GSK6853, a Chemical Probe for Inhibition of the BRPF1 Bromodomain.

The BRPF (Bromodomain and PHD Finger-containing) protein family are important scaffolding proteins for assembly of MYST histone acetyltransferase complexes. A selective benzimidazolone BRPF1 inhibitor showing micromolar activity in a cellular target engagement assay was recently described. Herein, we report the optimization of this series leading to the identification of a superior BRPF1 inhibitor suitable for in vivo studies.

There are at least 56 bromodomain modules contained within 42 human chromatin-regulator proteins.[1] Of these, the eight bromodomains of the BET (Bromodomain and Extra Terminal) subfamily (BRD2/3/4/T) have received most attention following the discovery that small-molecule inhibitors found during cellular screening campaigns bind to these targets and act by blocking their binding to Nε-acetyl-lysine (KAc) modified histones.[2−5] Inhibitors of BET proteins, such as I-BET762, RVX-208, OTX-015, and CPI-0610, have progressed into clinical evaluation for oncology and cardiovascular disease.[6,7]

The phenotypic effects of inhibiting other bromodomains remain to be discovered. The dearth of potent inhibitors of these bromodomains is one factor, but the problem is exacerbated by the significant and diverse phenotypic responses associated with inhibition of the BET proteins.[8] These can complicate the interpretation of phenotypic readouts obtained with poorly selective inhibitors. Biological evidence links several non-BET bromodomain-containing proteins to disease,[9] but because of the multidomain architecture and scaffolding function of these proteins the precise role of the bromodomain is often uncertain. For example, the BRPF (Bromodomain and PHD Finger-containing) family of BRPF1, BRPF2/BRD1, and BRPF3 (Figure S1, Supporting Information) act to facilitate the assembly of MYST-family histone acetyltransferase complexes.[10,11] In addition to the bromodomain, which binds acetylated histones,[12] the proteins contain other histone-binding domains (PZP and PWWP).[13−15] The availability of potent and selective chemical probes for BRPF bromodomains would significantly enhance our ability to dissect their biological role and better understand the therapeutic opportunities.

We recently reported the discovery of the BRPF1 bromodomain inhibitor 1 (GSK5959) (Table ).[17] In the BROMOscan panel of 34 bromodomain binding assays, it showed 10 nM BRPF1 inhibition, 90-fold selectivity over BRPF2, and over 500-fold selectivity over all members of the BET family. Others have independently found related BRPF1 inhibitors with similar properties.[18] Although the qualities of 1 made it an excellent probe to further elucidate the role of the BRPF1 bromodomain in many situations, its physicochemical properties and solubility are suboptimal. As part of our ongoing efforts to develop bromodomain probe molecules we sought to obtain an inhibitor with lower logD, solubility >100 μg/mL, and cellular activity < 100 nM.

Table 1

Summary of the Properties of GSK5959a

BRPF1	TR-FRET pIC50	7.1
	BROMOscan pKd	8.0
	NanoBRET pIC50	6.0
BRPF 2/3	TR-FRET pIC50	5.2/4.5
BRD4 BD1/2	TR-FRET pIC50	<4.3/<4.3a
Selectivity	BROMOscan	>90-fold
Chrom logDpH 7.4[16]		6.0
CLND solubility (μg/mL)		8

See also Table S1, Supporting Information.

Expansion of the SAR was carried out via analogue searches within the GSK collection, and bespoke syntheses (Scheme S1). The piperidine of 1 could be truncated to deliver the more potent and ligand efficient inhibitor 2, albeit at the expense of increased inhibition of BRD4 BD1 (Table ). Further reduction in the size of the 6-subsituent (3) was detrimental. However, with the ether-linked analogues an increase in potency and selectivity was observed on moving from 6-methoxy 4 to iso-propoxy 5, while ligand efficiency (LE)[19] was retained. Although the discovery of 5 represented a step toward an improved probe in terms of potency, LE, and logD, solubility still remained to be addressed.

