Duncan C.
Miller‡
a,
Mathew P.
Martin‡
b,
Santosh
Adhikari‡
a,
Alfie
Brennan
a,
Jane A.
Endicott
b,
Bernard T.
Golding
a,
Ian R.
Hardcastle
a,
Amy
Heptinstall
a,
Stephen
Hobson
a,
Claire
Jennings
b,
Lauren
Molyneux
a,
Yvonne
Ng
b,
Stephen R.
Wedge
b,
Martin E. M.
Noble
*b and
Celine
Cano
*a
aNewcastle Drug Discovery, Northern Institute for Cancer Research, School of Chemistry, Newcastle University, Newcastle upon Tyne, NE1 7RU, UK. E-mail: celine.cano@ncl.ac.uk; Tel: +44 (0)191 208 7060
bNewcastle Drug Discovery, Northern Institute for Cancer Research, Paul O'Gorman Building, Medical School, Framlington Place, Newcastle upon Tyne NE2 4HH, UK. E-mail: martin.noble@ncl.ac.uk; Tel: +44 (0)191 208 4466
First published on 15th February 2018
ATAD2 is an ATPase that is overexpressed in a variety of cancers and associated with a poor patient prognosis. This protein has been suggested to function as a cofactor for a range of transcription factors, including the proto-oncogene MYC and the androgen receptor. ATAD2 comprises an ATPase domain, implicated in chromatin remodelling, and a bromodomain which allows it to interact with acetylated histone tails. Dissection of the functional roles of these two domains would benefit from the availability of selective, cell-permeable pharmacological probes. An in silico evaluation of the 3D structures of various bromodomains suggested that developing small molecule ligands for the bromodomain of ATAD2 is likely to be challenging, although recent reports have shown that ATAD2 bromodomain ligands can be identified. We report a structure-guided fragment-based approach to identify lead compounds for ATAD2 bromodomain inhibitor development. Our findings indicate that the ATAD2 bromodomain can accommodate fragment hits (Mr < 200) that yield productive structure–activity relationships, and structure-guided design enabled the introduction of selectivity over BRD4.
ATAD2 comprises a four helical bundle (αZ, αA, αB, αC) and two loops (ZA and BC). The acetyllysine binding pocket, created by helices αB, αC and the ZA loop, is polar and shallow compared to several other bromodomains.13 The ZA loop of ATAD2, which forms a major part of the binding site is polar, whereas, the binding site in the bromodomain BRD4 is mostly hydrophobic.14 The flexibility of the ZA loop coupled to the shallow and polar nature of the binding site resulted in the druggability of the ATAD2 bromodomain being classified as ‘difficult’.13,14
Compounds 1 and 2a–b were recently disclosed as relatively potent inhibitors of the bromodomain of ATAD2.15–17 Compound 1 was not selective for ATAD2 over BRD4. Enhanced ATAD2 potency and selectivity over BRD4 was achieved in this series through introduction of a cyclic sulfone (2a) or a difluorocyclohexane moiety (2b). These groups form favourable interactions with the sidechains of Arg1007 and Arg1077 in the ATAD2 binding site, and achieve selectivity due to unfavourable interactions with lipophilic Trp81 and Met149 sidechains in BRD4. Compounds 3–5 were reported as ATAD2 inhibitors arising from a fragment screen,18 but the low potency of these fragments makes them unsuitable as chemical probes of ATAD2 function. More recently BAY-850 (6) has been reported as a potent ATAD2 inhibitor arising from screening of a DNA-encoded library with an unusual dimer-inducing mode of action, although not BRD4 selectivity data was reported.19
This work describes the development of ATAD2 inhibitors employing structure-guided optimization of a fragment hit, with potential for development into selective chemical tools to investigate ATAD2 bromodomain function in biological systems.
N 6-Alkylated target compounds were synthesised from 12 or 13 under microwave irradiation using an array of alkyl halides to give 15a–p and 16a–b, respectively. Subsequent N1-methylation gave compounds 17a–q and 18a–b (Scheme 2).
