Polycomb paralog chromodomain inhibitors active against both CBX6 and CBX8

Abstract

Methyllysine reader proteins bind to methylated lysine residues and alter gene transcription by changing the compaction state of chromatin or by the recruitment of other multiprotein complexes. The polycomb paralog family of methyllysine readers bind to trimethylated lysine on the tail of histone 3 via a highly conserved aromatic cage located in their chromodomains. Each of the polycomb paralogs are implicated in several disease states. CBX6 and CBX8 are members of the polycomb paralog family with two structurally similar chromodomains. By exploring the structure-activity relationships of a previously reported CBX6 inhibitor we have discovered more potent and cell permeable analogs. Our current report includes potent, dual-selective inhibitors of CBX6 and CBX8. We have shown that the −2 position in our scaffold is an important residue for selectivity amongst the polycomb paralogs. Preliminary cell-based studies show that the new inhibitors impact cell proliferation in a rhabdoid tumor cell line.

Graphical Abstract

Dual active CBX inhibitor: A library of peptidomimetic based inhibitors for chromobox homolog 6 and 8 (CBX6/8) was developed. Through SAR key residues important for selectivity and potency were identified. We arrived at a dual active CBX6/8 inhibitor with sub micromolar affinity and 18-fold selectivity over CBX7. Preliminary cell studies show our compounds are cell permeable and have an effect on cell proliferation.

Introduction

Post-translational modifications (PTMs) can be used as epigenetic signatures that control and regulate gene expression pathways. Lysine methylation is one such PTM that is important in several malignancies.^[1] Lysine methylation on histones, which package and order DNA, can activate or repress transcription of the associated genes. Methyllysine ‘reader’ proteins that recognize this mark form multi-protein complexes that regulate fundamental pathways critical for health and development.^[2]

Chromodomains (chromatin organization modifier domains) are a family of methyllysine readers that include proteins belonging to the Heterochromatin protein-1 (HP1) and polycomb sub-families. Both families recognize higher methylation states on H3K9 and H3K27.^[3] The polycomb family of methyllysine readers consists of five human paralogs: CBX2, CBX4, CBX6, CBX7, and CBX8. The human CBX polycombs are important regulatory proteins that are responsible for cellular differentiation during development, stem cell maintenance via transcriptional repression, and cancer progression.^[4] Each member can participate in different versions of the multi-protein complex called polycomb repressive complex 1 (PRC1). Each is functionally unique,^{[4b, 5]} and while the biological data are growing rapidly the phenotypes of chemical inhibition are only recently beginning to be discovered.

The challenge in targeting the polycomb paralogs with selective chemical inhibitors is due to the high degree of sequence similarity within the chromodomains. The polycomb paralog chromodomains all possess a conserved aromatic cage made up of 3–4 aromatic residues that recognize the side chain of Kme3. They achieve selectivity by recognizing residues adjacent to the Kme3 in a “β-groove” that engages the Kme3-containing partner in a beta-sheet like structure. The main differences between polycomb chromodomains are relatively small variations in the side chains that line the β-groove.

CBX7-selective inhibitors are by now well known,^[6] but other CBX’s are also arising as promising targets. CBX6 and CBX8 are both involved in regulating stem cell differentiation. CBX6 is an essential regulator of embryonic stem cells (ESCs).^[7] Depletion of CBX6 in ESCs leads to rapid cell differentiation. Cell-based studies with CBX6 mutants lacking the two methyllysine-binding tryptophan residues in the aromatic cage showed that the function of CBX6 in ESCs is strongly linked to the methyllysine reading function of the chromodomain.^[7] Unlike CBX7 (and apparently all other human polycomb paralogs), CBX6 does not act through binding H3K27me3.^{[3, 8]} Ablation of the H3K27 methyl-depositing PRC2 complex in ESCs did not change CBX6 recruitment to chromatin. CBX8 plays a critical role in dictating cell fate decisions and activation of differentiation genes.^[9] CBX8 acts as part of the PRC1 complex and is upregulated during ESC differentiation. CBX7 in ESCs acts to maintain pluripotency and is downregulated upon increased expression of CBX8 during differentiation.^{[4b, 5b, 9]}

