KDM3B inhibitors disrupt PAX3-FOXO1 oncogenic activity in fusion positive rhabdomyosarcoma.

doi:10.21203/rs.3.rs-2298627/v1

Download PDF

Article

KDM3B inhibitors disrupt PAX3-FOXO1 oncogenic activity in fusion positive rhabdomyosarcoma.

https://doi.org/10.21203/rs.3.rs-2298627/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 24 Feb, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Fusion-positive alveolar rhabdomyosarcoma (FP-RMS) is an aggressive pediatric sarcoma driven primarily by the PAX3-FOXO1 fusion oncogene, for which therapies targeting PAX3-FOXO1 are lacking. We screened 62,643 compounds using an engineered cell line that monitors PAX3-FOXO1 transcriptional activity identifying a hitherto uncharacterized compound, PFI-63. RNA-seq, ATAC-seq, and docking analyses implicated histone lysine demethylases (KDMs) as its targets. Enzymatic assays confirmed the inhibition of multiple KDMs with highest selectivity for KDM3B. Structural similarity search of PFI-63 identified PFI-90 with improved solubility and potency. Biophysical binding of PFI-90 to KDM3B was demonstrated using NMR and SPR. PFI-90 suppressed the growth of FP-RMS in vitro and in vivo through downregulating PAX3-FOXO1 activity, and combined knockdown of KDM3B and KDM1A phenocopied PFI-90 effects. Thus, we report novel KDM inhibitors with highest specificity for KDM3B. Its potent suppression of PAX3-FOXO1 activity can be exploited as a new therapeutic approach for FP-RMS and other transcriptionally driven cancers.

Biological sciences/Cancer/Cancer therapy/Drug development

Biological sciences/Cancer/Sarcoma

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma in children, accounting for 3% of all childhood tumors ¹. There are two major subtypes, embryonal RMS and alveolar RMS. Alveolar RMS is characterized by recurrent chromosomal translocations resulting in oncogenic PAX3-FOXO1 or PAX7-FOXO1 fusion transcription factors (fusion-positive RMS, FP-RMS) ². Clinically, FP-RMS is associated with metastatic disease and worse overall survival ³. Despite aggressive multimodal therapies including cytotoxic chemotherapeutic agents, the 5-year overall survival rate for patients with metastatic FP-RMS remains below 30% and is approximately 17% for those whose disease has relapsed ^{4, 5}. The poor outcomes highlight the need for new molecularly targeted therapies for this disease.

Expression of PAX3-FOXO1 is crucial for the survival and proliferation of FP-RMS ^{6, 7}. The fusion transcription factor establishes a core regulatory circuitry in FP-RMS that prevents the normal muscle differentiation program and locks the cell in a dedifferentiated proliferative state ⁸. To identify drugs that inhibit the transcriptional output of PAX3-FOXO1 and disrupt the core regulatory circuitry, a PAX3-FOXO1-selective cell-based functional reporter assay was utilized as previously described ^{7, 9}. ChIP-seq analysis showed that an intronic super enhancer within the ALK gene locus was highly bound by PAX3-FOXO1 and FP-RMS core regulatory transcription factors. This ALK super enhancer region was employed to express luciferase in a readout of PAX3-FOXO1-driven transcription. Using this cell-based assay, we screened a 62,643-member small molecule library and discovered a compound designated PFI-63 (PAX3-FOXO1 transcriptional signature Inhibitor 63) that blocks PAX3-FOXO1 downstream target gene expression. Surprisingly, the mechanism of action of PFI-63 was the inhibition of histone lysine demethylases (KDMs) with highest selectivity for KDM3B. Here, we present the structural, biochemical, and biological characterization of PFI-63 and a more potent analog, PFI-90. We show that PFI-90 exhibits potent in vitro cytotoxicity against FP-RMS with suppression of tumor growth in in vivo xenograft models.

