Development of potent dimeric inhibitors of GAS41 YEATS domain

Summary

GAS41 is an emerging oncogene overexpressed and implicated in multiple cancers, including non-small cell lung cancer (NSCLC). GAS41 is a dimeric protein that contains the YEATS domain, which is involved in the recognition of lysine-acylated histones. Here, we report the development of GAS41 YEATS inhibitors by employing fragment-based screening approach. These inhibitors bind to GAS41 YEATS domain in a channel constituting a recognition site for acylated lysine on histone proteins. To enhance inhibitory activity, we developed dimeric analog with nanomolar activity that blocks interactions of GAS41 with acetylated histone H3. Our lead compound engages GAS41 in cells, blocks proliferation of NSCLC cells and modulates expression of GAS41 dependent genes, validating on-target mechanism of action. This study demonstrates that disruption of GAS41 protein-protein interactions may represent an attractive approach to target lung cancer cells. This work exemplifies the use of bivalent inhibitors as a general strategy to block challenging protein-protein interactions.

Introduction

The recognition of post-translational modifications of histone proteins plays a fundamental role in transcriptional regulation (Allis and Jenuwein, 2016). YEATS domain-containing proteins belong to a relatively newly discovered family of epigenetic reader proteins that include four human paralogs: ENL (or MLLT1), YEATS2, AF9 (or MLLT3), and GAS41 (Glioma amplified sequence 41), also known as YEATS4 (Schulze et al., 2009). Biochemical studies revealed that YEATS domains bind to chromatin by recognizing histones with acetylated or crotonylated lysine side chains (Andrews et al., 2016; Li et al., 2014; Zhao et al., 2016; Zhao et al., 2017). GAS41 is an emerging oncogene overexpressed and implicated in multiple cancers. Amplifications of GAS41 have been identified in brain cancer patients, including 23% of glioblastomas and 80% of astrocytomas (Fischer et al., 1997; Fischer et al., 1996). GAS41 is also frequently amplified in sarcomas (Barretina et al., 2010; Italiano et al., 2008), as well as in colorectal (Tao et al., 2015) and gastric cancers (Kiuchi et al., 2018). In addition, an overexpression of GAS41 was detected in ~20% of non-small cell lung cancers (NSCLCs), but not in “normal” lung epithelial and fibroblast cells (Hsu et al., 2018; Pikor et al., 2013). Furthermore, the knockdown of GAS41 in a panel of NSCLC cell lines with GAS41 amplification strongly impaired cell growth and the formation of colonies (Hsu et al., 2018; Pikor et al., 2013). Altogether, GAS41 has emerged as a novel target in NSCLC and possibly in other cancers, and it represents an attractive protein for inhibitor development. Yet, to date, small molecule inhibitors of GAS41 have not been reported.

Biochemical studies revealed that GAS41 YEATS domain binds acetylated and crotonylated histone H3 peptides with moderate binding affinities (Cho et al., 2018; Hsu et al., 2018). One of the primary post-translational modifications recognized by GAS41 is histone H3 acetylated on K27 (H3K27ac), which may promote the recruitment of GAS41 to actively transcribed genes (Hsu et al., 2018). Previous studies by our lab and others reported the structural basis of the recognition of H3K27ac by the GAS41 YEATS domain (Cho et al., 2018; Hsu et al., 2018). Further analysis revealed that acetylated lysine binds in a channel on the GAS41 YEATS domain, which may constitute a binding site for peptidomimetics (Li et al., 2018). Here, we report the development of small molecule inhibitors of the GAS41 YEATS domain, identified by fragment-based screening and improved by chemical optimization, resulting in compounds with low-μM activity. The crystal structure of the GAS41-inhibitor complex demonstrates that these compounds bind to the acetylated lysine recognition channel on the YEATS domain and thus compete with acetylated histone binding. As GAS41 exists as a dimer in the cellular environment (Cho et al., 2018), we also developed dimeric compounds, resulting in nanomolar GAS41 inhibitors. Our most potent compound 19 demonstrates on-target activity and growth inhibition of NSCLC cells, supporting the utility of these compounds as chemical probes to further explore the role of GAS41 in lung cancer, and to develop more advanced compounds with optimized drug-like properties.

Results

Development of GAS41 inhibitors by employing fragment screening

To identify ligands binding to the GAS41 YEATS domain, we performed NMR-based screening of a fragment-like small molecule library against the ¹⁵N-labeled YEATS domain (residues 13–158). The screen yielded compound 1, which contains thiophene amide scaffold (Figure 1A) and induces extensive chemical shift perturbations upon binding to the YEATS domain (Figure 1A,B). To test the activity of GAS41 inhibitors, we utilized a fluorescence polarization (FP) assay. GAS41 YEATS domain binds acetylated histone H3 peptides with modest affinities (K_Ds approaching the 34- to 58-μM range) (Cho et al., 2018). Thus, to develop the FP assay, we employed a more potent di-crotonylated histone H3-derived peptide conjugated with fluorescein (FAM-H3K23crK27cr), which binds to the GST-fused GAS41 YEATS domain with a low micromolar affinity (K_D = 1.3 μM) (Figure S1). Such an enhancement in the binding affinity over mono-acylated peptides results from the bivalent interaction mode of the di-crotonylated peptide (Cho et al., 2018). We validated the FP assay in the competition experiment with the H3K27ac peptide and determined IC₅₀ = 243 μM (Table 1), which is consistent with the relatively weak binding affinity of the mono-acetylated peptide H3K27ac (Cho et al., 2018). When tested in the FP competition assay, fragment 1 exhibits comparable activity to the H3K27ac peptide, with IC₅₀ = 210 μM (Figure 1C, Table 1). We also developed an AlphaScreen competition assay using his₆-tagged full-length GAS41 and biotinylated-, di-crotonylated-H3 peptide (biotin-H3K23crK27cr), which yielded IC₅₀ = 73 μM for 1 (Figure 1D, Table 1). By comparison, H3K27ac yields IC₅₀ = 24 μM in such an assay (Table 1). We noted that the activities of GAS41 inhibitors measured by the AlphaScreen assay are typically several-fold stronger over the IC₅₀ values from the FP assay due to lower concentrations of reagents used in AlphaScreen.

Figure 1.

Identification and optimization of GAS41 YEATS inhibitors. A) Chemical structures of GAS41 inhibitors. B) Superimposed ¹H-¹⁵N HSQC spectra of 60 μM ¹⁵N-labeled GAS41 YEATS with 5% DMSO (black) and 1 mM fragment hit 1 (red). Peaks perturbed upon binding of 1 are boxed. C) Activities of compounds determined using the fluorescence polarization assay. Compounds were tested in duplicates; representative curves are shown. D) Activities of compounds determined using AlphaScreen assay. Compounds tested in duplicates; representative curves are shown. E) Binding affinities of 9 and 11 determined using ITC. K_D and stoichiometry values were determined from three and two independent experiments, respectively.

Table 1.

Activity of GAS41 inhibitors determined using FP and AlphaScreen assays.