Both CBX6 and CBX8 have arisen as potential targets in oncology. Overexpression of CBX6 promotes tumor growth and is linked to poor patient outcomes in hepatocellular carcinoma (HCC).^[10] In vitro studies in multiple HCC cell lines showed increased CBX6 expression to promote cell growth whereas knockdown of CBX6 showed reductions in proliferation. In vivo studies also showed CBX6 overexpression to promote HCC tumour growth.^[10] CBX6 is also upregulated in rhabdoid tumours, an aggressive pediatric cancer.^[11] The oncogenic role of CBX8 has been demonstrated in several malignancies including hepatocellular carcinoma (HCC),^[12] acute myeloid leukemia (AML),^[13] breast cancer^[14] and esophageal carcinomas.^[15] CBX8 is also overexpressed in bladder cancer,^[16] choriocarcinomas,^[17] and glioblastoma multiforme.^[18]

Isolated examples of selective inhibition of CBX6 and CBX8 have recently been reported. A peptidic CBX6 inhibitor (3) developed by structure-based design achieved selectivity over the other polycomb paralogs due to installation of a bulky group directed into protein’s (−2) pocket, which is larger in CBX6 than other polycombs.^[19] In separate work, a DNA-encoded library showed that CBX8-selective inhibition was achievable by combining large groups in the (−2) pocket with other changes in the ligand structure along the β-groove.^[20] Similarly a CBX8-selective positive allosteric modulator was found by incorporating larger groups in the (−2) pocket along with other structural changes.^[21]

In this work, we define the structure-activity relationships (SAR) of CBX6 inhibition, and also report a second generation of CBX6 inhibitors with improved cell permeability. Fluorescence polarization assays were employed to determine the compounds’ IC₅₀ values. The lower limit of the competitive fluorescence polarization assay is approximately the K_d value for the probe-protein complex.^[22] The K_d of the probe (2) is 4750 nM for CBX1, 300 nM for CBX2, 200 nM for CBX4, 47 nM for CBX6, 12 nM for CBX7 and 624 nM for CBX8.^{[19, 23]}{Milosevich, 2020 #38} Compounds with an IC₅₀ value lower than the probe’s K_d are approximate values as the IC₅₀ is outside the linear range of the assay. NMR characterization was used on a subset of compounds to show the solid phase synthesis approach used does not cause epimerization (Figure S24–38).^[6f] As we modified ligands in the β-groove region, we arrived at a novel class of dual-active CBX6- and CBX8-inhibiting ligands and identified the structural features that give rise to this dual selectivity. We provide preliminary biological data that demonstrate that the inhibitors have functional effects in a cancer cell line.

Results

We started our efforts with modifications to previously reported CBX6-targeting peptidic inhibitors,^{[19, 23]} and chose to screen each compound against CBX1/6/7/8. This panel was chosen in order to efficiently obtain data on CBX6- and CBX8-selectivity over CBX7, which is the CBX that intrinsically binds ligands most strongly. It also includes CBX1 (HP1β) as a representative of the HP1 class of chromodomains. Our previous report on CBX6 inhibition included the initial scaffold 1, the promiscuous inhibitor 2, CBX6-selective inhibitor 3 that is 6–90-fold selective for CBX6 over CBX1, 7 and 8, 4 the dye-free analog of 2; and compound 5 the dye-free analog of 3 (Figure 1).^{[19, 23]} As a first step towards making lower molecular weight compounds, we investigated 1, which is an analog of 2 that lacks the C-terminal Glu residue and dye label (Figure 1, Table 1). Compound 1 is 10-fold selective for CBX7 over CBX1 and 6- and 29-fold selective over CBX6 and CBX8, respectively.

Figure 1.

Chemical structures of compounds 1–6.

Table 1.

