][1,3]oxazine-4,5-dione] at the Kac binding site (Fig. 1 ▸ c and Supplementary Fig. S3a ).

The pyrano-1,3-oxazine moiety of NS5 binds to the deep pocket of BD2 and is stabilized by both hydrophobic and hydrophilic interactions. This moiety is surrounded by the residues Val376, Leu383, Cys425, Asn429 and Val435. The side chain of Asn429 makes a hydrogen bond to the O2 atom. A conserved water molecule bridges the O atom of this moiety and the hydroxyl group of Tyr386. Another conserved water molecule also bridges the O atom of this moiety and the carbonyl group of Met421. The pyrano-1,3-oxazine moiety also extends towards the WPF shelf region of BD2. The N1 atom of the moiety contributes a hydrogen bond to the carbonyl group of Pro371 (Fig. 1 ▸ d).

The linker N atom (N9), which connects the pyrano-1,3-oxazine moiety and the exposed chlorobenzene ring, forms a hydrogen bond to the backbone carbonyl group of Pro371. The exposed chlorobenzene group is positioned in the ZA channel and is sandwiched between Trp370 of the WPF shelf region and Leu381.

Partial electron density is observed in the chlorobenzene ring region, which may be due to fewer intermolecular interactions in this region. A similar observation has also been made in the recently reported BD1–NS5 complex structure (Deshmukh et al., 2020 ▸).

3.2. Comparison between BD2–NS5 and BD1–NS5 complexes

We have recently reported the crystal structure of the BRD2 BD1 bromodomain in complex with NS5, and the ligand possesses a significant inhibitory effect in glioblastoma, as confirmed by a cell-based assay in U87MG glioma cells (Deshmukh et al., 2020 ▸). As the electron density at the ligand-binding site was moderate, we thoroughly tested several crystals (soaked crystals or co-crystals) by varying the cryoprotection, ligand solvents and crystallization buffer conditions. The crystals in the absence of glycerol were more fragile and produced high-mosaicity diffraction. Hence, glycerol was necessary to produce better diffracting crystals. However, electron density for the ligand was apparent in all of the crystals, suggesting the unambiguous binding of the compound to BRD2 BD1 (Table 2 ▸; Supplementary Figs. S3b and S4).

When we compared the crystal structure of the BD2–NS5 complex with that of the BD1–NS5 complex, it was observed that the binding mode of the ligand in both structures is nearly identical (Fig. 2 ▸). However, near the pyrano-1,3-oxazine moiety binding regions, His433 in BD2 is replaced by Asp160 in BD1 and Val435 in BD2 is replaced by Ile162 in BD1. Near the exposed chlorobenzene ring region, Ala380 in BD2 is replaced by Lys107 in BD1. The side chain of Lys374 is flipped away from the edge of the binding pocket, whereas the side chain of Gln101 is positioned so that it makes a weak hydrophilic interaction with the linker N atom of NS5. However, in the BD2 complex a water-mediated interaction is observed at the equivalent position.

Figure 2.

Superposition of the BD2–NS5 complex with the BD1–NS5 complex (PDB entry 6jke). The NS5 ligands in BD1 and BD2 are shown as yellow and olive green sticks, respectively. Water-mediated interactions were observed at equivalent positions in both complexes. The binding-site residues corresponding to BD2–NS5 (deep teal) and BD1–NS5 (olive green) are shown as sticks. Water molecules are shown as red spheres.

The novel scaffold NS5 is a potential lead in further developing a ligand library to derive highly potent inhibitors against BRD2. The binding efficiency can be improved via designing analogs by (i) introducing rigidity by replacing the chlorobenzene ring with piperazine, indole, thiophene or oxazole at position 10 of NS5, (ii) introducing a long chain containing a hydroxyl group which may enhance interactions with the ZA loop region and (iii) the addition of smaller polar groups at position 9 which may increase the interaction with the WPF shelf region and nearby conserved water molecules.