Small Molecule Screen Identifies Inhibitor of PAX3-FOXO1 Activity

A small molecule library of 62,643 pure compounds (NCI Molecular Targets Program) was screened at a dose of 10 µM for 24-hours using a genetically engineered RH4 FP-RMS cell line expressing a PAX3-FOXO1-selective ALK super enhancer-driven luciferase ⁹. In parallel, we also tested the compounds on an RH4 cell line with a constitutively active CMV promoter-driven luciferase to monitor general transcriptional inhibition. We selected molecules with two characteristics: 1) Selective decrease in ALK super enhancer-driven luciferase relative to CMV promoter-driven luciferase; and 2) lack of non-specific toxicity at this early timepoint as measured by XTT cell viability assay (Fig. 1A). As described previously, we integrated the ALK luciferase, CMV luciferase, and XTT assays to generate a weighted average score that resulted in the identification of 573 hit molecules ⁷. Dose response data for these molecules was obtained in a follow-up screen consisting of five 10-fold dilutions ranging from 20 µM to 2 nM. This further reduced the candidate small molecules to 64 (Supplementary Table 1, Table 1), which were classified into 3 major groups ⁷. The first group showed decreased ALK super enhancer-driven luciferase concomitant with increased CMV-driven luciferase as observed with histone deacetylase inhibitors (HDACi) exemplified by N1302 (1-alaninechlamydocin) ⁷, which we designated HDACi-like compounds. A second group showed decreased ALK super-enhancer-driven luciferase without a significant change in CMV-driven luciferase, which we referred to as ALK-selective compounds. Lastly, compounds that strongly decreased both ALK super-enhancer and CMV-driven luciferase below 10% of control were denoted strong inhibitors (Fig. 1B). Besides N1302, there were 24 other compounds with known mechanisms of action including Midostaurin (PKC-412), a previously identified inhibitor of PAX3-FOXO1, which validated our screening methodology ¹⁰.

To investigate potential mechanisms of action of novel compounds, we manually selected few inhibitors from each group; HDAC-like, ALK-selective, and strong inhibitor for RNA-seq analysis following treatment of RH4 FP-RMS cells (Fig. 1B). Gene set enrichment analysis (GSEA) ^{11, 12} showed that PFI-63 (UPCMLD0ENAT5834780) treatment most closely phenocopied shRNA knockdown of PAX3-FOXO1 when looking at the sum of normalized enrichment scores of downregulated PAX3-FOXO1 gene sets. (Fig. 1C, Supplementary Fig. 1A and Table 2) ^{9, 13}. In addition, myogenesis and apoptosis gene sets were significantly enriched, suggesting that PFI-63 may induce FP-RMS differentiation and, ultimately, apoptosis. In line with this, western analysis demonstrated increased MYOG levels and PARP cleavage compared to DMSO control (Fig. 1D and Supplementary Fig. 1B). Direct enzymatic assay confirmed that caspase 3/7 activities were significantly increased (Supplementary Fig. 1C). Notably, PFI-63 action did not directly decrease the amount of PAX3-FOXO1 protein (Fig. 1D).

The half-maximal inhibitory concentrations (IC₅₀) of PFI-63 on FP-RMS cell lines RH4, RH30, and SCMC were in the single digit µM range (Fig. 1E). Next, we tested other pediatric solid cancer cell lines to determine broader applicability of PFI-63. IC50 values for PFI-63 on fusion negative RMS cell lines (RD: 7.16 µM and CTR: 2.92 µM), Ewing’s sarcoma (TC-32: 1.7 µM and A673: 3.7 µM) and osteosarcoma (OSA: 7.5 µM and HU09: 3.1 µM) were also in the single digit µM range (Supplementary Fig. 1D). We also tested PFI-63 on the NCI-60 cancer cell line panel ¹⁴, which showed that PFI-63 was mostly cytostatic in adult cancer cell lines (Supplementary Fig. 2). Given that PFI-63 downregulated PAX3-FOXO1 target gene expression in FP-RMS and had broad activity in other pediatric solid tumor cell lines, we next wanted to determine its molecular target(s).