IC₅₀ values for compounds 1, 4, 5 and 6 (μM).^[a]

Compound	CBX1	CBX6	CBX7	CBX8
1	14[b]	7.7[b]	1.4[b]	41[b]
4	N.T.[c]	5	1.8	28
5	N.T.[c]	18	210[b]	17[b]
6	>500[b]	8.5	>500	14[d]

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

^[b]Competitive FP assays performed in duplicate

^[c]Not tested

^[d]Shallow FP response. See SI.

To improve the cell-permeability and drug-like properties of our inhibitors we first focused on understanding the roles of the dye and the (−2) substituent in CBX6/8 selectivity (Table 1). Substituting the alanine in compound 4 to a valine in compound 5 produced a >100-fold decrease in binding to CBX7. Removal of the FITC-labeled lysine on the scaffold containing the (−2) valine residue (2 compared to 4) resulted in a 20-, 40- and 3-fold decrease in binding to CBX6, CBX7 and CBX8, respectively.^{[19, 23]} We knew from previous reports that removal of the FITC dye label makes these compounds significantly less potent. From the current experiments we can also see that the FITC removal affects CBX6 and CBX8 binding differently. Although 5 is still selective for CBX6 over CBX7, it is no longer selective for CBX6 over CBX8. Compound 5 is equipotent against CBX6/8 (IC₅₀ values of ~17 μM) and is 12-fold selective over CBX7. Compound 6, in which an isoleucine residue is present at the (−2) position, displayed no measurable binding to CBX1 or CBX7 (>500 μM). Compound 6 displayed similar potency for CBX6 and CBX8 (IC₅₀ values of 8.5 and 14 μM).

We focused next on optimization of compound 5 by removing the quaternary ammonium ion (Kme3) and the Glu residue, which we hypothesized would each cause poor cell permeability. We removed the C-terminal glutamic acid residue to reduce overall charge, leading to compound 7, and made an analogous compound 8 that swapped trimethyllysine (Kme3) for diethyllysine (Ket2) (Figure 2).^[6f] Removal of the C-terminal Glu gave decreased binding to CBX7 and no significant change in potency was observed for CBX6/8 (Table 2). Replacement of Kme3 with Ket2 gave a 7-fold increase in binding to CBX6, and no change in binding to CBX8. Compound 8 is 6-fold selective for CBX6 over CBX8 and displays no binding to CBX1/7. We next sought to substitute the (−1) residue in 8 with a more hydrophobic leucine residue (9, Figure 3). Compound 9 showed a slight decrease in potency for CBX6 and an increase in binding to CBX7 and CBX8 (Table 3). Neither compound showed activity against CBX1.

Figure 2.

Structure of compounds 7 and Ket2 analog 8

Table 2.

Binding affinity for compounds 7 and 8 (IC₅₀ values in μM).^[a]

Compound	CBX1	CBX6	CBX7	CBX8
7	> 500	24	> 250	18
8	> 500	3.2	> 500	18

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

Figure 3.

Chemical structures of compounds 8–12 with (−1) and N-cap substitutions.

Table 3.

Binding affinity for compounds 8–12 (IC₅₀ values in μM).^[a]

	R (N)	R (−1)	CBX1	CBX6	CBX7	CBX8
8			> 500	3.2	> 500	18
9			> 300	9.8[b]	23[b][c]	2.7
10			> 300	72	162	68
11			> 300	12	105	6.5
12			> 500	17	>300	6.2

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

^[b]Competitive FP assays performed in duplicate

^[c]Poor fit. See SI.

We next explored the roles of N-terminal capping (N-cap) residues on CBX6/8 selectivity, following leads from our collaborative studies of CBX8-selective inhibitors generated by the Krusemark lab.^[20] Using the scaffold with leucine at the (−1) position (9), we substituted the N-cap residue with two distinct heterocycles. Substitution of the p-bromobenzamide group with 1H-pyrazole-3-carboxamide (10) resulted in a 7-fold decrease in binding to CBX6 and a 25-fold decrease in binding to CBX8, reminiscent of results previously seen with a DEL (Table 3).^[20a] Substitution with the 5-methylisoxazole-3-carboxamide (11) gave a similar binding profile as 9 with CBX6/8 and showed activity against CBX7 with an IC₅₀ value of 105 μM (Table 3). A similar profile was observed for a 5-methylisoxazole N-cap ligand with a tyrosine residue at the (−1) position (12).

