diffracting crystals for apo structures may not be suitable for crystallization with a
ligand. Finally, crystallography is material, time, and financially intensive. Large
quantities (10–15 mg/mL) of pure and concentrated protein are needed. Protein
preparation, scoring crystals, cryo-protecting crystals, data collection, and data
processing require a trained researcher. Time at a synchrotron or home x-ray source
and crystallography tools are additional infrastructure needs.
A 2012 paper by Fish et al. [129] is an excellent example of the power of
crystallography in aiding the process of transforming an unselective, modestly
potent acetyl lysine mimic fragment into the potent, BET family-selective, commercially available probe PFI-1 (Fig. 4). Fragments mimicking the endogenous acetyl
lysine ligand are attractive starting scaffolds because they anchor binding through a
hydrogen bond to asparagine and a water-mediated bond to a tyrosine. To achieve
improved potency and selectivity, the starting scaffold must be extended to reach
portions of the binding pocket that differ between bromodomains. Inspired by the
BRD2 BD1 co-crystal structure by Chung et al. [112] (PDBID 4A9E), Fish and
co-workers chose 3,4-dihydro-3-methyl-2(1H)-quinazolinone 2 as acetyl lysine
mimics as a starting point. Compound 2 was active in an AlphaScreen assay against
BRD4 and CREBBP. While not universally active (no affinity for BPTF or BAZ2B),
selectivity for the BET family needed to be improved.
A crystal structure of 2 with BRD4 BD1 was instrumental in guiding decisions on
analog synthesis to achieve potency and selectivity (Figs. 18 and 19). Binding of 2 is
characterized by a hydrogen bond of the cyclic urea to N140 and lipophilic interactions with V87, L92, L94, Y97, F83, and I146 in the protein. The bromine at the C6
position extends into solvent, a logical place to add functional groups to the molecule
to access new binding interactions without disturbing the binding interactions
already present. The authors took advantage of the WPF shelf near the binding site
that is conserved among the BET family members, but not present in CREBBP, to
gain selectivity for the BET family. Sulfonamide linkers with hydrophobic caps
were used to introduce a pronounced kink at C6 allowing the proper vector to reach
and interact with the WPF shelf.
Two sulfonamide regioisomers were explored. The first generation of sulfonamides had the sulfur alpha to the dihydroquinazolinones. While these analogs
showed increased activity compared to the parent, the similar activity of the ethyl
and phenyl derivatives (e.g., 3, Fig. 18) suggests that lipophilicity of the aryl group is
not being efficiently delivered to the WPF shelf. This hypothesis was confirmed with
the crystal structure of 3, which has electron density consistent with multiple binding
poses (Fig. 19). Aryl groups are ideal for lipophilic and pi-stacking interactions
possible at the WPF shelf. The second generation of analogs with the nitrogen atom
alpha to the dihydroquinazolinone core showed significant increase in activity
compared to the first generation. A co-crystal structure of PFI-1 with BRD4 BD1
shows not only a binding interaction at the WPF shelf but also an interaction between
the sulfonamide oxygens and Q85 and a water-mediated hydrogen bond between the
carbonyl of L92 and the methoxy oxygen.
PFI-1 from the second generation of sulfonamides is a potent and BET familyselective chemical probe. A DSF assay was used to assess selectivity of PFI-1
Applied Biophysics for Bromodomain Drug Discovery
323
ligand. Finally, crystallography is material, time, and financially intensive. Large
quantities (10–15 mg/mL) of pure and concentrated protein are needed. Protein
preparation, scoring crystals, cryo-protecting crystals, data collection, and data
processing require a trained researcher. Time at a synchrotron or home x-ray source
and crystallography tools are additional infrastructure needs.
A 2012 paper by Fish et al. [129] is an excellent example of the power of
crystallography in aiding the process of transforming an unselective, modestly
potent acetyl lysine mimic fragment into the potent, BET family-selective, commercially available probe PFI-1 (Fig. 4). Fragments mimicking the endogenous acetyl
lysine ligand are attractive starting scaffolds because they anchor binding through a
hydrogen bond to asparagine and a water-mediated bond to a tyrosine. To achieve
improved potency and selectivity, the starting scaffold must be extended to reach
portions of the binding pocket that differ between bromodomains. Inspired by the
BRD2 BD1 co-crystal structure by Chung et al. [112] (PDBID 4A9E), Fish and
co-workers chose 3,4-dihydro-3-methyl-2(1H)-quinazolinone 2 as acetyl lysine
mimics as a starting point. Compound 2 was active in an AlphaScreen assay against
BRD4 and CREBBP. While not universally active (no affinity for BPTF or BAZ2B),
selectivity for the BET family needed to be improved.
A crystal structure of 2 with BRD4 BD1 was instrumental in guiding decisions on
analog synthesis to achieve potency and selectivity (Figs. 18 and 19). Binding of 2 is
characterized by a hydrogen bond of the cyclic urea to N140 and lipophilic interactions with V87, L92, L94, Y97, F83, and I146 in the protein. The bromine at the C6
position extends into solvent, a logical place to add functional groups to the molecule
to access new binding interactions without disturbing the binding interactions
already present. The authors took advantage of the WPF shelf near the binding site
that is conserved among the BET family members, but not present in CREBBP, to
gain selectivity for the BET family. Sulfonamide linkers with hydrophobic caps
were used to introduce a pronounced kink at C6 allowing the proper vector to reach
and interact with the WPF shelf.
Two sulfonamide regioisomers were explored. The first generation of sulfonamides had the sulfur alpha to the dihydroquinazolinones. While these analogs
showed increased activity compared to the parent, the similar activity of the ethyl
and phenyl derivatives (e.g., 3, Fig. 18) suggests that lipophilicity of the aryl group is
not being efficiently delivered to the WPF shelf. This hypothesis was confirmed with
the crystal structure of 3, which has electron density consistent with multiple binding
poses (Fig. 19). Aryl groups are ideal for lipophilic and pi-stacking interactions
possible at the WPF shelf. The second generation of analogs with the nitrogen atom
alpha to the dihydroquinazolinone core showed significant increase in activity
compared to the first generation. A co-crystal structure of PFI-1 with BRD4 BD1
shows not only a binding interaction at the WPF shelf but also an interaction between
the sulfonamide oxygens and Q85 and a water-mediated hydrogen bond between the
carbonyl of L92 and the methoxy oxygen.
PFI-1 from the second generation of sulfonamides is a potent and BET familyselective chemical probe. A DSF assay was used to assess selectivity of PFI-1
Applied Biophysics for Bromodomain Drug Discovery
323