We determined the velocity of the reaction by measuring the conversion of substrate to product at six time points spanning 100 min, analyzed these data by linear regression to determine initial steady-state enzyme velocity and fitted them to Michaelis-Menten kinetics. The K_m values of the peptide and SAM were then plotted as a function of UNC0638 concentrations. (a,b) UNC0638 is competitive with the H3K9 peptide substrate, as K_m^peptide increases linearly with compound concentration. (c,d) UNC0638 does not compete with the cofactor SAM, as K_m^app was not affected by the compound. (e,f) The X-ray cocrystal structure of the G9a-UNC0638-saH complex confirms the mechanism of action of UNC0638. UNC0638 (in gray, blue and red sticks) occupies the peptide binding groove and does not interact with the SAM binding pocket. The 7-(3-pyrrolidin-1-yl-) propoxy side chain of UNC0638 interacts with the lysine binding channel.

(a–d) MS of doubly charged peptides (KSTGGKAPR) corresponding to H3K9me2 (a,b) and H3K9me3 (c,d). In a,c, cells were treated with indicated shRNAs (control, D0 propionyl labeled; G9a, D0,D5 propionyl labeled, ~2.5 m/z heavier than the control; G9a + GLP, D5,D5 propionyl labeled, ~5 m/z heavier than the control). In b,d, cells were mock-treated (control, D0 propionyl labeled) or treated with 1 μM of BIX01294 (D0,D5 propionyl labeled, ~2.5 m/z heavier than the control) or UNC0638 (D5,D5 propionyl labeled, ~5 m/z heavier than the control) for 48 h.

(a) Time course of EGFP retrovirus activation during indicated treatments in J1 mES cells infected by HSC1-EF1α-EGFP-Puromycin retroviral vector. Plotted percentage of EGFP expression is the mean of the percentage of EGFP⁺ cells in three independent experiments. Error bars, s.d. (b) Analysis of mRNA levels of two G9a-regulated endogenous genes (MAGEA2 and DUB1) in J1 mES cells treated for 10 d with indicated compounds. The graph shows the normalized expression of MAGEA2 and DUB1 relative to the mRNA level detected in untreated cells (Δ(ΔCt)). (c,d) ChIP analysis of H3K9me2 enrichment at the EGFP gene (c) and MAGEA2 promoter (d) in cells treated with UNC0638 (500 nM) for 3 or 7 d. (e) Analysis of DNA methylation in the long terminal repeat (LTR) of the HSC1-EF1α-EGFP-Puromycin retroviral vector and in the MAGEA2 promoter after 10 d of treatment.

(a) UNC0638 (48 h) or G9a and/or GLP shRNAs, reduced H3K9 dimethylation levels. H3K9me2 antibody was used for cell immunostaining (in-cell western) and results normalized to cell number measured by uptake nucleic acid dye (DRAQ5). (b) UNC0638 was considerably more potent than BIX01294 and UNC0737 (negative control) in reducing cellular H3K9me2 levels, which were measured after MDA-MB-231 cells were treated with inhibitors for 48 h. Dashed line indicates level of H3K9me2 resulting from G9a and GLP knockdown. (c) Cellular levels of H3K9me2 were progressively reduced from 1 d to 4 d exposure to UNC0638 at three concentrations (80 nM, representing IC₅₀; 250 nM, representing IC₉₀; and 500 nM, representing 2 × IC₉₀). The reductions with 250-nM and 500-nM treatments after 4 d were equal or very close to that of G9a and GLP knockdown cells. Refreshing the inhibitor after 2 d (2+2d) increased inhibition by UNC0638 at 80 nM but had little further effect at 250 and 500 nM. The effects of UNC0638 were long-lasting. In cells with 2 d exposure to UNC0638, levels of H3K9me2 remained low after washout of compound followed by 2 d incubation without the inhibitor (2+2dw). (d) UNC0638 and UNC0737 had lower cellular toxicity than BIX01294 in MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide)) assays. (e,f) UNC0638 had good separation of functional potency (decrease in H3K9me2 levels) and toxicity (from the MTT assay), whereas BIX01294 had poor separation of these effects.

(a) Example of a genomic region (3p24.3) showing reductions in H3K9me2 after UNC0638 exposure (P = 2.0 × 10−10). Light blue bars show the log ratios of anti-H3K9me2 to IgG in control (top) and treated (bottom) samples. (b) Administration of UNC0638 decreased H3K9me2 in the MAGEA1 promoter. Log ratios are shown as in a. The promoter was defined to be 4 kilobase pairs upstream and 0.5 kilobase pairs downstream of the transcription start site. (c) UNC0638 did not change DNA methylation levels. MCF7 cells were exposed to either UNC0638 (at 70 or 320 nM), 5-azacytidine (DAC) or control. y-axis scale is the excess of the number of significantly hypomethylated probes over the number of hypermethylated probes on chromosome 3. Significance cutoff was set at P = 10−3 for two-sample t-test between treated and control log intensities.

Structure-based design and SAR exploration of the quinazoline template represented by BIX01294 led to the identification of UNC0321, a G9a and GLP inhibitor with high in vitro potency but poor cellular potency. The design and synthesis of several generations of new analogs aimed at improving cell membrane permeability while maintaining high in vitro potency resulted in the discovery of UNC0638, which has balanced in vitro potency and physicochemical properties aiding cell penetration. UNC0737, the N-methyl analog of UNC0638, was discovered as a structurally similar but less potent G9a and GLP inhibitor for use as a negative control.