Compound	Structure	IC50 FP (μM)	IC50 AlphaScreen (μM)
H3K27ac	Ac-ATKAAR{K-Ac}SAP-NH2	243 ± 45	24 ± 8
H3K14acK27ac	Ac-RKSTGG{K-Ac}APRKQLATKAAR{K-Ac}SAP-NH2	53 ± 3	1.4 ± 0.4
H3K18acK27ac	Ac-GGKAPR{K-Ac}QLATKAAR{K-Ac}SAP-NH2	50 ± 9	1.9 ± 0.2
1		210 ± 13	73 ± 9
2		137.5 ± 0.5	ND
3		69.3 ± 22.7	ND
4		580.5 ± 59.5	ND
5		235.75 ± 36.55	ND
6		41.05 ± 9.45	ND
7		20 ± 2.1	3.9 ± 0.7
8		163 ± 15	13 ± 1.2
9		12 ± 1.3	1.21 ± 0.04
10		10.47 ± 1.69	ND
11		4.2 ± 1.7	0.602 ± 0.046
12		3.75 ± 0.02	ND
13		15.6 ± 1.2	ND
14		11.8 ± 0.7	ND
15		1.7 ± 0.3	ND
16		1.5 ± 0.6	ND
17		1.4 ± 0.2	ND
18		0.223 ± 0.021	0.0264 ± 0.0018
19		0.096 ± 0.011	0.0181 ± 0.0003

ND – not determined.

To improve the potency of the fragment hit 1 we replaced a tert-butyl group with more polar moieties, including acetamide (2 and 3) and methoxy (compound 4) groups (Table 1). We found that the introduction of acetamide yielded either two-fold (2, IC₅₀ = 137 μM) or four-fold (3, IC₅₀=69 μM) improvements in the activity over 1, while the methoxy group decreased the activity by two-fold (4, IC₅₀ = 580 μM) (Table 1). Further modifications of 3, such as introduction of an azetidine ring resulted in two isomers, among which 5 is less active (IC₅₀ = 235 μM), while 6 has improved inhibitory activity (IC₅₀ = 41μM). Then, we replaced azetidine with a larger positively charged pyrrolidine ring and synthesized two stereoisomers, which are analogs of (S)-proline (compound 7) and (R)-proline (compound 8). Interestingly, the analog with (S)-proline (compound 7) is eight-fold more potent than with (R)-proline (IC₅₀s in FP assay are 20 μM and 163 μM, respectively, Figure 1C, Table 1). The preference for the binding of (S)-enantiomer 7 was further validated in the AlphaScreen assay, which resulted in the IC₅₀ values of 3.9 μM and 13 μM for 7 and 8, respectively (Figure 1D, Table 1). Subsequent optimization of 7 was focused on the replacement of the pyrrolidine with the azetidine moiety, which was further substituted with different aromatic rings. We then found that compound 9 (with thiazole) and 10 (with pyridine) yielded an additional two-fold improvement in the activity, resulting in low μM GAS41 YEATS inhibitors (Table 1). Characterization of 9 using FP and AlphaScreen assays yielded IC₅₀ = 12.0 μM and 1.2 μM, respectively.

GAS41 inhibitors block the acetyl-lysine binding channel in the YEATS domain

To understand the binding mode of GAS41 inhibitors, we crystallized the YEATS domain of GAS41 in complex with compound 9, and determined the crystal structure of complex at 2.10 Å resolution (Figure 2A, Table 2). We found that 9 binds in a channel that constitutes a recognition site for the acetyl-lysine of H3 (Cho et al., 2018), and is comprised of the side chains of H43, H71, S73, Y74, W93, F96 and the backbones of G92, G94 and E95 (Figure 2B,C). The thiophene amide carbonyl group in 9 forms two hydrogen bonds with the backbone amides of W93 (dist C=O···HN = 2.6 Å) and G94 (dist _C=O···HN = 3.0 Å) (Figure 2C), recapitulating the interactions of the side chain carbonyl group of the acetyl-lysine present in the histone peptides (Figure 2D) (Cho et al., 2018). The proline moiety in 9 forms another hydrogen bond involving the carbonyl of E95 (dist _C=O···HN = 2.9 Å) (Figure 2C), and the positive charge of the secondary amino group of the proline moiety is compatible with the negatively charged binding site on the GAS41 YEATS domain (Figure 2E). We also found that the thiophene ring in 9 is involved in π–π stacking interactions with the side chains of H71 (4.8 Å distance between ring centroids and 85° angle between aromatic rings), W93 (5.0 Å distance and 37° angle), and F96 (4.8 Å distance and 25° angle) (Figure 2C). Overall, the binding mode of 9 represents a favorable shape and charge complementarity to the acetyl-lysine binding pocket on the GAS41 YEATS domain (Figure 2E).

Figure 2.

Crystal structure of 9 in complex with GAS41 YEATS domain. A) Binding site of 9 (compound shown in sticks with green carbons) with mFo-DFc electron density map contoured at 3.0 σ. Selected GAS41 YEATS residues are shown as sticks with white carbons. B) Structure of the YEATS domain shown as a ribbon with selected side chains (shown as salmon sticks and transparent surface) and backbone atoms (transparent surface in light blue) forming the binding site and interacting with 9 (green carbons). C) Details of interactions of 9 with GAS41 YEATS domain. Hydrogen bonds are shown as dashed lines with indicated distances. D) Superposition of the crystal structure of GAS41 YEATS domain (gray carbons) with bound compound 9 (green carbons) and H3K23acK27ac peptide (PDB code 5VNB, orange carbons). Only a fragment of the H3K23acK27ac peptide showing the side chain of K27ac is shown. Hydrogen bonds involving carbonyl groups from 9 are shown as dashed lines. E) Electrostatic potential mapped onto surface of GAS41 YEATS calculated using APBS server (Jurrus et al., 2018). Scale in dimensionless units (k_b × T)/e_c, where k_b is the Boltzmann constant, 1.3806504 × 10⁻²³ J × K⁻¹, T is the temperature in K, and e_c is the charge of an electron, 1.60217646 × 10⁻¹⁹ C.

Table 2:

Summary of crystallographic statistics for GAS41 YEATS domain in complex with 9.

	GAS41 YEATS-9
Data Collection
Space group	C121
Cell dimensions
a, b, c (Å)	117.27, 112.50, 63.22
α, β, γ (°)	90.00, 116.74, 90.00
Resolution (Å)	39.65 – 2.10
Rmerge	0.09 (0.48)*
CC1/2	0.994(0.809)
<I>/<σ(I)>	11.74 (2.49)
Completeness (%)	98.1 (98.2)
Redundancy	3.9 (3.8)
Refinement
Resolution (Å)	39.65 – 2.10 (2.17 – 2.10)
No. reflections	41875 (4184)
Rwork / Rfree (%)	20.7/26.4 (25.3/31.0)
No. atoms
Protein	4108
Ligand	124
Water	728
Mean B-factors (Å2)
Protein	30.5
Ligand	30.6
Water	38.6
r.m.s. deviations
Bond lengths (Å)	0.002
Bond angles (°)	0.47

^*Values in parentheses correspond to highest resolution shells

Structural data suggested that thiazole could be replaced by a larger aromatic system, and we synthesized compounds 11 and 12 bearing 2-benzothiazolyl- and 2-benzothiophenyl substituents, respectively, which demonstrated approximately three-fold improvement over 9. For further evaluation, we selected compounds 9 and 11, and validated their activities in the AlphaScreen assay, yielding IC₅₀ values of 1.2 and 0.6 μM for 9 and 11, respectively. To quantify the binding affinity of these two compounds, we employed isothermal titration calorimetry (ITC) and determined K_D values of 7.9 μM and 4.7 μM for 9 and 11, respectively (Figure 1E). Binding of these compounds is primarily driven by favorable enthalpy change (Table S1).