PFI-63 Targets Histone Lysine Demethylases

First, we searched for other gene expression signatures that were significantly enriched following treatment with PFI-63 ^{11, 15}. Remarkably, among the most highly upregulated gene sets by PFI-63 were genes associated with GSK-J4 and JIB-04 treatment (Fig. 1F) ^{16, 17}. GSK-J4 and JIB-04 are inhibitors of the Jumonji domain (JmjC) family of histone lysine demethylases (KDMs) that are involved in epigenetic control of transcription. This suggested that PFI-63 may act by inhibiting KDMs. In support of this notion, submission of the PFI-63 chemical structure to the SEA (Similarity Ensemble Approach) Search Server (https://sea.bkslab.org/) to generate predictions for potential molecular targets revealed that PFI-63 could be a ligand for multiple JmjC KDMs (Supplementary Table S1, Table 2) ^{18, 19}. Moreover, ATAC-seq after treatment with PFI-63 to assess new open areas of chromatin showed enrichment of a gene set induced by treatment with the KDM1A/LSD1 inhibitor SP2509 (Supplementary Fig. 3A) ²⁰. Therefore, we hypothesized that PFI-63’s mechanism of action to be the inhibition of KDMs. Hence, we assessed the direct inhibition of KDMs by PFI-63 using cell free in vitro enzyme inhibition assays. Confirming our hypothesis, PFI-63 inhibited multiple KDMs including KDM3B, KDM4B, KDM5A, and KDM6B, with highest potency against KDM3B at an IC₅₀ of 7 µM (Fig. 1G). PyMOL alignment of PFI-63 to the enzyme pocket of KDM3B is shown in Fig. 1H. In contrast, PFI-63 had no significant inhibitory activity against HDAC1, HDAC2, HDAC3, or PRMT5 up to a concentration of 10 µM. Given these findings, we performed western analysis of PFI-63-treated FP-RMS cell lines for changes in methylation status of histone 3 lysines: H3K9 (demethylated by KDM3B, KDM4B, KDM1A), H3K4 (KDM1A, KDM5A) and H3K27 (KDM6B). After 24-hour treatment of RH4 and RH30 cells with PFI-63, we found significant increases in the methylation of all three lysine substrates compared to DMSO treatment, confirming that multiple histone demethylases are targeted by PFI-63 in FP-RMS cells (Fig. 1I and Supplementary Fig. 3B). Given the novel semi-selectivity of PFI-63 towards KDM3B, we next investigated KDM3B expression levels in RMS cell lines using the Broad Institute’s Dependency Map (DepMap) portal. We observed that RMS cell lines had the second highest average levels of expression of KDM3B compared to other cancer lines, indicating that KDM3B is significantly expressed in RMS and a potential target for therapy (Supplementary Fig. 3C).

Identification of Chemical Analog of PFI-63 with improved drug-like properties

PFI-63 had reduced solubility in water with a partition coefficient logP of 2.36 and thus has poor drug-like properties and makes preclinical in vivo testing a challenge. Hence, we conducted a molecular similarity search with PFI-63 in the Enamine catalog and identified 26 commercially available compounds of similar structure (Supplementary Table 1, Table 3). Performing the ALK super enhancer-driven luciferase screen with the 26 compounds identified one new compound designated as PFI-90 (Compound 26) (Fig. 2A), which showed significant inhibition of PAX3-FOXO1 driven luciferase (Supplementary Table 1, Table 3). PFI-90 had increased water solubility with a partition coefficient logP of 0.82 and showed low to sub-micromolar IC₅₀ activity in FP-RMS (Fig. 2B). PFI-90 was also tested on a large panel of adult cancer cell lines (NCI-60) which again showed mostly non-cytotoxic growth inhibitory effects except for renal cancer cell lines where cell killing was observed at the highest dose tested (Supplementary Fig. 4).