![]() | ||
Scheme 2 Reagents and conditions: (a) RX, MeCN, 170 °C, μW, 45 min, 22–93%; (b) MeI, Cs2CO3, DMF, 100 °C, μW, 30 min, 74–94%; (c) NaOH, EtOH, H2O, 100 °C, 23 h, 32%; Structures of 17a–q and 18a–b are given in Tables 2 and 3. |
Diversification of the C3-position could be achieved using Knoevenagel condensation of aldehydes with 11 to provide 19a–j in high yield (Scheme 3). Subsequent alkene reduction, followed by methylation provided 21a–f. Treatment of methoxypyridines 21a–f with benzyl bromides at high temperature provided the target pyridones (22a–f) in moderate yields.
![]() | ||
Scheme 3 Reagents and conditions: (a) RCHO, piperidine, THF, 100 °C, 30 min, 50–88%; (b) H2, 10% Pd/C, THF, MeOH, r.t., 2 h; (c) MeI, Cs2CO3, DMF, 50 °C, 1.5 h, 55–56% over 2 steps; (d) 4-chlorobenzyl bromide, MeCN, 170 °C μW, 45 min, 50–64%. Structures of 22a–f are given in Table 4. |
Methylation of the N1 position (17a) resulted in a change in binding mode (Fig. 2A), whereby the fragment core rotates and the pyrrolidinone carbonyl now forms a hydrogen bond with the carboxamide sidechain of Asn1064, and a second through a water-mediated hydrogen bond to Tyr1021. In this binding mode the N1-methyl group projects towards the conserved water molecules. Rotation of the pyrrolidinone core allows the pyridone carbonyl to interact through a water molecule with the backbone NH of Asp1014 on the ZA loop. One of the C3-geminal methyl groups occupies a small hydrophobic pocket formed between the sidechains of ZA-loop residues Tyr1021, Val1018 and the Tyr1063 residue that reside next to the conserved asparagine. Additional hydrophobic interactions are formed with the base of the binding pocket, where Val1008 and Ile1074 interact with the second C3-geminal methyl group and pyrrolidinone core, respectively. This binding orientation was designated as binding mode 2. Intriguingly, when the N1 substituent was further elaborated to an ethyl group (14a: Fig. 2B) the compound reverted to binding mode 1, suggesting the conserved water-rich pocket was unable to accommodate the larger ethyl group.
Fragments 7, 17a, and 14a bound very weakly and were unable to displace the acetylated histone ligand sufficiently in a homogeneous time-resolved fluorescence (HTRF) assay to enable a Kd to be determined, although some inhibition was observed at the highest concentration assayed (1000 μM; Table 1). Surface plasmon resonance (SPR) was used to assess the direct binding of the fragments and provide Kd values for binding to both ATAD2 and BRD4. SPR can yield artefactual data where the concentration of a small-molecular analyte exceeds ∼250 μM, and accordingly, this was the highest concentration used in the assays. Where the Kd exceeded this limit, projected Kds were determined by extrapolation, using the observed response units for compound binding (RU) relative to the molecular weight of a known control compound.23 Compounds 7 and 17a bound too weakly to ATAD2 to allow Kd values to be calculated. However, 17a did show detectable binding to BRD4. Compound 14a showed a remarkable increase in affinity towards BRD4, reporting a Kd of 12.7 μM and a binding constant 100-fold weaker towards ATAD2 (Table 1).
ID | R | ATAD2 IC50![]() |
ATAD2 LE | ATAD2 Kd![]() |
BRD4 Kd![]() |
---|---|---|---|---|---|
a HTRF format assay.
b SPR format assay.
c Isothermal titration calorimetry. Ligand efficiency (LE) = 1.4(−log![]() |
|||||
7 | H | >1000 | N/A | >2000 | >2000 |
2400c | |||||
17a | Me | >1000 | N/A | >2000 | 1500 |
14a | Et | >1000 | N/A | 1300 | 12.7 |
14b | Bn | 810 ± 57 | 0.21 | >2000 | 1050 |
14c | –CH2CH2Ph | 868 ± 33 | 0.19 | 1400 | 800 |
14d |
![]() |
>1000 | N/A | >2000 | 1700 |
14e |
![]() |
>1000 | N/A | >2000 | >2000 |
14g | –CH2CH2OH | >1000 | N/A | >2000 | >2000 |
The two binding modes offered different vectors to explore structure activity relationships (SARs) of substituents directed towards the RVF shelf and the ZA loop. Thus a set of N1-alkylated pyrrolidinopyridones was prepared to attempt to access this region while maintaining binding mode 1 (Table 1). Lipophilic benzyl (14b) and phenethyl (14c) substituents exhibited sub-millimolar HTRF IC50 values, and showed the first detectable signs of direct binding to both ATAD2 and BRD4 when analysed by SPR. Co-crystallisation confirmed that 14b retained binding mode 1 (Fig. 2C). Hydrophilic substituents (14d–g) gave no measurable ATAD2 inhibition or binding in HTRF or SPR format experiments.
