Identification and Characterization of the First Fragment Hits for SETDB1 Tudor Domain

SET domain bifurcated protein 1 (SETDB1) is a human histone-lysine methyltransferase, which is amplified in human cancers and was shown to be crucial in the growth of non-small and small cell lung carcinoma. In addition to its catalytic domain, SETDB1 harbors a unique tandem tudor domain which recognizes histone sequences containing both methylated and acetylated lysines, and likely contributes to its localization on chromatin. Using X-ray crystallography and NMR spectroscopy fragment screening approaches, we have identified the first small molecule fragment hits that bind to histone peptide binding groove of the TTD of SETDB1. Herein, we describe the binding modes of these fragments and analogues and the biophysical characterization of key compounds. These confirmed small molecule fragments will inform the development of potent antagonists of SETDB1 interaction with histones.


INTRODUCTION
Post-translational modifications (PTMs) such as methylation, phosphorylation, acetylation, ubiquitination, and SUMOylation of histone proteins impact gene expression by altering chromatin compaction or recruiting effector proteins. [1][2][3] Methylation of histones occurs predominantly at the lysine and arginine sidechains of histone 3 (H3K4, H3K9, H3K27, H3K36, H3K79 and H3R2) and histone 4 (H4K20 and H4R3). Unlike some other PTMs, methylation of a lysine or arginine sidechain does not change overall charge but results in subtle alterations in size and hydrophobicity, which can influence chromatin packing. 4 Transcriptional activation or repression is dependent on the position and degree of methylation 5 and aberrant methylation of histones can result in dysregulation of cell growth and development of cancer. 6 Many of the histone methyl transferases have been extensively studied and selective inhibitors have been effective in preclinical models. [7][8][9] Human SETDB1 is a histone-lysine methyltransferase (HMT), which specifically trimethylates histone H3 lysine 9 (H3K9me3). It is therefore deemed central in the silencing of specific euchromatic genes. 10 Recent studies have also suggested that SETDB1 could be an oncogene. Its overexpression is crucial to the growth of non-small and small cell lung cancer cell lines and has been reported in various human cancers. [11][12][13] Overexpressed SETDB1 increases H3K9me3 leading to alterations in gene expression. 14 SETDB1 contains a tandem tudor domain (TTD), a methyl-DNA binding domain (MBD), and a SET domain. 15 The MBD of SETDB1 promotes direct binding to methylated DNA sites and mediates trimethylation of H3K9 in a site-directed fashion. The SET domain catalytically "writes" the epigenetic methyl mark, which then goes on to recruit proteins belonging to the Heterochromatin Protein 1 (HP1) family to control gene repression. 10 Tudor domains are associated with the recognition of different methylation levels of lysine and arginine residues and the TTD of SETDB1 has been associated with the recognition of di-and trimethylation states of H3K9. 16,17 To the best of our knowledge, and despite the evident therapeutic importance of SETDB1, no antagonists of this protein have been reported to date. We therefore undertook fragment based approach toward discovery of ligands for SETDB1. We successfully crystallized the truncated tandem tudor domain of SETDB1 in complex with a histone 3 peptide dimethylated at lysine 9 and acetylated at lysine 14 (H3K9me2K14ac). 18 The co-crystal structure shows that H3K9me2K14ac interacts with the protein at two interconnected pockets within the TTD: a small pocket where the methyl lysine (Kme) binds (hereafter called the Kme binding pocket) and a relatively larger and deeper pocket engaged by the acetyl lysine (the Kac pocket) ( Figure 1). We used a crystal soaking fragment screening approach to identify small-molecule ligands occupying the Kme and/or Kac pockets of SETDB1-TTD. Crystal structures revealed binding modes that could enable structure-guided hit optimization toward the development of SETDB1 antagonists.