To assess selectivity of 9 and 11 toward the GAS41 YEATS domain over the other family members, we utilized NMR. A series of ¹H-¹⁵N HSQC experiments clearly show extensive chemical shift perturbations in the GAS41 YEATS domain induced by both compounds, but no binding to the YEATS domains of the three homologs AF9, ENL, and YEATS2 (Figures S2, S3). By employing ITC, we showed that compound 9 is not binding to AF9 and ENL YEATS domains (Figure S3). Furthermore, 9 and 11 selectively stabilize GAS41 YEATS but not AF9, ENL and YEATS2 in thermal shift assays (Figure S4). Collectively, the optimization of fragment 1 resulted in compounds 9 and 11, with increased activities by two orders of magnitude, which selectively bind to the GAS41 YEATS domain but not to the other members of the YEATS family.

Dimerization markedly enhances the activity of GAS41 inhibitors

The structure of 9 bound to the GAS41 YEATS domain revealed that the inhibitor occupies the majority of the binding channel, making it challenging to enhance its potency further (Figure 2B,D). Based on our previous findings demonstrating the dimeric structure of the GAS41 YEATS domain (Cho et al., 2018), we explored the development of bivalent inhibitors to enhance their activity. First, we compared the activities of mono-acetylated histone peptide H3K27ac with di-acetylated H3K14acK27ac and H3K18acK27ac in the FP and AlphaScreen assays, and we found five- to ten-fold enhanced IC₅₀ values for the di-acetylated peptides (Table 1). Then, we employed the crystal structure of the YEATS domain with bound 9 to design dimeric analogs of compounds 9 and 11. To explore a suitable site for linking monomeric compounds, we selected fully solvent exposed Cγ position in a proline ring of compound 7 and introduced hydroxy and propargyloxy groups. This yielded compounds 13 and 14, both of which had comparable activities to 7 (Table 1). A similar modification introduced into 11 yielded 15 with comparable activity, again validating proline Cγ as a suitable site for introduction of a linker (Table 1). Based on these results, we synthesized dimeric compounds 16 and 17 by introducing, respectively, alkane and alkyne-based six-carbon linkers. We found that dimers 16 and 17 are approximately one order of magnitude more potent when compared to monomers (Table 1). Subsequently, we developed compound 18, which is a dimeric analog of 9 harboring a saturated six-carbon linker (Figure 3A). We tested binding of both compounds to the dimeric GAS41 YEATS domain using ITC and found that monomeric 9 binds with K_D = 11.8 μM and 1:1 stoichiometry, while the dimer 18 clearly demonstrates enhanced affinity with K_D = 195 nM and 0.5:1 stoichiometry (Figure 3B, Table S1). This validates that dimerization significantly improves the affinity towards dimeric GAS41 YEATS domain. We also developed another dimeric analog, compound 19 with a four-carbon linker designed based on monomer 11 (Figure 3A). While dimer 18 is relatively polar and suitable for biochemical assays, the more hydrophobic 19 is expected to demonstrate enhanced cellular permeability. We then tested the biochemical activity of dimeric compounds 18 and 19 and found that both analogs have IC₅₀ values in the nanomolar range and moreover are significantly more potent than the corresponding monomers (Table 1). Specifically, in the FP assay, we found 18 to be 54-fold more potent than 9 (IC₅₀ values of 223 nM versus 12 μM), and 19 had 44-fold improved inhibitory activity over monomeric 11 (IC₅₀ values of 96 nM versus 4.2 μM). Similar improvements in the activities of the dimeric analogs (33- to 46-fold) over monomers were also observed in the AlphaScreen assay, resulting in the IC₅₀ values of 18 and 26 nM for 19 and 18, respectively (Figure 3C, Table 1). We also tested stabilization of full length GAS41 by 9 and 18 using thermal shift assay (Huynh and Partch, 2015). Both compounds increase protein melting temperature (T_M) by ~ 3 °C; however, the dimer stabilizes GAS41 at ten-fold lower concentration when compared to the monomer (Figure 3D). Furthermore, both monomer 9 and dimer 18 selectively stabilize the YEATS domain of GAS41 but not AF9, ENL and YEATS2 (Figure S4).

Figure 3.

Activities of dimeric GAS41 inhibitors 18 and 19. A) Structures of dimeric inhibitors 18 and 19. B) Binding affinities of 9 and 18 towards GST-GAS41 determined using ITC. K_D and stoichiometry values were determined from two independent experiments. C) Activities of inhibitors determined using AlphaScreen assay. Representative data shown from two independent experiments. IC₅₀ calculated as mean ± s.d. from two independent experiments. D) Thermal shift assay showing stabilization of full-length GAS41 with different concentrations of 9 and 18. Two independent experiments were performed. E) AlphaScreen assay demonstrating inhibitor-induced dimerization of his₆-tag and biotin-labeled Avi-tag GAS41 YEATS domains. Representative data shown from two independent experiments. F) ¹H-¹⁵N HSQC spectrum of 60 μM ¹⁵N-labeled GAS41 YEATS domain (black) and in the presence of 60 μM 9 (red, left spectrum) or 30 μM 18 (red, right spectrum).

Dimeric GAS41 inhibitors induce YEATS domain dimerization

To determine whether bivalent inhibitors induce dimerization of the YEATS domains, we developed an AlphaScreen assay employing two different GAS41 YEATS domain constructs with, respectively, his₆-tag and biotin-labeled Avi-tag. Titration of both proteins with either 18 or 19 resulted in an increased luminescence signal, reflecting the formation of the compound-induced dimeric YEATS domain complex (Figure 3E). The signal was decreased at the highest concentrations of both compounds, indicating saturation of the YEATS domain via an excess of dimeric ligands, known as the Hook effect. The stronger signal for 18 likely reflects the higher solubility of the thiazole analog when compared to the benzothiazole analog 19. For comparison, we tested monomeric compounds 9 and 11 and observed no dimerization of YEATS domains (Figure 3E). To validate further the specificity of compound 18, we performed a competition assay and demonstrated that monomeric compound 9 disrupts 18-induced YEATS domain dimerization (Figure S5).