3.3. BRD2 BD2 in complex with L10a

We have recently reported another novel scaffold molecule, a phenanthridinone derivative (L10), which is an inhibitor of BRD2 BD2 (Mathur et al., 2018 ▸). In our previously reported crystal structure of BD2 in complex with the core 6(5H)-phenanthridinone (L10a) moiety of phenanthridinone, the electron density corresponding to the ligand region was moderate, which was probably due to an alternate conformation (Mathur et al., 2018 ▸). To reconfirm the ligand binding, we soaked the ligand L10a with BD2 crystals in different crystallization conditions. X-ray data were collected from a crystal which was obtained in the condition 50 mM Tris pH 7.5, 50 mM NaCl, 25% PEG MME 2000. The structure of the L10a complex was refined to 1.7 Å resolution, as shown in Fig. 3 ▸(a). As expected, it revealed specific binding of L10a to the BD2 protein as shown from the polder difference Fourier map and |2F _o| − |F _c| map (Fig. 3 ▸ b and Supplementary Fig. S3c ). It is intriguing to note that the new crystallization condition contained no glycerol, suggesting that glycerol competes with the ligand binding and thus causes weak electron density corresponding to the ligand molecule (Supplementary Fig. S2).

Figure 3.

The crystal structure of BRD2 BD2 in complex with L10a. (a) A cartoon representation of the BD2–L10a complex and the chemical structure of L10a. (b) A polder difference Fourier map of L10a, contoured at 3.0σ, at the binding site. (c) A closer view of the intermolecular interactions between BD2 and L10a. Asn429 forms a hydrogen bond to L10a. A water-mediated interaction was observed through Tyr376. The interacting residues and ligand are shown as sticks. A water molecule is shown as a sphere. Hydrogen bonds are shown as broken lines.

As observed previously, the ligand binds at the base of the Kac binding pocket. The interactions observed for phenanthridinone and the core moiety of L10 are nearly identical. Both hydrophobic and hydrophilic interactions stabilize the compound. The side chain of Asn429 significantly contributes to hydrophilic interactions with the N and O atoms of L10a (Fig. 3 ▸ c). A conserved water molecule bridges the O atom of L10a and Tyr386. The cyclic ring of L10a makes hydrophobic interactions with the surrounding residues Val435, Pro371, Val376, Phe372, Leu383 and Leu381.

The binding efficiency of the L10a scaffold can be improved by designing analogs by increasing the hydrophobic inter­actions of the base of the phenanthridinone ring with the WPF shelf of the ZA loop region to tether the molecule at the base of the binding pocket. In addition, the introduction of electron-withdrawing groups on the ring, exposed to the solvent region, would help in forming hydrogen bonds (towards NE2 of His433) and hydrophobic interactions (towards Pro430) at the edge of the cavity.

3.4. Microscale thermophoresis (MST) assay

A microscale thermophoresis (MST) assay was carried out to evaluate the binding of the NS5 ligand to the wild-type BD1 and BD2 bromodomains of BRD2 and their mutants (Table 3 ▸; Figs. 4 ▸ and 5 ▸). A binding affinity (K _d) value of 2 µM was observed for wild-type BD1 (Fig. 4 ▸ a), which is similar to the reported value (1.2 µM; Deshmukh et al., 2020 ▸). For the wild-type BD2 protein, the binding affinity had a value of 43.5 µM (Fig. 5 ▸ a). The several-fold difference in K _d values suggests that the binding of NS5 to BD2 is substantially weaker than that to BD1.

Table 3. The binding affinity of the bromodomains for NS5 and L10a.

Bromodomain	Binding affinity (K d) (µM)
NS5
BD1, wild type	2.0
BD1, W97A	8.3
BD1, N156A	75.6
BD1, I162A	11.6
BD2, wild type	43.5
L10a
BD2, wild type	54.8
BD2, N429A	106.9

Figure 4.

Quantitative binding assay of wild-type and mutant BD1 with NS5. A concentration-dependent binding response of NS5 with BD1 significantly shows a binding affinity of 2 µM. The mutants of BD1 show weaker or abolished binding affinity to NS5. (a) WT/NS5, 2 µM; (b) W97A/NS5, 8.3 µM; (c) N156A/NS5, 75.6 µM; (d) I162A/NS5, 11.6 µM; (e) P98A/NS5, no binding.

Figure 5.

Quantitative binding assay of wild-type and N429A mutant BD2 with NS5. The dose–response curve of MST analysis significantly establishes binding of NS5 with a K _d value of 43.5 µM, while the N429A mutant abolishes the binding of NS5.