To explore the SAR in the b-groove we combined the N-cap heterocyclic substitutions with aromatic and hydrophobic rings in the (−3) position (Figure 4, Table 4). Leucine was held constant as the (−1) residue. Replacement of the phenylalanine residue in 10 with a cyclohexyl group (13) showed a slight 2–3 fold increase in binding to CBX6, 7 and 8. The same substitution in the 5-methylisoxazole-3-carboxamide containing 11 to give compound 14 did not significantly change binding to CBX6 and CBX8 but did show a decrease in binding to CBX7 (Table 4). Exchanging the (−3) phenylalanine (10, 11) for a cyclohexyl (13, 14) showed a non-additivity in the SAR. When the N-cap was a methylisoxazole this exchange at the (−3) position caused a decrease in CBX7 affinity, while when the N-cap was a pyrazole the affinity for CBX7 increased.

Figure 4.

Overall chemical structures of compounds 10, 11 and 13–19 with (−3) and N-cap substitutions. See Table 4 for compound numbers

Table 4.

Binding affinity for compounds 10, 11 and 13–19 (IC₅₀ values in μM).^[a]

	R (N)	R (−3)	CBX1	CBX6	CBX7	CBX8
10			> 300	72	162	68
13			> 300	32	104	25
11			> 300	12	105	6.5
14			> 300	8.3	171	5.8
15			> 500	12	>250	0.6[b]
16			N.T.[c]	9.3	>500	1.8
17			N.T.[c]	68	>500	20
18			> 500	33	>250	17
19			N.T.[c]	5.0	209	1.6

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

^[b]Estimated value because IC₅₀ is outside of linear range of assay.^[22–23]

^[c]Not tested

Modifying the aromatic residues at the (−3) position while holding the N-cap methylisoxazole constant showed a decrease in binding to CBX7. Replacement of the phenyl moiety (11) with a 4-methoxyphenyl group (15) resulted in no significant change in binding to CBX6. CBX7 binding was decreased while CBX8 was increased 11-fold. Substituting the methoxy for fluorine (16) and iodine (17) showed no binding to CBX7. The p-fluorophenyl (16) analog showed no significant change in CBX6 and 3-fold decrease in CBX8 binding, while the p-iodophenyl (17) showed a 6- and 33- fold decrease in binding to CBX6 and CBX8. Exchanging the phenyl moiety (11) with a pyridine-3-yl group at the (−3) position decreased binding to CBX6/8 by a factor of 2–3 and abolished binding to CBX7 (18, Table 4). Replacing the pyridine-3-yl with a thiophene (19) showed a 7–10 fold increase in CBX6/8 binding.

Substitution of the methylisoxazole N-cap gave large improvements in potency. We expanded the protein panel to include CBX2 and CBX4. CBX2 has a highly similarly −2 pocket compared to CBX6 and CBX8. CBX4 is structurally analogous to CBX7. We synthesized compounds 20 and 21, containing a propyl substitution in the 5-position of the isoxazole ring (Figure 5, Table 5). The 5-propylisoxazole N-cap combined with the (−3) cyclohexyl moiety (20) displayed potent binding to CBX6 with an IC₅₀ value of 2.8 μM and is 14- and 3-fold selectivity over CBX7 and CBX8. 5-propylisoxazole (R_N) with a phenylalanine (−3) and threonine (−1) lead to a potent CBX8 inhibitor (21) with an IC₅₀ of 2.2 μM. This inhibitor was >100-fold selective over CBX2, CBX4, and CBX7, and 3-fold selective over CBX6. The 5-phenylisoxazole N-cap substitution in compounds 22 and 23 led to the most potent inhibitors. The addition of a 5-phenylisoxazole N-cap substitution with a (−3) phenyl group provided 22, which is the most potent compound of the series (Table 5). Compound 22 is equipotent for CBX6 and CBX8 with IC₅₀ values of 200 nM, and is 18- to 50-fold selective over CBX2, CBX4 and CBX7. We then swapped the (−1) leucine for a tyrosine to obtain compound 23. Compound 23 was one of the more potent compounds however it had lost significant selectivity against CBX7. Substitution of the methyl group in the 5-methylisoxazole N-cap with larger alkyl chains and aromatic groups improves potency due to increased engagement with the extended β-groove of CBX6/7/8.