To assess the dimerization of the YEATS domain using an independent method, we compared the binding of monomeric 9 and dimeric 18 to the GAS41 YEATS domain using NMR. While 9 induces chemical shift perturbations upon binding, the addition of 18 to the YEATS domain at a 1:2 stoichiometric ratio resulted in a very substantial signal broadening, consistent with the formation of a larger dimeric complex (Figure 3F). A similar effect was observed for compounds 11 and 19 (Figure S2). Overall, these data clearly demonstrate that dimeric inhibitors facilitate specific dimerization of the GAS41 YEATS domain, which further explains their enhanced inhibitory activity against full-length GAS41.

Dimeric inhibitor blocks activity of GAS41 in cancer cells

To test whether inhibitors block GAS41 interactions in cells, we developed a split luciferase NanoBiT assay (Dixon et al., 2016; Oh-Hashi et al., 2016) by transfecting HEK293T cells with LgBiT-GAS41 and SmBiT-H3.3. The co-expression of both proteins resulted in a strong luciferase signal, reflecting the interaction of GAS41 with in situ acetylated histone H3.3 (Figure S6A). The introduction of a point mutation W93A in LgBiT-GAS41, which has been reported to abolish histone recognition by GAS41 (Hsu et al., 2018), largely diminished the luminescence signal, thereby validating the NanoBiT assay for this system (Figure S6A). Subsequently, we tested the activity of 19, a more cell-permeable dimeric inhibitor of GAS41 in the NanoBiT assay, and we found a dose-dependent reduction in the luminescence signal, indicating the disruption of GAS41 interactions with acetylated H3.3 in cells with IC₅₀ ~ 6 μM, (Figure 4A). Importantly, treatment with 19 did not reduce the luminescence signal for the W93A GAS41 mutant, which supports the specific on-target activity of this compound (Figure 4A).

Figure 4.

Cellular activity of 19. A) NanoBiT assay showing the treatment of 293T cells co-transfected with SmBiT-H3.3 and LgBiT-GAS41-WT or LgBiT-GAS41-W93A mutant with 19 for 24 h. B) Inhibition of cell proliferation of H1299 cells treated with 19 and 11. C) Inhibition of cell growth in A549 or A549 GAS41-KO cells by 19. D) Growth inhibition of H1299 and H1993 cells by 19. E) CETSA experiment in H1299 cells treated with DMSO or 12 μM 19. Quantification of band intensities is shown. Three independent experiments were performed and representative data are shown. F) Change in the expression level of E2F2, FOXM1, MCM6 and GDF15 in H1299 cells after 7 day treatment with 19 measured by quantitative RT-PCR. Data in panels A and B represent two independent experiments, each performed in duplicates. Data in panels C, D and F represent two independent experiments, each performed in triplicates. ns – not significant, * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001 calculated using Student’s 2-tailed t-test.

As previously found, GAS41 is amplified in NSCLC cell lines and, moreover, is required for NSCLC proliferation (Hsu et al., 2018; Pikor et al., 2013). To investigate the activity of GAS41 inhibitors, we treated H1299 cells with monomeric 11 and dimeric 19 for 4 days, and found that only the more potent 19 induced dose-dependent cell growth inhibition, with GI₅₀ ~ 3 μM (Figure 4B). To determine whether the cellular activity of 19 is dependent on the presence of GAS41, we developed A549 GAS41 knockout (GAS41-KO) cells using the CRISPR/CAS9 system (Figure S6B). We found GAS41-KO cells were viable, but they grew more slowly when compared to the parental A549 cells, with a ~70% reduced growth rate measured at day 14 (Figure 4C). Concurrently, treatment with 19 inhibited growth of A549 cells by ~40% at 12 μM of the compound, but it had no effect on the GAS41-KO cells (Figure 4C), thereby validating specific growth inhibition that is dependent on GAS41 expression. Next, we evaluated the effect of 19 on the growth of two NSCLC cell lines with GAS41 amplification, H1299 and H1993 (Hsu et al., 2018; Pikor et al., 2013). Treatment with 19 reduced the growth of both cell lines, with GI₅₀ ~ 6 μM at day 14 (Figure 4D). Such an effect correlates closely with the activity of 19 in inhibiting GAS41 binding to H3.3, as measured with the NanoBiT assay (Figure 4A). We further tested whether 19 engages GAS41 in cells using CETSA assay (Martinez Molina et al., 2013). Treatment of H1299 cells with 12 μM 19 clearly leads to stabilization of GAS41 in lung cancer cells (Figure 4E). To validate further the on-target activity of 19, we tested the expression of GAS41 target genes in H1299 cells (Hsu et al., 2018). Treatment with 19 resulted in a statistically significant decrease in the expressions of the E2F2, FOXM1 and MCM6 genes (Figure 4F), which were shown previously to be downregulated by GAS41 knockdown (Hsu et al., 2018), but not unrelated genes (Figure S6C). Treatment with 19 increased expression of GDF15, a gene upregulated upon GAS41 knockdown (Hsu et al., 2018), frequently downregulated in NSCLC tissues, and its low expression being associated with poor clinical outcome in NSCLC patients (Lu et al., 2018). Altogether, dimeric inhibitor 19 disrupts the binding of GAS41 to acetylated H3.3 in cells and induces on-target growth inhibition in NSCLC cell lines.

Discussion

The family of YEATS proteins represents an attractive target for the development of small molecule inhibitors (Barretina et al., 2010; Fischer et al., 1997; Fischer et al., 1996; Hsu et al., 2018; Italiano et al., 2008; Kiuchi et al., 2018; Pikor et al., 2013; Tao et al., 2015). YEATS domains are readers of histone proteins with post-translationally acetylated or crotonylated lysine side chains (Andrews et al., 2016; Li et al., 2014; Zhao et al., 2016; Zhao et al., 2017). Unlike in bromodomains, the acyl-lysine binding site in the YEATS domains is partially solvent-exposed and has a channel-like structure, rather than featuring a deep pocket (Linhares et al., 2020). Such a structural feature presents a difficulty in the development of highly potent small molecule ligands. While several recent studies described inhibitors of ENL (Heidenreich et al., 2018) or dual inhibitors of ENL/AF9 (Moustakim et al., 2018) YEATS domains, their activities are in the low or sub-micromolar range, with the IC₅₀ values above 200 nM. Thus, potencies comparable to bromodomain inhibitors have not been achieved, which emphasizes the challenges in targeting YEATS domains.

Here, we describe the discovery of small molecule inhibitors of the GAS41 YEATS domain. We identified a hit compound by screening a fragment library and improved its potency over 100-fold, leading to the development of low-μM GAS41 YEATS domain inhibitors 9 and 11. The co-crystal structure of 9 in complex with the GAS41 YEATS domain revealed that the inhibitor binds in a channel involved in the recognition of the acylated lysine side chain of histone H3 and forms a well-defined network of interactions, stabilizing the complex. However, despite the relatively small size of 9, the inhibitor occupies nearly the entire binding pocket, which hinders further optimization to enhance the potency. Therefore, to develop better inhibitors, we synthesized dimeric analogs 18 and 19, which could simultaneously interact with both YEATS domains in the GAS41 dimer. As expected, the dimeric compounds are more potent and possess activities improved by two orders of magnitude over the monomeric analogs 9 and 11. This results from the avidity effect that we previously observed in the recognition of the di-acetylated histones via the dimeric GAS41 protein (Cho et al., 2018). Exploring bivalent inhibitors might offer a more general means of targeting difficult protein–protein interactions, as has been demonstrated for other epigenetic reader domains with small and partially solvent exposed binding sites, such as Spindlin-1 (Fagan et al., 2019) and L3MBTL3 (James et al., 2013).