RESULTS AND DISCUSSIONS
Hit Identification. In order to identify small molecule antagonists of SETDB1-TTD, we screened a fragment library of 252 compounds by soaking the apo crystals of SETDB1-TTD with pools of 3 compounds and identified a single fragment hit (1). Co-crystal structure of 1 with SETDB1-TTD revealed that it binds in the Kme pocket ( Figure 2). However, fragment 1 did not show any HSQC shift at concentrations as high as 10 mM, indicating very weak binding.
As a major part of the H3K9me2K14ac peptide occupies and forms hydrogen bonding interactions in and around the Kac pocket, fragments binding in the Kac pocket are likely to have higher ligand efficiency and can potentially be optimized to improve affinity. To test this hypothesis, we selected and soaked 5 constrained analogues of known bromodomain inhibitors 19 synthesized in our laboratory (Supplementary Figure 1). Bromodomain inhibitors are known acetyl lysine mimics 20 and could potentially bind to the Kac pocket of the SETDB1-TTD. Out of 5 fragments, one fragment (2) was found to bind to SETDB1-TTD occupying the TTD Kac pocket ( Figure 2). Binding of 2 to SETDB1-TTD was also confirmed by HSQC NMR at 5mM (Supplementary Figure 2). In order to explore the SAR of these hits and assess the feasibility of improving their activity, a small set of analogs was synthesized and tested by HSQC NMR and/or by SPR.

Chemistry
The tricyclic triazole 2 was prepared from commercially available 2-aminobenzyl alcohol (11), which was reacted with ethylchloroacetate (12) in the presence of sodium hydride to give the lactam. Treatment of the lactam with Lawesson's reagent led to the corresponding thioamide 13, which was subsequently condensed with acethydrazide (14) to give the desired tricycle 2 in 6% yield overall yield (Scheme 1).     We also synthesized the corresponding urea 6, sulfonamide 7, and carboxylic acid 8 ( Figure 5).

OH
All three compounds were predicted to form additional or alternate hydrogen bonds with the protein, however, only the sulfonamide analog 7 showed a strong HSQC NMR shift at 5 mM  Despite forming a number of hydrogen bonding interactions with the protein, 5 showed only weak binding by ITC and SPR, which is likely due to incomplete occupation of the large pocket that is occupied by the native peptide H3K9me2K14ac. We therefore decided to grow the fragment 5 into the pocket occupied by the peptide. We initially focused on simple substitutions at the para position of the phenyl ring of 5 to rescue hydrophobic interactions observed between 2 and I333. However, none of the para-substituted analogues showed significant improvement of binding compared to 5 as measured by SPR (Table 1 and Supplementary Figure 4).     Amberlite IRA-67 was added to free-base the amine followed by addition of acetic anhydride (1.5 equiv). The reaction was allowed to stir for 6 -12 hours and after removal of the solvent, crude products were purified by reverse phase or normal phase column chromatography.

2-(3-(((tert-butoxycarbonyl)amino)methyl)-5-methyl-4H-1,2,4-triazol-4-yl)benzoic acid
(33). Anhydride 28 (2.65 g, 16.4 mmol, 1.0 equiv.) 21 and hydrazide 32 ( mg, mmol, 1.2 equiv.) 22 were refluxed EtOH for 3 hours . 23 The reaction was cooled down to room temperature and the reaction was dry loaded in celite and purified on C18 functionalized silica column using a gradient  overnight. The next day the reaction was cooled down to room temperature and the LAH was quenched using a Fieser workup. After filtration, the resulting filtrate was dry loaded into celite and purified using C18 functionalized silica using a gradient from 5:  The reaction was monitored by LC-MS. Once the reaction was completed, the reaction was cooled to r.t. and then N-acetyl-N-methoxyacetamide (2.11 g, 16.1 mmol, 6.0 equiv.) were added to the reaction mixture and stirred overnight at room temperature. The resulting reaction mixture was dry loaded in celite and purified on C18 functionalized silica column using a gradient from 5:95