Figure 5.

Chemical structure of peptidic ligands 20–23 with isoxazoles, (−1) and (−3) substitutions.

Table 5.

Binding affinity for compounds 20–23. (IC₅₀ values in μM).^[a]

Compound	CBX1	CBX2	CBX4	CBX6	CBX7	CBX8
20	> 300	N.T.[b]	N.T.[b]	2.8	40	7.5
21	N.T.[b]	246	330	7.5	268	2.2
22	> 500	6.8	9.8	0.2	3.6	0.2[c]
23	N.T.[b]	21[d]	18[d]	2.2	5.4	1.0

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

^[b]Not tested

^[c]Estimated value because IC₅₀ is outside of linear range of assay.^[22–23]

^[d]Shallow FP curve. See SI.

We aligned structures of CBX6, CBX7, and CBX8 to learn more about the SAR of these inhibitors. We chose structures of CBX6 (3I90) in complex with histone tail peptide H3K27me3,^[3] along with CBX7 (5EPJ) and CBX8 (5EQ0) in complex with UNC3866,^[6f] which is analogous to the inhibitors in this paper. The ligands for CBX6 and CBX7 were hidden and UNC3866 complexed with CBX8 was used in the analysis. The p-(t-butyl)benzamide N-capping group of UNC3866 sits in a surface groove defined by D50, R52, L53, which are conserved in CBX6, CBX7 and CBX8 (Figure S62). The aligned crystal structures show no large structural differences in this region, even though we know that ligand differences in this region are prime determinants of selectivity. The main variation in the β-groove comes from residue 51, which diverges significantly between CBX7 (P51) vs. CBX6 (S51) or CBX8 (A51). This residue is directed away from the N-cap substituent, but may cause the N-cap binding pocket to change shape in ways that significantly influence ligand binding. The impact of such subtle differences on binding would be hard or impossible to predict, but it is clear empirically that the isoxazoles in the N-cap pocket are strong determinants of potency and selectivity due to subtle differences in the proteins’ β-groove.

Larger hydrophobic substitutions in the (−2) position of the ligand allow for CBX8 selective inhibitors. We synthesized analogs of compound 11 containing a cyclopropyl, cyclopentyl and sec-butyl group at the (−2) position (Figure 6). Compound 24 containing a cyclopropyl group was nearly equipotent for CBX6/8 and displayed between 4- and 7-fold selectivity over CBX7 (Table 6). Compared to the isopropyl-containing compound (11), the cyclopropyl substitution increased binding to CBX2/4/6/7/8 by factors of 2–22. Replacing the cyclopropyl with a cyclopentyl group (25) dramatically reduced binding to CBX6/7 and decreased binding to CBX8 by a factor of 6 (IC₅₀ of 18 μM, Table 6). Compound 25 is 10x and 20x selective for CBX8 over CBX6 and CBX7 respectively. Addition of a sec-butyl group (26) further decreased binding to CBX6 and abolished binding to CBX7 at the concentrations tested. All CBX polycomb paralogs tested showed decreased binding with increasing size of the substituent at the (−2) position.

Figure 6.

Chemical structure of peptidic ligands 11, 24–26 with substitutions at the (−2) position.

Table 6.

Binding affinity for compounds 11, 24–26. (IC₅₀ values in μM).^[a]

	R (−2)	CBX1	CBX2	CBX4	CBX6	CBX7	CBX8
11		> 300	116	210	12	105	6.5
24		> 500	9.2	9.5	5.0	22	3.0
25		> 500	N.T.[b]	N.T.[b]	178	372[c]	18
26		> 500	N.T.[b]	N.T.[b]	262	> 500	48

^[a]Competitive FP assays performed in triplicate unless stated otherwise. Raw data and asymmetric 95% confidence intervals are reported in SI.