In this study, we developed GAS41 inhibitor 19 with very potent in vitro activity, which blocks GAS41 interactions with acetylated H3.3 in cells and inhibits proliferation of NSCLC cells in a GAS41-dependent manner, supporting on-target activity. Thus, compound 19 represents a valuable chemical tool and may further serve to explore the inhibition of GAS41 in other cancer models. Intriguingly, the activity of 19 in cell-based assays is substantially weaker than in biochemical assays, and such a discrepancy may suggest limited cell permeability of the dimeric compound 19 due to increased molecular weight. Alternatively, such an effect may reflect challenges in the disruption of the PPIs involving epigenetic complexes formed in the cellular environment. For example, GAS41 has been found to be a subunit of TIP60 and SRCAP complexes (Park and Roeder, 2006), and releasing such complexes from chromatin may require far more potent inhibitors than needed for the disruption of the binary GAS41-H3K23crK27cr system. Furthermore, a similar discrepancy between biochemical and cell-based activity has been observed for ENL YEATS domain inhibitor 92 (Moustakim et al., 2018). In summary, our study reports selective GAS41 inhibitors, which represent valuable chemical biology tools and molecular scaffolds for the development of a second generation of GAS41 inhibitors with increased activity in cancer models. The concept of the dimeric or bivalent inhibitors reported here may represent a more general approach to successfully block complex epigenetic protein–protein interactions.

RESOURCE AVAILABILITY

Lead contact

Requests for further information or reagents should be directed to the lead contact and corresponding author, Tomasz Cierpicki (tomaszc@umich.edu)

Material Availability

Bacterial strains and Antibodies were obtained from the commercial or academic sources described in the STAR methods Key resources table. Material generated in this study will be made available upon reasonable request.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-GAS41 monoclonal antibody	Santa Cruz	Catalog number: sc-393708
Rabbit anti-GAPDH monoclonal antibody	Cell Signaling	Catalog number: 3683
Bacterial Strains
Escherichia Coli: One Shot BL21(DE3) Competent Cells	Thermo Fisher Scientific, MA, USA	Catalog number: C600003
Escherichia Coli: BL21-CodonPlus (DE3)-RIL competent cells	Agilent, CA, USA	Catalog number: 230245
Escherichia Coli: Rosetta (DE3) Competent Cells	EMD Millipore, MA, USA	Catalog number: 70954
Chemicals, Peptides, and Recombinant Proteins
Compound 1	This work	N/A
Compound 2	This work	N/A
Compound 3	This work	N/A
Compound 4	This work	N/A
Compound 5	This work	N/A
Compound 6	This work	N/A
Compound 7	This work	N/A
Compound 8	This work	N/A
Compound 9	This work	N/A
Compound 10	This work	N/A
Compound 11	This work	N/A
Compound 12	This work	N/A
Compound 13	This work	N/A
Compound 14	This work	N/A
Compound 15	This work	N/A
Compound 16	This work	N/A
Compound 17	This work	N/A
Compound 18	This work	N/A
Compound 19	This work	N/A
FAM-H3K23crK27cr (FAM-QLAT[K-cr]AAR[K-cr]SAPA-NH2)	Genscript, NJ, USA	N/A; designed and special ordered for FP assay
H3K23crK27cr-biotin (AT[K-ac]AAR[K-ac]SAPG[K-Biotin])	Genscript, NJ, USA	N/A; designed and special ordered for AlphaScreen assay
Ni-NTA Agarose resin	Qiagen, Hilden, Germany	Catalog number: 30230
Glutathione Sepharose 4B GST-tagged protein purification resin	Cytiva, MA, USA (U.S. headquarters; formerly GE Healthcare Life Sciences)	Product: 17075601
PreScission Protease	Cytiva, MA, USA (U.S. headquarters; formerly GE Healthcare Life Sciences)	Product: 27084301
Q-Sepharose Fast Flow resin	Cytiva, MA, USA (U.S. headquarters; formerly GE Healthcare Life Sciences)	Product: 17051010
FuGENE® HD Transfection Reagent	Promega, WI, USA	Catalog number: E2311
Ammonium sulfate (15N2, 99%)	Cambridge Isotope Laboratories, MA, USA	Item number: NLM-713-50
Deposited Data
Crystal structure of 9 in complex with GAS41 YEATS	This work	7JFY
Crystal structure of GAS41 YEATS	Cho et al., 2018	5VNA
Experimental Models: Cell Lines
A549	ATCC	CCL-185
A549 GAS41-knockout cell line	This work	N/A
H1299 cell line	ATCC	CRL-5803
H1993 cell line	ATCC	CRL-5909
Recombinant DNA
pQE-80L-GAS4113–158	Cho et al., 2018	N/A
pGST-GAS411–148	Cho et al., 2018	N/A
pMA-GAS41-FL	Life Technologies, CA, USA	N/A; cDNA codon-optimized and synthesized by special order for this work
pMOCR-GAS41-FL	This work	N/A
pET32a-BirA	Gray et al., 2016	N/A
pET32p-Avi-GAS411–148	Cho et al., 2018	N/A
pET32a-GAS419–151	This work	N/A
pET32a-AF91–140	Gift from John Bushweller, Ph.D., University of Virginia	N/A
pHis-parallel2-ENL1–148	Gift from John Bushweller, Ph.D., University of Virginia	N/A
pMA-YEATS2201–350	Life Technologies, CA, USA	N/A; cDNA codon-optimized and synthesized by special order for this work
pGST-YEATS2201–350	This work	N/A
pBiT1.1-C[TK/LgBiT]	Promega, WI, USA	NanoBiT PPI Starter System
pBiT1.1-C[TK/LgBiT]-GAS41 WT	This work	N/A
pBiT1.1-C[TK/LgBiT]-GAS41 W93A	This work	N/A
pBiT1.1-C[TK/SmBiT]-H3.3	Promega, WI, USA	N/A; designed and special order for NanoBiT assay
Critical Commercial Assays
NanoBiT PPI Starter Systems	Promega, WI, USA	Catalog number: N2014
4D-Nucleofector™ System	Lonza Group AG, Basel, Switzerland	Catalog number: AAF-1002B
AlphaScreen Histidine (Nickel Chelate) Detection Kit	PerkinElmer, MA, USA	Part number: 6760619C
RNeasy Mini Kit	Qiagen, Hilden, Germany	Catalog number: 74106
High-Capacity cDNA Reverse Transcription Kit	Thermo Fisher Scientific, MA, USA	Catalog number: 4368814
Software and Algorithms
HKL2000	Otwinowski et al., 1997	https://hkl-xray.com/hkl-2000
CCP4	Winn et al., 2011	https://www.ccp4.ac.uk/
COOT	Emsley et al., 2010	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PHENIX	Adams et al., 2010	http://www.phenix-online.org/
MOLREP	Vagin et al., 2010	http://legacy.ccp4.ac.uk/html/molrep.html
MOLPROBITY	Williams et al., 2018	http://molprobity.biochem.duke.edu/
Origin 7.0	OriginLab, MA, USA	https://www.originlab.com/
Prism 8.0	GraphPad, CA, USA	https://www.graphpad.com/
TopSpin 2.1	Bruker, MA, USA	https://www.bruker.com/nc.html
NMRPipe	Delaglio, et al., 1995	https://www.ibbr.umd.edu/nmrpipe/index.html
SPARKY	T. D. Goodard and D. G. Kneller, UCSF	https://www.cgl.ucsf.edu/home/sparky/

Data and Code Availability

Crystallographic statistics is shown in Table 2 and coordinates have been deposited in the Protein Data Bank under accession number 7JFY. Software used for the biological studies were obtained from the commercial or academic sources described in the STAR methods Key resources table.