To further check the specificity binding pocket, the assay was performed for the BD1 mutants W97A, P98A, N156A and I162A. The W97A mutant yielded a K _d value of 8.3 µM, which is fourfold weaker than the binding to the wild type (Fig. 4 ▸ b), whereas the P98A mutant abolished the ligand binding (Fig. 4 ▸ e). This suggests that the WPF shelf lies in the ZA channel, which is vital for NS5 binding. The binding affinities for the other mutants, N156A and I162A, positioned in the pyrano-1,3-oxazine moiety binding region, were also determined. Intriguingly, the N156A mutant resulted in a binding affinity of 75.6 µM, which is about 38-fold weaker than the wild-type interaction (Fig. 4 ▸ c). The weaker affinity correlates well with the crystal structure, in which the critical hydrophilic residue, Asn156, contributes two hydrogen-bond interactions with this moiety. The I162A mutant gave a weaker binding affinity value of 11.6 µM (Fig. 4 ▸ d). Disrupting these hydrophilic and hydrophobic interactions will significantly affect the binding of NS5 to BD1. Further, the BD2 mutant N429A yielded no binding to NS5, indicating the disruption of vital interactions (Fig. 5 ▸ b). Thus, the overall results suggest that NS5 has higher specificity and affinity towards BD1 compared with BD2.

The binding-affinity measurement of ligand L10a with BRD2 BD2 was determined to be 54.8 µM (Fig. 6 ▸ a). For the N429A mutant, the binding affinity was significantly reduced, with a K _d value of 106.9 µM (Fig. 6 ▸ b), suggesting the importance of intermolecular hydrogen-bond interactions between Asn429 and the ligand L10a.

Figure 6.

Quantitative binding assay of wild-type and N429A mutant BD2 with L10a. The assay yielded a K _d value of 54.8 µM for wild-type BD2 and an affinity of 107 µM, approximately a twofold change, for the N429A mutant.

4. Conclusion

In macromolecular crystallography, standardization of crystallization conditions, obtaining protein crystals, maintaining their stability and seeking suitable cryo-conditions are rigorous processes and principal aspects for obtaining protein–ligand complex structures in the rational drug-discovery process. In our drug-discovery studies of BRD2 bromo­domains, we experienced the binding of β-mercaptoethanol, which is covalently linked to cysteine at the Kac binding pocket of BRD2 BD2 (Mathur et al., 2018 ▸). The presence of β-mercaptoethanol in the Kac binding pocket suggests that this reducing agent should be avoided in order to obtain appropriate binding assays and protein–ligand complex structures. Moreover, in our reassessment process of the ligands NS5 and L10a as well as other ligands (not shown), we found that several protein buffer components, crystallization reagents and ligand solvents such as glycerol, PEG, β-mercaptoethanol, Tris, DTT and DMSO favor binding at the Kac binding pocket (Supplementary Fig. S2).

After careful optimization of every reagent, solvent and precipitant, we established that avoiding or reducing the abovementioned reagents, solvents and precipitants was essential, ultimately leading us to obtain unambiguous electron density for the ligands in the crystal structures of the complexes. Using wild-type BRD2 bromodomains and their mutants, structural and biochemical studies confirm the binding of the phenanthridinone core moiety and pyrano-1,3-oxazine derivative ligands to BRD2 bromodomains. Thus, both compounds are novel molecules for the development of a library of potent inhibitors for regulating the function of BRD2 in cancer cells.

5. Related literature

The following references are cited in the supporting information for this article: Larkin et al. (2007 ▸) and Robert & Gouet (2014 ▸).

Supplementary Material

PDB reference: BD1–NS5 complex, 7env

PDB reference: BD2–L10a complex, 7enz

PDB reference: BD2–NS5 complex, 7eo5

Funding Statement

This work was funded by Department of Science and Technology, Ministry of Science and Technology, India grants DST-FIST: SR/FST/LS-I/2017(C) and IF190143 to Padmanabhan Balasundaram and Aishwarya H. Arole; Department of Biotechnology, Ministry of Science and Technology, India grant BT/PR7079/BID/7/426/2012 to Padmanabhan Balasundaram; Indian Council of Medical Research grant 45/51/2018/PHA/BMS to Prashant Deshmukh.