Table 2

SAR at the Benzimidazolone 6-Positiona

See also Table S1, Supporting Information.

CLND (μg/mL).

The X-ray complex of BRPF1 with 1 (Figure )[17] suggested that the 3- and 4-positions on the benzamide provided vectors suitable for projecting a solubilizing group directly out into solvent, but presented no prospect of gaining interactions with the protein. In contrast, elaboration of the phenyl 2-methoxy also offered the potential to interact with residues of the ZA loop, such as Glu655, while retaining the intramolecular hydrogen-bond seen between the 2-methoxy and the amide NH of 1. Compounds 6–12 were accessed via amide coupling of the appropriate benzoic acids with the aromatic amine (Scheme S1).

Figure 1

X-ray structure of BRPF1 with 1 (PDB 4UYE).

While introduction of these groups generally improved solubility relative to 5, they all resulted in weaker inhibition of BRPF1 (Table ). This observation was particularly disappointing for side-chains containing hydrogen-bond donors (6 and 7), which were no more potent than close analogues without this functionality (10, 11), hydrogen bond acceptors (9), or basic groups (8 and 12). These results suggest that introduction of substituents larger than a methyl may give rise to steric clashes with the residues of the ZA loop and suboptimal binding conformations.

Table 3

Benzamide ortho-Position SARa

See also Table S1, Supporting Information.

CLND (μg/mL).

With the goals of expanding the diversity of substituents on the 5-position of the benzimidazolone core, improving solubility by lowering logD, and exploring alternatives to the intramolecular methoxy/amide hydrogen-bond, further 5-position analogues were synthesized (Schemes S1 and S2; Table ). All substituents showed improved solubility relative to 5, but aliphatic groups, whether linear (14 and 15) or cyclic (16 and 17) and positively charged (16) or neutral (14–15, 17) had little effect on potency relative to the 5-H (13) and were over 10-fold less potent than the 2-methoxybenzamide 5.

Table 4

Modifications at the 5-Positiona

See also Table S1, Supporting Information.

CLND (μg/mL).

Consistent with our previous observations for a 6-piperidine,[17] the unsubstituted benzamide 18 is also weaker than the 2-methoxy analogue 5. Presumably this is because the unsubstituted phenyl of 18 is unable to form the intramolecular hydrogen bond with the amide N–H. Anticipating that a 2-pyridyl could mimic this aspect of the 2-methoxy substituent, 19 was synthesized and shown to reduce logD relative to 5 while maintaining potency. The other pyridyl isomers (20 and 21) had activity comparable to the phenyl 18. We interpret this data as supporting the hypothesis that the intramolecular hydrogen bond plays an important role in restricting the degrees of freedom and increasing binding affinity (Figure S2). Further evidence came from the observation that the activities of compounds possessing a 2-heteroatom or 2-OMe group (22–23) retain higher potency than those without. Incorporation of a 6-N into the phenyl ring of 5 is detrimental to activity (compare 24 to 5), probably as a result of electrostatic repulsion between the lone pair of the pyridyl nitrogen and the amide carbonyl. This effect would be greater with the 6-OMe (25), explaining the further drop in potency.

The compounds in Table all met the objective for increased solubility relative to 5. However, it was disappointing that the best balance between BRPF1 potency and selectivity over BRD4 BD1 was achieved with the poorly soluble 2-methoxyphenyl 5. Unable to improve upon the 2-methoxybenzaminde in the 5-position, we switched our efforts to the benzimidazolone core itself. Inspection of the structure of 1 bound to BRPF1 (Figure ) suggested it may be possible for a chain possessing a solubilizing group to pass through the channel between Phe714, Asn708, and Tyr707, although the space for such a group was very limited. Extension of the 3-Me to a hydroxyethyl (Scheme S3) (26) gave a substantial increase in solubility, but with an unacceptable 40-fold drop in BRPF1 inhibition (Table ).