EXPERIMENTAL MODELS AND SUBJECT DETAILS

Protein expression in E. coli

Following BL21 strains were used in the studies: BL21(DE3) from Thermo Fisher Scientific, Rosetta (DE3) from Novagen and BL21 CodonPlus (DE3) RIL from Agilent. Cells were grown in LB (Luria-Bertani) or ¹⁵N labeled M9 minimal media to OD₆₀₀ between 0.6 and 0.8, induced at 16 °C for 16 h with 0.4 mM IPTG.

Cell line and cell culture

A549 (male, CLL-185), H1299 (male, CRL-5803), H1993 (female, CRL-5909) and 293T (female, CRL-3216) cells were obtained from ATCC and supplied mycoplasma free. A549, H1299 and H1993 cells were maintained in RPMI1640 (Gibco, 11875–119) with 10% fetal bovine serum (FBS) and 1% Penicillin and Streptomycin (P/S, Gibco, 15140–122). 293T cells were cultured in Dulbecco Modified Eagle medium (DMEM, Gibco, 11995–073) with 10% FBS and 1% P/S. All cells were grown at 37°C/5% CO2 incubator.

METHOD DETAILS

Protein expression and purification

The following constructs were used for these studies: GAS41 YEATS domain for the dimerization AlphaScreen and NMR (residues 13 – 158 in pQE-80L vector with N-terminal His₆-tag); GAS41 YEATS domain for Fluorescence Polarization and crystallization (residues 1 – 148 in pGST vector with N-terminal GST-tag and N-terminal TEV cleavage site); GAS41 YEATS domain for ITC (residues 9 – 151 in pET32a vector with N-terminal Trx-His₆-tag and N-terminal PreScission cleavage site); full-length GAS41 for AlphaScreen and TSA (synthetic gene encoding full length GAS41, synthesized by Life Technologies and sub-cloned into pMOCR vector with an N-terminal Mocr-His₆ tag); biotinylated-GAS41 YEATS domain for the dimerization AlphaScreen (obtained via sub-cloning of N-terminal Avi-tag and Trx-His₆-tag and N-terminal PreScission cleavage site into a modified pet32a vector with GAS41 YEATS residues 1 – 148); YEATS2 YEATS domain (synthetic gene encoding residues 201 – 350 from Life Technologies, subcloned into pGST vector); AF9 YEATS domain (residues 1 – 140 in pET32a vector, courtesy of Dr. John Bushweller PhD, University of Virginia); and ENL YEATS domain (residues 1 – 148 in pHis-parallel2 construct, courtesy of Dr. John Bushweller, PhD, University of Virginia). Proteins were expressed using E.Coli BL21(DE3) One Shot (ThermoFisher), BL21(DE3) CodonPlus RIL (Agilent), or Rosetta DE3 (Millipore Sigma) cells, in either LB or ¹⁵N M9 minimal media, and induced with 0.4 mM IPTG for 16 h at 18 °C. Cells were re-suspended in lysis buffer (50 mM Tris, pH 7.5, 300 mM NaCl, 1 mM TCEP, and 0.5 mM PMSF, with 10 mM imidazole added for His₆-tagged constructs), lysed using a cell disrupter, and applied to Ni-NTA (Qiagen) affinity column or a Glutathione Sepharose 4B column (Cytiva). Proteins were eluted using buffer containing 300 mM imidazole or 10 mM reduced glutathione, and dialyzed against buffer containing 50 mM Tris (pH 7.5), 150 mM NaCl buffer, and 1 mM TCEP. pGST-GAS41 YEATS (residues 1 – 148), pET32a-GAS41 YEATS (residues 9 – 151), pET32a-AF9 YEATS, and pGST-YEATS2 YEATS constructs were then proteolytically cleaved with TEV or PreScission proteases, and re-applied to Ni-NTA (Qiagen) or Glutathione Sepharose 4B (Cytiva) columns to extract Trx-His₆- or GST-affinity tags, respectively.

For crystallization experiments, GAS41 YEATS (residues 1 – 148) was applied to S75 column for size exclusion chromatography purification, and exchanged into buffer containing 20 mM Tris (pH 7.5) and 300 mM NaCl. For AlphaScreen experiments, the construct encoding biotinylated GAS41-YEATS domain (residues 1 – 148) was co-expressed in E. Coli BL21(DE3) cells with BirA biotin ligase as previously described (Gray et al., 2016) and purified in the same manner as for His₆-tagged GAS41 constructs. ENL YEATS domain was purified through refolding: cells were harvested and lysed in buffer containing 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM TCEP, and 0.5 mM PMSF, and washed repeatedly with buffer containing 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM TCEP, 0.5 mM PMSF, and 1% Triton X-100. Inclusion bodies were refolded in 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM TCEP, 0.5 mM PMSF, and 6 M Guanidine Hydrochloride. Solubilized protein was exchanged into refolding buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM TCEP) and soluble protein was purified via Ni-NTA affinity chromatography in the same manner as for His₆-tagged GAS41 constructs.

NMR binding experiments

NMR ¹H-¹⁵N HSQC experiments were performed at 30 °C with 60 μM GAS41, AF9 or ENL YEATS domains in buffer containing 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM TCEP, 7% D₂O, and 5% DMSO. YEATS2 YEATS domain was prepared in buffer containing 50 mM HEPES, pH 7.5, but otherwise conditions were identical. Compounds were added from DMSO stocks to final concentrations indicated in the figure labels. All NMR spectra were acquired on a 600 MHz Bruker Avance III spectrometer equipped with cryoprobe running Topspin version 3.7. Processing and spectral visualization were performed using NMRPipe and Sparky.

Fragment screening by NMR Spectroscopy

Fragment-screening samples were prepared with 60 μM ¹⁵N-labeled GAS41 YEATS (residues 13–158) in buffer containing 50 mM Tris, pH 7.5, 200 mM NaCl, 5% DMSO, and 7% D₂O. The library of fragments used for screening was a combination of both commercially available and in-house synthesized compounds. Fragments were screened in mixtures of 10 compounds per sample at 500 μM final concentration per compound. Hit confirmation was performed with individual compounds at 1000 μM. ¹H-¹⁵N HSQC spectra were acquired at 30 °C on a 600 MHz Bruker Avance III spectrometer equipped with cryoprobe running Topspin version 2.1. Processing and spectral visualization were performed using NMRPipe (Delaglio et al., 1995) and Sparky (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco).