^[b]Not tested

^[c]Shallow FP curve. See SI.

CBX8 was best able to tolerate larger substitutions directed into the (−2) pocket of the protein. We and others have reported CBX8-selective ligands that engage the large size (−2) pocket within CBX8 compared to other members of the polycomb paralog CBX family.^[20–21] The SAR reported demonstrates the (−2) pocket of CBX8 to be capable of engaging and binding to ligands with bulkier (−2) substitutions. The residues lining the (−2) pocket are nearly identical in CBX6/8 with the exception of an isoleucine residue in CBX6 that is replaced with a leucine in CBX8. Compound 25, containing a cyclopentyl group at the (−2) position displays promising selectivity for CBX8. We predict that combining the (−2) cyclopentyl group with other substitutions could improve the potency and selectivity of 25 for CBX8.

CBX6 is upregulated in rhabdoid tumours,^[11] a highly aggressive pediatric malignancy driven by loss of SMARCB1 subunit of the SWI/SNF chromatin remodeling complex.^{[11, 24]} We used immunoblot analysis to profile CBX6 expression in a panel of kidney cell lines, including non-transformed epithelial cells (HK-2), clear cell renal cancers (Caki1, Caki2, 786O), kidney adrenal caner (SW-13) and kidney rhabdoid tumor cell line G401, as well as muscle rhabdoid cell line A204. There was no detectable CBX6 protein in the majority of the cell lines, while the kidney rhabdoid line G401 displayed a high expression level of CBX6, consistent with patient transcriptional data (Figure S63a). We used CRISPRCas9 to knock out CBX6 in the G401 cell line (Figure S63b) and determined that loss of CBX6 significantly reduces the growth rate (Figure S63c). To investigate whether CBX chromodomain inhibition can recapitulate the effects of CBX6 knockout in rhabdoid cells, we measured cell proliferation of G401 rhabdoid cells following treatment with 8, 11, 22 and 23. Wild-type G401 cells were treated with the inhibitors for 72 hours and cell counts were measured using CellTitre-Glo (Figure 7a). While all four inhibitors displayed significant toxicity at 100 μM, only compound 23 reduced growth to a similar degree as CBX6 knockout and was active at 50 μM. To define whether compound 23 is targeting CBX6 function, we treated wild-type and CBX6 KO G401 cells (Figure 7b) with 50 μM compound 23. Compound 23 decreased cell proliferation in wild-type but did not affect proliferation in KO cells. This provides indirect evidence that the compound is targeting CBX6. However, inhibition of other CBX chromodomains, particularly with 100 μM treatment, cannot be excluded. Further experiments are needed to determine whether CBX7 and CBX8 play a role in rhabdoid tumor viability, and whether these ligands might also be inhibiting them.

Figure 7.

Inhibitors are cell permeable and able to engage CBX proteins. A) G401 rhabdoid cells are affected by treatment of 8, 11, 22, 23 and show decreased cell proliferation. Cell counts were analyzed using CellTitre-Glo following 72-hour treatment. B) Treatment with compound 23 with wild type G401 decreased cell proliferation while no effect was observed in CBX6 KO. P value calculated using students t-test, n = 6. *** p<.001

Discussion

These studies show the critical role the (−2) position of CBX ligands for selectivity within the polycomb paralog family. The ability of ligands with larger groups in the (−2) position to bind the CBX proteins depends on the peptide scaffold and neighboring residues. Previous work demonstrated dye-labeled ligands with a valine residue in the (−2) position to be selective for CBX6. Removal of the dye from these scaffolds showed reduced selectivity for CBX6 over CBX8, however CBX6 was still better able to tolerate ligands containing bulkier groups within the (−2) pocket (compound 3, 6, Table 1). Larger groups in the (−2) position within the scaffolds reported demonstrate selectivity for CBX8. The aromatic cage, hydrophobic clasp and (−2) pocket are tightly connected and residues in the ligand to the right and left of the Ket2 can change the shape and size of the (−2) pocket. Despite the highly similar sequence homology of the residues in the (−2) pocket in CBX6/8, our results show a clear difference in the binding effects within this pocket depending on the chemical structure of the ligand and resulting protein-ligand interactions.