Similar to PFI-63, cell free in vitro enzyme inhibition assays confirmed that PFI-90 had highest inhibitory selectivity towards KDM3B (Fig. 2C). Western analysis showed an increase of H3K4me3 and H3K9me2 but only mild increase of H3K27me3 at doses up to 4µM (Fig. 2D and Supplementary Fig. 5A). This was consistent with PFI-90 having higher potency for KDM3B, KDM4B, KDM5A, and KDM1A which target H3K9 (KDM3B, KDM4B, KDM1A) and H3K4 (KDM1A, KDM5A) than KDM6B (H3K27). RNA-seq analysis at two time points, 6 hours and 24 hours after PFI-90 treatment showed significant enrichment of the myogenesis gene set at the 6-hour time point without enrichment of the apoptosis gene set (Fig. 2E and Supplementary Table 3, Table 1). However, at the 24-hour time point, myogenesis was no longer enriched and there was an enrichment of the apoptosis gene set along with downregulation of PAX3-FOXO1 targets (Supplementary Table 3, Table 2). Western analysis showed an increase of MYOG at 6 and 12 hours with a sharp decrease at 24 hours consistent with the GSEA analysis (Fig. 2F and Supplementary Fig. 5B). Additionally, there was initiation of PARP cleavage at 24 hours indicating apoptosis which was consistent with the GSEA result at the 24-hour time point. We also quantitated caspase 3/7 for apoptosis and found robust increase in caspase after PFI-90 treatment (Supplementary Fig. 5C). Lastly, we examined the transcript levels of the KDMs after treatment of PFI-63 and PFI-90 and found no significant change compared to DMSO treatment, indicating that the biological effect is due to enzymatic inhibition and not due to downregulation of KDM transcription (Supplementary Fig. 5D).

Biophysical Interaction Between PFI-90 and KDM3B

To establish that PFI-90 directly binds to KDM3B, we performed a series of bio-orthogonal ligand-binding detection experiments using nuclear magnetic resonance (NMR) techniques. For the NMR experiments, we employed the truncated KDM3B protein previously used to determine the crystal structure of the KDM3B enzymatic pocket ²¹. We initially obtained standard proton (¹H), carbon (¹³C), and HSQC NMR spectra of samples containing PFI-90 and 2-OG alone and together in buffer, as well as in the presence or absence of the KDM3B construct (Supplementary Fig. 6A). These results were used as reference for ligand-observed NMR experiments, allowing for unambiguous assignment of peaks to the compound structure. We then subjected these samples to Water-Ligand Observed via Gradient Spectroscopy (WaterLOGSY) ¹H NMR. We observed positive proton peaks for PFI-90 only in the presence of KDM3B indicating protein binding (Fig. 2G). When KDM3B was absent, PFI-90 had negative peaks indicating lack of binding as expected. Our non-binding negative control compound N-Methyl-L-Valine (NMV) showed negative peaks both with and without KDM3B indicating no binding as expected.

Next, we subjected the samples to Carr-Purcell-Meiboom-Gill Relaxation-Editing ¹H NMR (CPMG) to confirm binding using an orthogonal NMR method. CPMG analysis of PFI-90 showed binding of PFI-90 to KDM3B indicated by the attenuation in the PFI-90 peaks when compared to PFI-90 without KDM3B (Fig. 2H). Again, our non-binding negative control molecule NMV showed that there was no binding by CPMG as indicated by lack of attenuation of NMV peaks in the presence of KDM3B. Interestingly, nearly all ¹H signals of PFI-90 showed attenuation to ~ 75%, indicating that the vast majority of the PFI-90 structure binds to the enzymatic portion of the KDM3B protein in solution.

Finally, we also used Surface Plasmon Resonance (SPR) to determine binding of PFI-90 to synthesized full length KDM3B which showed robust enzymatic activity in demethylating H3K9me2 (Supplementary Fig. 6B). SPR demonstrated binding between PFI-90 to KDM3B with a K_d of 7.68 x10^− 6 M (Fig. 2I). Hence, we show, using 3 different methods that PFI-90 biophysically binds to KDM3B.