Fluorescence Polarization Assay

5’ 6-Fluorescein (FAM)-labeled di-crotonylated Histone H3 peptide probe H3K23crK27cr (FAM-QLAT[K-cr]AAR[K-cr]SAPA-NH2) was synthesized by Genscript. For competition experiments, 1 μM GST-GAS41(1–148) protein was incubated with competitor dilutions at 1% DMSO in assay buffer containing 50 mM Tris pH 7.5, 150 mM sodium chloride, 1 mM TCEP, 0.01% BSA, and 0.01% Tween-20 for 1 h in a 96-well black plate (Corning). 25 nM FAM-H3K23crK27cr peptide was added and the plate was incubated for 1 h before fluorescence polarization data was measured at 525 nM on a Pherastar plate reader (BMG Labtech). Data was fit in Prism 7.0 (GraphPad), and IC₅₀ values were derived using the equation log(inhibitor) vs. response – Variable slope (four parameters).

ITC experiments

GAS41 (residues 9–151), GST-GAS41 (residues 1–148), AF9 and ENL YEATS domains were dialyzed repeatedly against 50 mM phosphate buffer, pH 7.5, and 150 mM NaCl (ITC Buffer) at 4 ^oC. Compounds were dissolved in DMSO, and compound and protein solutions were prepared at 200–400 μM and 8–40 μM in ITC buffer, respectively, with 1% DMSO. The calorimetric cell was titrated with compound in 10–20 μL aliquots at 300 s intervals. Titrations were performed using a VP-ITC calorimetric system (MicroCal) at 25 ^oC. Data were analyzed by Origin 7.0 (OriginLab) to obtain K_D and stoichiometry.

Thermal shift assay

Thermal denaturation of MOCR-his₆-GAS41 (full-length) and His₆-GAS41 YEATS (residues 13–158) were tested at 5 μM protein concentrations in 50mM HEPES, pH 7.5, 150mM NaCl, 1mM TCEP buffer with SYPRO Orange dye using dilution recommended by the manufacturer (Invitrogen S6650). Compounds were tested at final 1% DMSO. After 1 h incubation, samples were dispensed as technical quadruplets on a PCR plate (BioRad hard, black shell, 96, clear well plates). Melting temperature was measured using a Biorad CFX96 Real-time System, C1000 Touch Thermal Cycler in FRET scan mode, and temperature was increased from 10°C to 95°C over 1 h. Data were analyzed using BioRad CFX Manager 3.1.

Crystallization and structure determination

GAS41 YEATS (residues 1 – 148) at 7.3 mg mL⁻¹ (400 μM) was incubated with 450 μM compound 9 for 6 h at room temperature in 20 mM Tris, pH 7.5, and 300 mM NaCl. Initial crystals of GAS41 YEATS-inhibitor 9 complex were obtained via screening using the hanging-drop vapor diffusion technique over one week at 4 °C, and optimization yielded final crystallization conditions of 0.1 M Tris, pH 7.3, 0.2 M MgCl₂, and 2.2 M NaCl. Prior to data collection, crystals were exchanged into cryoprotectant solution, comprised of precipitant solution with 750 μM 9, 1% DMSO, and 25% ethylene glycol, and frozen in liquid nitrogen. Diffraction data for GAS41 YEATS-inhibitor 9 complex were collected at the 21-ID-F beam line at the Life Sciences Collaborative Access Team at the Advanced Photon Source. Data were integrated and scaled using HKL-2000 (Otwinowski and Minor, 1997), and the structure was solved by molecular replacement with MOLREP (Vagin and Teplyakov, 2010) in CCP4 suite (Winn et al., 2011), using native-GAS41 YEATS structure as a search model (PDB 5VNA). Further model building was performed in COOT (Emsley et al., 2010), and structure was refined using REFMAC (Murshudov et al., 1997) and PHENIX program suite (Adams et al., 2010). Structures were validated using MOLPROBITY (Williams et al., 2018) and ADIT servers. The coordinates of the GAS41 YEATS in complex with 9 have been deposited in the Protein Data Bank under accession number 7JFY.

AlphaScreen assays

Competition experiments were performed using full-length GAS41. 100 nM MOCR-his₆-GAS41 was incubated with competitors (compounds or peptides) diluted 100-fold from DMSO stocks into 50 mM HEPES pH 7.5, 100 mM NaCl, 1 mM TCEP, 0.05% BSA, 0.01% Tween-20 for 1 h at 1% DMSO in a 96-well ½-Area AlphaPlate (PerkinElmer). H3K23crK27cr-biotin was added to a final concentration of 25 nM and incubated for 1 h. Nickel Chelate Acceptor AlphaScreen beads (PerkinElmer) were added to a final concentration of 10 μg/mL and incubated for 1 h. Streptavidin Donor beads (PerkinElmer) were added to a final concentration of 10 μg/mL and incubated for 2 h. All AlphaScreen assay incubations were performed at room temperature. Luminescence signal was measured on a Pherastar plate reader (BMG Labtech).

For dimerization experiments, 500 nM his₆-GAS41(13–158) and 250 nM Avi-GAS41-YEATS were incubated in 50 mM HEPES pH 7.5, 100 mM NaCl, 1 mM TCEP, 0.05% BSA, 0.01% Tween-20 in a 96-well ½-Area AlphaPlate (PerkinElmer) for 0.5 h. Subsequently, monomeric (9, 11) and dimeric (18, 19) compounds were re-suspended in DMSO and diluted 100-fold to a final concentration of 250 nM at 1% DMSO. AlphaScreen beads were added as in competition experiments and luminescence signal was measured on a Pherastar plate reader (BMG Labtech).

For competition experiments with the dimeric complex, 500 nM his₆-GAS41 (residues 13–158) and 250 nM Avi-GAS41-YEATS were incubated in assay buffer with 100 nM dimeric inhibitor 18 for 0.5 h and followed with addition of competitor 9 in final 1% DMSO. AlphaScreen beads were added as in competition experiments and luminescence signal was measured on a Pherastar plate reader (BMG Labtech).

NanoBiT assay

To detect inhibition of protein-protein interaction in cells by compounds the NanoBiT assays (N2014, Promega) were processed according to the manufacturer’s instructions. Briefly, GAS41-WT and GAS41-W93A mutant were cloned into pBiT1.1-C[TK/LgBiT] vector and verified by sequencing. SmBiT-H3.3 were purchased from Promega. 293T cells (4 × 10⁵) were plated into 6-well plates (DMEM with 10 % FBS) and incubated for 5 h. The LgBiT-GAS41 and SmBiT-H3.3 plasmids were co-transfected using FuGENE HD (E2311, Promega) for 42 h. 5 × 10⁴ cells were transferred into 96-well white plates (DMEM with 10 % FBS and 1% Penicillin and Streptomycin) and treated with compounds for 24 h. After the Nano-Glo Live Cell Reagent was added to each well, the luminescence was measured immediately using PHERAstar reader (BMG Labtech).