Table 5

Modifications at the N3-Positiona

See also Table S1, Supporting Information.

As the 3- and 5- positions seemed unproductive we returned to the 6-position of the benzimidazolone. The crystal structure of 1 in complex with BRPF1 (Figure ) showed the proximity of the backbone carbonyl of Asn651 to the C4 carbon of the piperidine. It was postulated that introduction of a basic nitrogen in this position could pick up an additional hydrogen bond to this carbonyl, while also improving solubility.

To test this hypothesis, compounds 28–31 were synthesized via SNAr reaction or Suzuki chemistry on an appropriately substituted core (Schemes S4 and S5). The piperazine 28 was well tolerated, delivering the desired improvement in physicochemical properties while maintaining potency at BRPF1 (Table ). Methylation of the piperazine to give 4-methylpiperazine 29 was slightly detrimental. The C-linked piperidine 30 and 1,2,3,6-tetrahydropyridine 31, however, were significantly less potent, with the sp3 hybridized carbon being more deleterious than the sp2. This is in accordance with our previous observation that the 6-H and 6-Me (3) analogues were ∼10-fold less active than the 6-OMe (4). Collectively, the results for compounds 1–5, 28, and 30–31 rule out steric demands as the explanation for the reduced inhibition of carbon-linked 6-substituents. Instead, they provide support for the hypothesis put forward in our previous paper that linking via a heteroatom plays an important role by electrostatically influencing the conformation of the 5-benzamide.[17] This is further supported by the distribution of torsion angles and distances in an analysis of small molecule X-ray structures (Figure S3).

Table 6

Modifications at the 6-Positiona

See also Table S1, Supporting Information.

CLND (μg/mL).

Despite the 40-fold variation in the BRPF1 activities of analogues 1 and 27–31, there was comparatively little change in potency versus the closely related BRPF2 bromodomain (Table S1). The largest selectivity window over BRPF2 is therefore approximately 160-fold (27 and 28).

Having achieved with 28 acceptable solubility, while maintaining significant BRPF1 inhibition, we next sought to investigate whether improved potency and selectivity was readily achievable. Interrogating the crystal structure of 1 in complex with BRPF1, it was intriguing to note that the 2-position of the piperidine presented a vector toward Pro658 (Figure A). This residue is one of the few that are not conserved across the BRPF subfamily, being substituted for Ser592 in BRPF2 or Asn619 in BRPF3. Analysis of the BRPF1 site using GRID probes suggested the potential for favorable interaction with a methyl group in the small indentation in the surface near Pro658 (Figure A).[20]

Figure 2

(A) GRID map analysis using the BRPF1 structure of compound 1 (PDB 4UYE). Methyl probe is shown in orange mesh. (B). X-ray structure of BRPF1 (cyan, PDB 5G4R) with 34, overlaid with BRPF2 apo (magenta, PDB 3RCW).

This hypothesis was first explored with the 2-methyl pyrrolidines 32 and 33. Indeed, a significant increase in BRPF1 inhibition was observed, especially with the (R)-enantiomer, with minimal effect on the inhibition of BRPF2 (Table S1, compare 32 with 27), BRPF3, and BRD4 (Table ). Transferring these findings to the more soluble piperazines (34–35) again showed the (R)-enantiomer 34 to be the most potent. The comparatively weaker inhibition with the 2-(S)-enantiomer 35, and 3-methyl enantiomers (36–37), confirm the specificity of this interaction.

Shortly afterward a 2.0 Å crystal structure of the BRPF1 bromodomain in complex with 34 was obtained (Figure B). This showed that 34 binds to BRPF1 in a very similar way to 1. As expected, the 2-methylpiperidine occupies the pocket close to Pro658, and the piperazine 4-NH makes a direct hydrogen bond to the backbone carbonyl of Asn651. Overlaying the apo structure of BRPF2 highlights the structural differences between the isoforms, which the 2-methyl substituent is able to exploit.

