Discovery of a Highly Selective Cell‐Active Inhibitor of the Histone Lysine Demethylases KDM2/7

Abstract Histone lysine demethylases (KDMs) are of critical importance in the epigenetic regulation of gene expression, yet there are few selective, cell‐permeable inhibitors or suitable tool compounds for these enzymes. We describe the discovery of a new class of inhibitor that is highly potent towards the histone lysine demethylases KDM2A/7A. A modular synthetic approach was used to explore the chemical space and accelerate the investigation of key structure–activity relationships, leading to the development of a small molecule with around 75‐fold selectivity towards KDM2A/7A versus other KDMs, as well as cellular activity at low micromolar concentrations.

Chemical modifications of DNAand its associated histones regulate gene expression across the entire genome,a nd therefore have aprofound impact on anumber of fundamental biological processes. [1] As aresult, targeting the epigenetic pathways responsible for these chemical modifications may represent ap ivotal approach to addressing disease at the transcription level. [2] In order to realize the potential of epigenetics in drug discovery,atoolkit of chemical probes that selectively target individual epigenetic proteins and allow researchers to clearly identify their downstream effects is invaluable. [3] Significant progress has been made towards the development of al ibrary of chemical probes that target the proteins involved in histone acetylation, in particular the bromodomain family of epigenetic readers. [4] In contrast, proteins involved in the dynamic methylation of histone lysine residues have proven to be more challenging targets, especially the histone lysine demethylases (KDMs). [5] Them ajority of KDMs belong to the Jumonji C( JmjC) family of enzymes,which contain acatalytically-active Fe II ion in the active site and require a2 -oxoglutarate (2-OG) cofactor for demethylation in the catalytic JmjC domain. [6] TheJ mjC KDMs may be divided into six sub-families (KDM2-KDM7) based on substrate specificity,w ith KDM2 and KDM7 being closely related. [7] Am ajor challenge in generating chemical probes for KDMs is achieving selectivity between these structurally similar sub-families.C urrently, most inhibitors of the JmjC KDMs are iron-chelating 2-OG competitors ( Figure 1). [8][9][10][11][12][13][14] Although many of these molecules achieve high levels of in vitro potency, they are frequently limited by al ack of selectivity and activity in cells.P eptide inhibitors that either mimic the histone   substrate or bind KDMs allosterically have also been reported, [15] however peptides are often limited by their low cellular permeability.H erein, we describe the discovery of af irst-in-class,c ell-permeable KDM2A/7A inhibitor that exhibits more than or equal to 75-fold selectivity relative to other JmjC KDM sub-families. KDM2A catalyses the demethylation of mono-and dimethylated lysine 36 on histone H3 (H3K36). [16] The enzyme has been reported to be involved in the regulation of NF-kBs ignalling [17] and the control of stem-cell differentiation and proliferation. [18] Its overexpression in gastric and small-cell lung cancer cells suggests that inhibiting KDM2A may represent as trategy for targeting certain cancers at the transcription level. [19,20] All KDM2A inhibitors described to date are 2-OG competitors,a nd none are truly selective.Inaddition, many 2-OG competitors show reduced activity in cells,mainly due to poor cellular permeability and the high intracellular concentration of 2-OG. [21] We therefore envisioned an inhibitor that would mimic the structure of the histone substrate rather than the 2-OG co-substrate,a s postulated for the KDM7A inhibitor E67-2. [22] To identify astarting point, alibrary of known binders to methyllysine reader domains and histone methyl transferases was screened for inhibitory activity against apanel of KDMs, since we reasoned that such as pecialized library would be more likely to contain molecules that also interact with demethylases.Compound 1,which was prepared as aputative methyllysine binding domain inhibitor, [23] was identified as ap romising candidate for further optimization against KDM2A. However,a ttempts to significantly improve potencyb yf unctionalizing at the indole NH position and varying the aromatic substituent on the indole C-2 position were unsuccessful. Thepyrrolidine moiety in compound 1 was shown to be critical for potency,which we postulated was due to its role as aH3K36me2 mimic. Preliminary docking studies of 1 with KDM2A (PDB ID:2 YU1;F igure S11 and Section S8 in the Supporting Information) indicated ap otential for occupancyo ft he peptide binding site on the enzyme. Based on this initial model, we subsequently hypothesized that inhibitory activity towards KDM2A might be improved by replacing the original indole scaffold with as aturated indoline ring system. We envisioned an exploration of threedimensional chemical space around this core,with the aim of augmenting selectivity through increasing complexity. [24] To achieve this,amodular synthetic approach was employed to generate al ibrary of indoline-containing compounds and identify key structure-activity relationships (Scheme 1A).
Ab ase-mediated 5-endo-trig cyclization of aC -2-substituted aromatic imine afforded the racemic indoline core with two adjacent stereocenters.S ubsequent acylationo ft he indoline ring system conferred stability towards oxidation and provided ahandle for modulating polarity.Finally,metalcatalysed cross-coupling of the aryl bromide provided access to av ariety of linkers between the indoline core and pyrrolidine capping group.I nt otal, 45 racemic compounds were synthesized, and IC 50 values for inhibition of KDM2A were determined using two orthogonal enzyme activity assays:AlphaScreen [25] and RapidFire MS [26] (see Section S3.1 in the Supporting Information for complete inhibition data).