Generation of A549 GAS41 knockout cell line

A549 GAS41 knockout cell line was generated using Gene Knockout Kit (Synthego) according to manufacturer’s protocol. Briefly, the A549 cells were transfected with GAS41 sgRNA-CAS9 ribonucleoprotein (RNP) complex using Lonza 4D-Nucleofector™ system. GAS41 sgRNA (ATAGTTTACGGTAATGTTGC) was designed in exon 2 of GAS41 and synthesized from Synthego. For clonal isolation of knockout single cell was plated by limiting dilution on 96 well plates. The GAS41 level of single clone was checked by Western blot. The cells were lysed by RIPA buffer and supernatants were collected after centrifuge. The concentration of cell lysates was measured by BCA assay (Invitrogen). The same amounts of cell lysates were run in SDS-PAGE and transferred on PVDF membrane (GE). After blocking in skim milk, the membrane was probed with anti-GAS41 antibody (sc-393708, Santa Cruz) and anti-mouse secondary antibody. The luminescence with ECL reagents (Kwik Quant) was detected by Kwik Quant imager.

Cell proliferation assay

For cell proliferation assay H1299 cells (1 × 10³) were seeded into 12-well plate to get around 5% confluence next day in RPMI with 10 % FBS and 1 % Antibiotic-Antimycotic. Next day, the media was changed to RPMI with 2 % FBS and 1 % Antibiotic-Antimycotic and 11 or 19 was treated in final DMSO concentration at 0.25 %. During 4 days of incubation with the compounds the cell confluence was analyzed by the IncuCyte or count cells in the end of incubations.

For growth curves 2.5 × 10⁴ of A549, H1299, and H1993 cells were cultured in RPMI with 10% FBS and 1% penicillin/streptomycin into 12-well and treated with 12, 6, or 3 μM 19 in final DMSO concentration at 0.25 %. The cells were counted and re-plated with fresh medium and compounds each 3 or 4 days by 14 days. Growth curved were plotted using cell numbers at days 0, 4 (or 3), 7, 11 (or 10), and 14. H1299 cells treated for 7 days were harvested for qRT-PCR analysis.

CETSA assay

H1299 cells were treated with DMSO or 12 μM 19 for 24 h. Treated cells were collected, divided and heated at different temperatures for 3 min. Subsequently, the cells were lysed using RIPA buffer and sonicated. Lysates were centrifuged at 15,000 rpm for 10 min at 4 °C and mixed with4x Laemmli sample buffer (Bio-rad / 161–0747). The samples were analyzed by SDS-PAGE followed by Western blot using GAS41 (Santa Cruz, sc-393708) and GAPDH (Cell Signaling, 3683) primary antibodies. The membrane with ECL reagents (Kwik Quant) was detected by ChemiDoc™ Touch Imaging System (Bio-Rad).

Quantitative RT-PCR (qRT-PCR)

H1299 cells were collected at 7 days of treatment with 19 in cell proliferation assay. RNA was isolated according to RNeasy Mini Kit protocol (74106, Qiagen) and quantified Nanodrop 2000 UV-V. 500 ng of total RNA was conducted for reverse transciption PCR (RT-PCR) using High-Capacity cDNA Reverse transcription Kit (4368814, Applied Biosystem). SYBR Green Master Mix for real-time PCR were purchased from Applied Biosystem. Real-time PCR was processed using the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Relative quantification of each gene transcript was carried out using the comparative Ct method as described in the Applied Biosystems User Bulletin no. 2. The primers were E2F2 forward (5’- GCC TAT GTG ACT TAC CAG GAT ATC C −3’), E2F2 reverse (5’-CCT TGA CGG CAA TCA CTG TCT −3’), FOXM1 forward (5’- GCA GCA TCA AGC AAG AGA TG −3’), FOXM1 reverse (5’-GCC GCT CAG ACA CAG AGT TC −3’), MCM6 forward (5’-CGT CTG GAA AAA AGC GAC TTG −3’), MCM6 reverse (5’-TGC TTA GTG CCG AGG ATT CG −3’) (Hsu et al., 2018), GDF15 forward (5’-CAA CCA GAG CTG GGA AGA TTC G −3’), GDF15 reverse (5’-CCC GAG AGA TAC GCA GGT GCA-3’), PEG10 forward (5’- ACC ACC AGG TAG ATC CAA CCG A −3’), PEG10 reverse (5’- TGT CAG CGT AGT GAC CTC CTG T −3’), SEMA3C forward (5’- ACC CAC TGA CTC AAT GCA GAG G −3’) and SEMA3C reverse (5’- CAG CCA CTT GAT AGA TGC CTG C −3’).

Chemical Synthesis and Characterization

Details of chemistry methods and characterization of compounds is included in Supplemental Information (Methods S1).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses and number of independent biological experiments are presented in the figure legends. The IC50 values from Fluorescence Polarization and AlphaScreen assays were determined in Prism 8.0 (Graphpad) software from plots of log[inhibitor] and % inhibition using the nonlinear regression, variable slope model. IC50 values are reported as mean +/− standard deviation from two independent experiments. Analysis of ITC data was performed using Origin 7.0 (OriginLab) software and K_D, stoichiometry, ΔEnthalpy, and TΔEntropy values are reported as mean +/− standard deviation with n number of independent experiments denoted in the table legend. Thermal shift assay EC50 values are reported as mean +/− standard deviation from two independent experiments analyzed using Prism 8.0 (Graphpad) software. The statistical tests for viability at 14 days, qRT-PCR and NanoBiT were calculated using Prism 8.0 (Graphpad) software. Two-tailed t-tests were used for comparisons. Error bars are plotted as standard deviation. The quantification and data standardization for qRT-PCR were using the ΔΔCT method.

Significance.

YEATS domain proteins are readers of histone proteins with acetylated or crotonylated lysine side chains and are involved in the formation of epigenetic complexes regulating gene expression. Epigenetic reader proteins are frequently dysregulated in cancer and represent attractive targets for development of small molecule inhibitors. GAS41 (Glioma amplified sequence 41) is an emerging oncogene overexpressed and implicated in non-small cell lung cancer (NSCLC). Here, we developed small molecule inhibitors of GAS41 by employing a fragment screening approach combined with medicinal chemistry. Structural analysis revealed that these inhibitors bind to GAS41 YEATS domain in a channel involved in the recognition of the acylated lysine side chain. However, because of the small size of the binding site, development of very potent inhibitors proved challenging. Since full length GAS41 is a dimeric protein, we developed dimeric inhibitors that simultaneously interact with two YEATS domains. Our lead compounds bind to dimerized GAS41 YEATS domains with nanomolar affinity, demonstrating two orders of magnitude enhancement in potency over monomeric inhibitors. The most potent inhibitor we developed, engages GAS41 in cells and blocks its interaction with acetylated histone H3, leading to growth inhibition in NSCLC cells. This work demonstrates advantages of bivalent inhibitors as an approach for targeting challenging protein-protein interactions.