To facilitate interpretation of phenotypic assays we designed a less active analogue of 34 in which the 5-amide was alkylated, 38 (Scheme S5, GSK9311). We reasoned that this would prevent the 5-aryl group from adopting the conformation observed in the structures of 1 and 34 because of the loss of all internal hydrogen-bonding interactions, and steric clashes between the amide N-alkyl group and both the piperazine and the o-methoxy substituents. Usefully, this change resulted in a 100-fold drop in BRPF1 inhibition. This result can be rationalized by a crystal structure of 38 in BRPF1 in which the benzamide group packs poorly against P658 and E661 (Figure ).

Figure 3

X-ray structures of BRPF1 with 34 (cyan, PDB 5G4R) and 38 (magenta, PDB 5G4S).

Compound 34 was tested in the BROMOscan panel of bromodomain binding assays (DiscoveRx). Consistent with the data above, it showed excellent BRPF1 potency (pKd 9.5) and greater than 1600-fold selectivity over all other bromodomains tested (Table S2). To our knowledge this is the most potent and selective inhibitor of any single bromodomain reported to date.

All biochemical data presented thus far was determined using recombinant, truncated bromodomains. Potent binding to full-length endogenous BRPF1 (pIC50 8.6) was confirmed in a chemoproteomic competition binding assay using pull-down of lysate from HuT-78 cells followed by immuno readout (Figure ). Selectivity over endogenous BRD3 was also shown.

Figure 4

Potency for endogenous cellular BRPF1. pIC50 of 34 for BRPF1 and BRD3 measured in a chemoproteomic competition binding assay (Supplementary Methods, Supporting Information).

Compounds 34 and 38 were assessed in a cellular protein interaction assay measuring the displacement of NanoLuc-tagged BRPF1 bromodomain from Halo-tagged histone H3.3 to demonstrate cell permeability and disruption of chromatin binding (Supplementary Methods, Supporting Information).[21] Compound 34 exhibited potent inhibition (Figure S7) with pIC50 7.7, a 50-fold improvement over 1. The less active analogue 38 gave a reduced effect (pIC50 5.4), in line with its biochemical data.

Screening 34 against a panel of 48 unrelated assays revealed only off-target activities that are relatively weak compared to the BRPF1 potency (Table S3). However, to minimize the chance of off-target effects we recommend that a concentration of no higher than 1 μM should be used in cell-based assays. To assess the suitability of compound 34 for in vivo studies, its DMPK characteristics were determined in male CD1 mice via iv, po, and ip administration. The results indicate that the ip route of administration would be suitable for dosing this molecule in potential PK/PD models. In order to be able to compare biochemical potency measurements to free blood concentrations, the fraction unbound (fub) in the CD1 mouse was also determined with a resulting value of 7.9% (Table and Table S4).

Table 7

Summary of Properties of Compound 34a

BRPF1	TR-FRET pIC50	8.1
	BROMOscan pKd	9.5
	NanoBRET pIC50b	7.7
BRPF 2/3	TR-FRET pIC50	5.1/4.8
BRD4 BD1/2	TR-FRET pIC50	4.7a/<4.3
selectivityc	BROMOscan	>1600-fold
Chrom logDpH 7.4[16]		2.0
CLND solubility (μg/mL)		140
iv CLb (mL/min/kg)/t1/2 (h)		107/1.7
F% ip/po (3 mg/kg)		85/22

See also Table S1, Supporting Information.

Promega Corp., Figure S5.

DiscoveRx Corp., Table S2.

To conclude, we have optimized compound 1 to produce 34 (GSK6853), a potent (300 pM), soluble, cell active (20 nM), and highly selective inhibitor of the BRPF1 bromodomain, and demonstrated properties suitable for cellular and in vivo studies. We hope that disclosure of this chemical probe will stimulate further investigation into the functions of the BRPF1 bromodomain.