Key structure-activity relationships are summarized below (Scheme 1B,c ompounds 2-12). We examined different linkers and found that triazole (2), ether (3), and alkyne (4) linkers were well tolerated, with significantly lower IC 50 values than the original hit. Reduction of the alkyne functional group in 4 to an alkene (5)o ra na lkane (6)a lso improved potency. Molecules containing apyridine ring at the indoline C-2 position were marginally more active than analogues bearing other aromatic groups such as furan (7 or 8)a nd significantly more active than as ubstituted benzene (9). In addition, pyridine-containing compounds displayed the highest selectivity towards KDM2A (Section S3.1). Exploration of substituents at the all-carbon quaternary stereocenter as in 10 and 11 demonstrated that aPh,CN combination gave rise to the most potent series of compounds.U nfortunately, 12,t he most potent inhibitor identified, was found to be reactive in aqueous solution due to the susceptibility of the aaminoacetyl group to hydrolysis.H owever,t he N-acetyl group present in compounds 2-10 proved inert to hydrolytic cleavage.T he optimal length of the linker connecting the indoline core to the pyrrolidine capping group was found to be 7-8 atoms,and replacing pyrrolidine with other secondary amines or ac yclopentyl ring led to as ignificant drop in potency( Section S3.1).
The(S,S)enantiomers of 3 and 6 were found to be slightly more potent than their respective (R,R)analogues,and (S,S)-6 (IC 50 :0.16 mm)was assessed further in avariety of biological assays.I ni mmunofluorescence assays using HeLa cells ectopically expressing catalytically active KDM2A, ad osedependent increase in H3K36me2 staining was observed upon incubation with (S,S)-6,r eflecting augmented cellular H3K36me2 levels (Figure 2A and Section S3.3). No significant change in H3K36me2 fluorescence was observed for cells containing constitutively inactive KDM2A (Section S3.3). [28] Cytotoxicity towards HeLa and HAP1 cells was observed at higher concentrations (EC 50 22 mm and 7.1 mm respectively), and (S,S)-6 showed as imilar effect on the viability of human fibroblasts (HDFa; EC 50 :1 0mm)t oG SK-J4, aw ell-characterized chemical probe for KDM5/6 [9] (Section S3.4). This suggests ap otential activity window for investigating the effects of KDM2A inhibition within cells.T o profile its selectivity,(S,S)-6 was tested for inhibitory activity against ap anel of KDMs,m ethyllysine binding motifs,a nd epigenetic enzymes.I tw as found to be remarkably selective towards KDM2A (! 100-fold) relative to representatives of the other KDM sub-families,except closely related KDM7A (H3K9/K27me2/me1 demethylase), [29] where it was similarly potent ( Figure 2B). [30] (S,S)-6 was inactive towards ar epresentative panel of methyllysine binding domains,m ethyl transferases,and histone acetyl transferases (Section S3.2). To our knowledge,t his is the first time aK DM2A/7A-selective small molecule has been shown to inhibit demethylation in cells,with asignificant reduction in demethylation achievable at low mm concentrations.T oe xplore its cellular activity further, the effect of (S,S)-6 on gene expression in HAP1 cells was monitored using ahighly multiplexed 3' mRNAsequencing method. [31] Within adiverse panel of in-house compounds, our series of indoline-containing inhibitors was represented by (S,S)-6 and the less active close analogue 18 (IC 50 :1 7mm, Figure 2C).
Obtaining ac o-crystal structure of (S,S)-6 bound to KDM2A proved challenging,a nd hence,n on-denaturing mass spectrometry (MS) experiments were performed to determine the binding stoichiometry of (S,S)-6 to KDM2A. KDM2A was incubated with (S,S)-6 and subsequently introduced into am ass spectrometer under conditions optimized for the preservation of noncovalent interactions. [32] Thenative mass spectrum (Figure 3) shows 1:1b inding of (S,S)-6 to KDM2A. To verify the identity of bound (S,S)-6,w e performed tandem-MS on the 14 + charge state,r esulting in the removal of (S,S)-6 as asingly charged species ( Figure 3C, see inset).
Kinetic analyses subsequently revealed that (S,S)-6 does not display competitive inhibition kinetics with respect to either 2-OG or the peptide substrate (Section S6), thus suggesting ad ifferent mode of inhibition to the majority of previously discovered KDM inhibitors. [33] Consistent with this observation, (S,S)-6 did not displace fluorescent methylstat (a "bivalent" substrate-cofactor tracer for KDM2A) in fluorescence polarisation assays.T op robe the (S,S)-6 binding site further, KDM2A was subjected to ap hotoaffinity labelling profile with ad iazirine-containing analogue of (S,S)-6,a nd LC-MS/MS experiments were conducted (Section S7). The majority of covalently modified residues were found to be either aspartic or glutamic acids,t hus suggesting the formation of ar elatively long-lived electrophilic intermediate following photo-induced isomerization of the diazirine to ad iazo compound. [34] While this precludes the unambiguous determination of the inhibitor binding site,the observed lack of labelling within the JmjC domain active site (Section S7) is consistent with the observed lack of competitive inhibition with respect to either 2-OG or the peptide substrate.This may indicate the presence of an alternative (allosteric) binding site specific to KDM2A/7A, although further investigation is necessary to demonstrate this clearly.
In conclusion, we have developed ap otent and selective first-in-class inhibitor of the histone lysine demethylases KDM2A/7A. Compound (S,S)-6 displays more than 75 fold selectivity towards KDM2A/7A versus other JmjC lysine demethylases and is,t oo ur knowledge,t he first reported selective KDM2A/7A inhibitor that has been demonstrated to reduce H3K36me2 demethylation within cells.T his study demonstrates how the generation of three-dimensional scaffolds bearing significant saturation and multiple chiral centres can lead to the discovery of selective compounds that may be useful in the study of achallenging epigenetic target.

Conflict of interest
Theauthors declare no conflict of interest.