To attempt to exploit binding mode 2, a set of N6-alkylated pyrrolidinopyridones were also prepared (Table 2) with N6-benzyl (17c) and N6-phenethyl (17d) analogues giving sub-millimolar ATAD2 IC50 values. As expected, 17c and 17d retained binding mode 2 (Fig. 2D and E), with the benzyl group sitting above the RVF shelf, forming a hydrophobic interaction with the sidechain of Lys1011. Pairwise comparison with N1-des-methyl analogue 15c confirmed the importance of a methyl group in the conserved water pocket. 17c and 17d bound more tightly to BRD4 than to ATAD2 in the SPR experiments.
ID | R1 | R2 | ATAD2 IC50![]() |
ATAD2 LE | ATAD2 Kd![]() |
BRD4 Kd![]() |
---|---|---|---|---|---|---|
a HTRF format assay.
b SPR format assay. Ligand efficiency (LE) = 1.4(−log![]() |
||||||
17b | H | Pr | >1000 | N/A | >2000 | 1000 |
15c | H | Bn | >1000 | N/A | 2000 | 1400 |
17c | Me | Bn | 681 ± 120 | 0.21 | 700 | 101 |
17d | Me | –CH2CH2Ph | 741 ± 74 | 0.20 | 1000 | 161 |
ID | R | ATAD2 IC50a (μM) | ATAD2 LE | ATAD2 Kdb (μM) | BRD4 Kdb (μM) |
---|---|---|---|---|---|
a HTRF format assay.
b SPR format assay. Ligand efficiency (LE) = 1.4(−log![]() |
|||||
17e | 4-CO2Me | 651 ± 1 | 0.18 | 900 | 141 |
17f | 4-CO2H | 533 ± 40 | 0.19 | 600 | 1500 |
17g | 4-CN | 562 ± 67 | 0.20 | 900 | 142 |
17h | 4-CONH2 | 680 ± 66 | 0.18 | 700 | 300 |
17i | 4-SO2Me | 359 ± 37 | 0.19 | 600 | 200 |
17j | 4-Cl | 190 ± 63 | 0.24 | 200 | 40 |
17k | 4-Br | 236 ± 36 | 0.23 | 138 | 48 |
17l | 4-Me | 274 ± 75 | 0.23 | 400 | 58 |
17m | 4-CF3 | 179 ± 43 | 0.21 | 400 | 124 |
17n | 3,4-DiCl | 299 ± 28 | 0.21 | 159 | 65 |
17o | 2-CN | >1000 | N/A | >2000 | 300 |
17p | 3-CN | 668 ± 28 | 0.19 | 800 | 184 |
17q | 2-CO2H | >1000 | N/A | >2000 | 300 |
With the aim of improving binding affinity towards ATAD2 the sidechain of Arg1007 was targeted through substitution of the 4-position of the N6-benzyl group. Compound 17f showed similar ATAD2 affinity to 17c, and the co-crystal structure with ATAD2 (Fig. 2F) confirmed that the carboxylate did not form the projected charge–charge interaction with Arg1007. Other analogues bearing functionality with the potential to form hydrogen bonds with Arg1007 gave no further improvement in inhibition (17g–i). Lipophilic substituents at the 4-position of the benzyl group (17j–m, Fig. 2G) provided the most potent benzylic analogues, but continued to be selective for BRD4. Addition of a second chloro substituent at the 3-position (17n) resulted in a modest drop in potency. N6-Benzyl analogues with substituents at the 2- and 3-positions of the benzylic ring (17o–q) were prepared to attempt to introduce further interactions with the ZA loop, or to favourably influence the conformation of the ligand, but no further improvement in ATAD2 affinity was observed.