Aromatic and hydrophobic residues are well tolerated at the (−3) position of the ligand. The greatest selectivity for CBX8 within the (−3) series was achieved through the addition of a 4-methoxy phenyl moiety. Further exploration of substituted aromatic rings in the (−3) position of the ligand will aid in a better understanding of selectivity determinants.

The 5-phenylisoxazole N-cap moiety in compound 22 was the most potent dye-free CBX6/8 inhibitor in our series of compounds. Replacement of the 5-methyl to the 5-phenyl group on the terminal isoxazole ring improved binding to CBX6 and CBX8 by a factor of 60 and 33. These results clearly demonstrate the potential for potent inhibitors to be developed through extension of the ligand to engage the proteins’ peptide binding β-groove.

Developing cell-active chromodomain inhibitors is challenging and there are limited examples that have achieved this goal. In spite of the significant interest in CBX’s as drug targets, very few phenotypes for CBX inhibition have been reported. Of the 15 reports of chromodomain inhibitors,^{[6, 19–21, 23, 25]} there are 6 examples of cell-active CBX inhibitors.^{[6a, 6f, 6h, 20b, 21, 26]} The small molecule inhibitor MS37452 (K_d 30 μM) was shown to be active in cells but required compound treatment of >250 μM to observe decreased amounts of CBX7 on a known target locus.^[6a] The aforementioned UNC3866, with approximately 5% cell permeability, was the first example of a cell-active peptidic chromodomain inhibitor.^[6f] The use of DNA-encoded libraries to target the CBX proteins resulted in the discovery of a CBX8-selective and cell-permeable peptidic inhibitor.^[20b] Another example of DNA-encoded libraries targeting CBX proteins resulted in a selective and cell-active CBX2 inhibitor.^[26] In addition, the first positive allosteric modulator of CBX7 was reported and shown to regulate PRC1 activity in cellular assays.^[6h] A second allosteric modulator was recently posted for CBX8 that showed activity in two different cancer cell lines.^[21] The cell-based studies reported in this work demonstrate that our second generation inhibitors are able to enter cells and engage CBX proteins. Further studies will be required to determine the observed effects are due only to engagement of CBX6 or also through CBX7/CBX8, and to what extent they are inhibited in a cellular context.

Conclusion

CBX6 and CBX8 are emerging therapeutic targets for multiple aggressive malignancies. We have reported the first dual-selective inhibitors for CBX6 and CBX8, and demonstrated their entry into cells. These are among the first examples of inhibitors that bind specific CBX polycomb paralogs with no measurable activity against CBX7. The inhibitors and SAR described in this work will aid in paving the way for future efforts in chromodomain inhibition. Cell permeable chemical probes are needed to unravel the mysteries of CBX6 and to elucidate the paralog-specific roles of both CBX6 and CBX8.

Experimental Section

All experimental procedures along with compound characterizations are reported in the Supporting Information.

Abbreviations

CBX

chromobox

DEL

DNA-encoded library

DNA

deoxyribonucleic acid

ESC

embryonic stem cell

FITC

fluorescein isothiocyanate

FP

fluorescence polarization

Glu

glutamic acid

H3

histone 3

H3K27

lysine 27 on histone 3

H3K9

lysine 9 on histone 3

H3K9me3

trimethylated lysine 9 on histone 3

H3K27me3

trimethylated lysine 27 on histone 3

HCC

hepatocellular carcinoma

HP1

heterochromatin protein

IC₅₀

inhibitory concentration that reduces effect by 50 percent

Ket2

diethyllysine

Kme3

tri-methyllysine

KO

knockout

N.T.

not tested

PRC1

polycomb repressive complex 1

PRC2

polycomb repressive complex 2

PTM

post-translational modification

SAR

structure activity relationship