Chemical Inhibition of Protein Methyltransferases Matthieu Schapira  Cell Chemical Biology  Volume 23, Issue 9, Pages 1067-1076 (September 2016) DOI: 10.1016/j.chembiol.2016.07.014 Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 1 Chemical Coverage of PMTs PMTs are composed of 50 SET domain lysine methyltransferases (left) and 13 known Rossmann fold methyltransferases (right) that methylate arginine (PRMTs) or lysine residues (this number may grow as more Rossmann fold enzymes get characterized; Richon et al., 2011). Potent, selective, cell-active inhibitors that have been reported so far are shown. These compounds can compete with peptide substrates (orange), the methyl-donating cofactor (pink), allosterically destabilize the active conformation (blue) or target non-PMT subunits of multiprotein PMT complexes (black). References are A-366 (Sweis et al., 2014); A-893 (Sweis et al., 2015); BAY-598 (Eggert et al., 2016); Constel_cmp3 (Garapaty-Rao et al., 2013); EPZ004777 (Daigle et al., 2011); EPZ015666 (Chan-Penebre et al., 2015); EPZ020411 (Mitchell et al., 2015); EPZ031686 (Mitchell et al., 2016); pinometostat (EPZ-5676) (Daigle et al., 2013); tazemetostat (EPZ-6438) (Knutson et al., 2013); GSK126 (McCabe et al., 2012); GSK343 (Verma et al., 2012); MI-503 (Borkin et al., 2015); Novartis_cmp12,13 (Chen et al., 2016); OICR-9429 (Grebien et al., 2015); PFI-2 (Barsyte-Lovejoy et al., 2014); SGC0946 (Yu et al., 2012); SGC707 (Kaniskan et al., 2015); UNC1999 (Konze et al., 2013); UNC638 (Vedadi et al., 2011). Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 2 Multimodular Nature of PMTs PMTs are often composed of multiple structural domains, each carrying distinct functions. MLL is composed not only of a catalytic domain that methylates H3K4 (PDB: 2W5Z), but also includes a CXXC domain that binds unmethylated CpG dinucleotides (PDB: 4NW3), and an acetyl-lysine binding bromodomain juxtaposed to a methyl-lysine binding PHD finger (PDB: 3LQI). This modular arrangement allows for the direct interpretation of and rapid response to diverse epigenetic signals. No catalytic inhibitor of MLL was reported to date, and targeting other domains may be an alternative strategy. Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 3 Structural Mechanism of Cofactor and Substrate Competitors (A) Substrate peptide and cofactor (SAM) of PMTs bind at distinct sites (stars indicate the departing methyl group of SAM and the methyl-accepting nitrogen of lysine). (B) A structure of the H3K4 PMT SETD7 in complex with SAM and PFI-2, a substrate competitor (PDB: 4JLG), shows that the substrate-binding site (orange) of SET domain PKMTs is sandwiched between the I-SET and post-SET domains. The latter is also a major contributing element to the cofactor-binding pocket (purple). (C) A crystal structure of PRMT6 in complex with the cofactor product SAH and the pan-PRMT substrate competitor MS023 (PDB: 5E8R) illustrate the typical homo-dimeric arrangement (white/blue) of PRMTs, and a canonical N-terminal α helix that wraps around the bound cofactor and contributes to both cofactor- and substrate binding pockets (purple and orange, respectively). The post-SET, I-SET domains, and αX helix are conformationally dynamic elements. (D) The H3K27 PKMT EZH2 is the catalytic subunit of the PRC2 complex, which also includes the scaffolding proteins EED and SUZ12. The structure shown is a composite image of the PRC2 in complex with a histone peptide (orange) occupying the substrate binding site (PDB: 5HYN) and in complex with a pyridone inhibitor (blue; PDB: 5IJ7) that occupies a pocket at the interface of the SET domain and an N-terminal activation loop of EZH2 stabilized by EED. This inhibitor overlaps minimally, but sufficiently, with the cofactor-binding site to compete with SAM (Brooun et al., 2016; Justin et al., 2016). Mutations Y661D, at the I-SET domain, and Y111L, at the activation loop, that confer resistance to clinical inhibitors are 550 residues apart but less than 4.0 Å away and both map at the inhibitor-binding site. (E) In its active state, the activation loop of the Rossmann fold H3K79 PKMT DOT1L folds on the cofactor (PDB: 1NW3) (left). Picomolar inhibitors, including the clinical compound pinometostat (EPZ-5676), conserve the adenosine moiety of SAM and feature a large hydrophobic moiety that locks the activation loop in an inactive state and exploits a hydrophobic cavity that is obstructed in the SAM-bound conformation (PDB: 4ER5) (middle). Non-SAM analogs with improved pharmacokinetics occupy this hydrophobic cavity without extending into the binding cleft of the cofactor's adenosine moiety (PDB: 5DSX) (right). Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 4 Structural Plasticity of PMTs (A) The post-SET domain (yellow) of PKMTs can adopt different conformations in the apo state (left), and when the cofactor- or substrate-binding sites are occupied (right), as illustrated in structures of SETD7 (apo, PDB: 1H3I; ternary complex with SAM and PFI-2, a substrate competitor, PDB: 4ILG). (B) Left: the I-SET domain of PKMTs can pivot about a conserved pair of glycine residues between conformationally active and inactive states, as illustrated by structures of the H4K20 PKMT SETD8 with or without a substrate peptide (yellow, blue, respectively; PDB: 1ZKK, 4IJ8). Proper binding of substrate peptides (orange) and enzymatic activity of MLL proteins (middle) or EZH2 (right), depend on the stabilization of the I-SET domain in an active state, respectively by RBBP5, a component of MLL complexes (PDB: 5F6K), or by a distant EZH2 activation loop with the help of EED and SUZ12, two subunits of the PRC2 complex (PDB: 5HYN). (C) The N-terminal helix of PRMTs is folded on the cofactor in the active state (yellow) but can adopt major conformational transformations, as shown in two superimposed CARM1 structures, where it partially turns into a β strand and shifts to an entirely different face of the protein (red) in the absence of cofactor (apo, PDB: 3B3J; cofactor-bound, PDB: 3B3F). Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 5 Structural Chemistry of Substrate Competitors (A) Binding of substrate competitors often relies on the presence of SAM, probably because they are sandwiched between the I-SET and post-SET domain, which is often stabilized by the cofactor. (B) Direct interactions with SAM can also contribute to potent binding (SETD7, PDB: 4JLG; PRMT5, PDB: 4X61). (C) Loss of a cation-π interaction (black arrows) between the PRMT5 inhibitor EPZ0015666 and SAM or the close analog synefungin results in over 200-fold reduction in binding potency (Chan-Penebre et al., 2015). (D) Fragment 3 which, in the context of the potent inhibitor A-366, occupies the methyl-lysine binding channel of G9a, is inactive against G9a (Nguyen et al., 2013). Fragment 7 occupies the methyl-arginine-binding channel of PRMT6, makes a buried electrostatic interaction with E155, and binds with an excellent ligand efficiency of 0.56, even though part of the binding site (yellow) is disordered in the absence of ligand (structures with and without fragment 7, PDB: 5EGS, 4HC4) (Ferreira de Freitas et al., 2016). Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 6 Allosteric Inhibition of PMTs (A) SGC707 (green) binds at the junction of the β barrel and the Rossmann fold of PRMT3, at a cavity that is distant from the site of methyl transfer (substrate- and cofactor-binding sites are colored orange and purple, respectively). A pyrrolidine-amide extension of the inhibitor is buttressed against the N-terminal helix of the homodimeric subunit, and antagonizes its active conformational state (yellow) (PDB: 4RYL). (B) Left: in its active state, the N-terminal helix of PRMT6 (αX) interacts with a loop of the β barrel (βL) (PDB: 4Y30). Right: The β loop is shifted away from the active site in a conformation where a compound occupies an allosteric site. Interaction with the α helix is lost, the helix is disordered, and the structural integrity of the active site is compromised, suggesting a possible mechanism of allosteric inhibition (PDB: 4QPP). Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 7 Disrupting Chromatin Complexes (A) MLL is a component of a large multi-protein chromatin complex. Binding to Menin at its N terminus mediates association with LEDGF at leukemic loci and with the androgen receptor in castration-resistant prostate cancer. MI-503 occupies the MLL binding site of Menin, disrupts the MLL-Menin interaction, antagonizes the recruitment of MLL complexes to chromatin, and kills MLL-translocated leukemia and prostate cancer cells (Malik et al., 2015; Yokoyama et al., 2005). Binding of WDR5 at the C terminus of MLL mediates association with the p30 oncogenic variant of the transcription factor C/EBPα. OICR-9429 occupies the MLL-binding pocket of WDR5, disrupts the MLL-WDR5 interactions, reduces the activity of MLL at p30-occupied loci, and kills p30 expressing acute myeloid leukemia (Grebien et al., 2015). (B) Binding of the scaffolding protein EED to the H3K27 PMT EZH2 within the PRC2 chromatin complex is necessary for enzymatic activity. A synthetic stapled helix occupies the EZH2-binding site of EED, disrupts the EZH2-EED interaction, and inhibits the methylation of H3K27 by the PRC2 complex (Kim et al., 2013). The stapled helix antagonizes a non-catalytic oncogenic function of EZH2 in tumors that are resistant to catalytic inhibitors (Kim et al., 2015). At the opposite surface of the EED scaffold, pharmacological targeting of the H3K27me3-binding pocket of EED should antagonize the stimulatory effect of this peptide on PRC2. Cell Chemical Biology 2016 23, 1067-1076DOI: (10.1016/j.chembiol.2016.07.014) Copyright © 2016 Elsevier Ltd Terms and Conditions