Attention turned to exploring substitution at the C3-geminal dimethyl group. The pro-S methyl group occupies a small lipophilic pocket in the ligand binding site of ATAD2, whereas the pro-R methyl provides opportunities for fragment growth. It was envisioned that larger substituents at the pro-R position may occupy the cleft between the ZA and BC loops where the substrate acetylated lysine sidechain enters the active site, and may also provide improved selectivity over the more sterically demanding BRD4. Synthetic chemistry to access chiral C3 analogues that retained one methyl but differed in the second C3 substituent proved challenging, with mono-alkylation strategies found to be low yielding. Knoevenagel condensation was successful when the aldehyde coupling partner did not contain an enolisable proton, allowing the synthesis of 22a–f (Table 4) as racemic mixtures. Symmetrical C3-diethyl analogues 18a and 18b were also prepared.
ID | R1 | R2 | X | ATAD2 IC50a (μM) | ATAD2 LE | ATAD2 Kdb (μM) | BRD4 Kdb (μM) |
---|---|---|---|---|---|---|---|
a HTRF format assay.
b SPR format assay. Ligand efficiency (LE) = 1.4(−log![]() |
|||||||
22a | Me | Bn | Cl | >1000 | N/A | 500 | >2000 |
22b | Me |
![]() |
Cl | >1000 | N/A | 300 | 1700 |
22c | Me |
![]() |
Cl | 352 | 0.17 | 110 | 1500 |
22d | Me | CH2C(CH3)3 | Cl | 489 ± 18 | 0.18 | 125 | >2000 |
22e | Me |
![]() |
Cl | 214 | 0.19 | 500 | >2000 |
22f | Me |
![]() |
Cl | 304 ± 83 | 0.20 | 113 | 800 |
18a | Et | Et | Cl | 163 ± 23 | 0.22 | 107 | 1000 |
18b | Et | Et | CF3 | 163 ± 37 | 0.20 | 300 | >2000 |
The crystal structure of 18a bound to ATAD2 (Fig. 3A) showed that the diethyl analogues delve further into the hydrophobic pockets formed within the ZA-loop residues Tyr1021, Val1018 and with Ile1074 in the base of the binding site. Gratifyingly, these analogues resulted in a significant improvement in selectivity over BRD4 while retaining ATAD2 binding and inhibition (Table 4) (Fig. 3B).
Superimposition of the crystal structure of 18a with the first bromodomain of BRD4 (PDB entry 2OSS) indicates that the selectivity for ATAD2 over BRD4 may be due to a steric clash of one of the ethyl groups with Leu94 and the other with Ile146 in the BRD4 structure (Fig. 3C).
Overall, we have demonstrated that structure-based optimisation of ATAD2 fragments was able to bring millimolar hits into a micromolar potency range. Although inhibition of ATAD2 in the pyrrolidinopyridone series reached a plateau at 100–200 μM, we have succeeded in identifying interactions that can enhance potency while introducing selectivity versus BRD4. Application of these lessons might contribute to the development of improved chemical probes of ATAD2's role and structure–function relationship, if applied to fragments with superior initial ligand efficiency.
Mathew Martin – Protein expression, crystallisation and structure determination, assay development and manuscript preparation. †Santosh Adhikari – Synthesis and characterization of compounds.
Alfie Brennan – Synthesis and characterization of compounds.
Jane A. Endicott – Project conception, experimental design and interpretation.
Bernard T. Golding – Experimental design and interpretation, data analysis.
Ian R. Hardcastle – Experimental design and interpretation, data analysis.
Amy Heptinstall – Synthesis and characterization of compounds.
Stephen Hobson – Synthesis and characterization of compounds.
Claire Jennings – Protein expression, crystallisation and structure determination.
Lauren Molyneux – Synthesis and characterization of compounds.
Yvonne Ng – Assay development.
Stephen R. Wedge – Project conception, experimental design and interpretation.
Martin E. M. Noble – Project conception, experimental design and interpretation, data analysis and manuscript preparation.
Céline Cano – Project conception, experimental design and interpretation, data analysis and manuscript preparation (submitting author).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8ob00099a |
‡ These authors contributed equally to the work described in this paper. |
This journal is © The Royal Society of Chemistry 2018 |