KDM3B is Essential for Survival in FP-RMS

Given that our novel compound targets KDM3B, we investigated whether KDM3B was essential for survival of FP-RMS. We developed a CRISPR library with sgRNA designed across all KDM3B exons. As controls, we used sgRNA for 8 essential genes including PAX3-FOXO1, MYOD, and 20 non-targeting sgRNA from the Brunello CRISPR library ²². We found that 7.1% of sgRNAs which target KDM3B were depleted by > 4-fold in the RH4 FP-RMS cell line with hot spots in the enzymatic domain of KDM3B, indicating that it is in fact essential for survival of FP-RMS (Fig. 3A). It is noteworthy that KDM3B has not previously been recognized as a dependent gene in FP-RMS (e.g., in DepMap), most likely due to a limited selection of sgRNAs against this gene.

KDM Knockdown Recapitulates the PFI-90-induced Transcriptome

To demonstrate that PFI-90 downregulation of PAX3-FOXO1 target genes is through the inhibition of KDMs especially KDM3B, we knocked down the top 4 KDMs inhibited by PFI-90, namely KDM3B, KDM1A, KDM4B, and KDM5A using CRISPRi. We achieved effective knockdown of KDMs measured at the transcript level and at the protein level (Fig. 3B-C and Supplementary Fig. 7A-C). RNA-seq analysis followed by GSEA enrichment showed that KDM3B knockdown significantly downregulated all 5 PAX3-FOXO1-related gene sets that were downregulated by PFI-90 (Fig. 3D and Supplementary Table 3, Table 3–10). Comparison of KDM knockdown to PAX3-FOXO1 knockdown using shRNA showed that KDM3B knockdown decreased PAX3-FOXO1 gene sets most robustly (Supplementary Fig. 7D). However, we did not see enrichment of the upregulated gene sets observed following PFI-90 treatment, namely, those involved in myogenesis or apoptosis. Given that KDM3B removes transcriptionally repressive methyl groups from H3K9, enrichment of the same downregulated gene sets was in line with inhibition of KDM3B function. We reasoned that since H3K4 methylation was also increased and since methylated H3K4 is a marker of transcriptional activation, we investigated upregulated GSEA gene sets of KDM1A and KDM5A knockdowns, both of which target H3K4. Interestingly, we found enrichment for myogenesis and apoptosis following KDM1A knockdown but not in KDM5A knockdown cells. KDM1A knockdown also downregulated PAX3-FOXO1 gene sets although not as robustly as KDM3B knockdown. The combined knockdown of KDM1A and KDM3B showed enrichment for myogenic differentiation and apoptosis gene sets together with downregulation of PAX3-FOXO1 targets, most closely recapitulating the PFI-90-induced transcriptome in RH4 cells (Fig. 3D).

scRNA-seq Shows PFI-90-Induced Decrease PAX3-FOXO1 Signature, Myoblast Signature, and G2/M Population

In order to study the effects of PFI-90 at the single-cell level, we performed scRNA-seq analysis after treating cells with PFI-90 for 16 hours, a time point before apoptosis. Integrating the PFI-90 and DMSO samples together, Cluster 2 showed the largest fold difference in cell number between the PFI-90 and DMSO groups with 82.0% of the cluster comprised of DMSO treated cells (Fig. 4A-B). Cluster 2 was also the largest cluster of DMSO treated cells overall and hence represented a major population of RH4 cells during exponential growth. Clusters 3 and 8 showed the greatest fold change following PFI-90 treatment compared to DMSO. To characterize the clusters, we examined myogenic differentiation using the top 50 differentially expressed genes from a previously published scRNA-seq profile (Supplementary Table 4, Table 1–2) ²³. RMS cells are actively dividing in culture, resembling highly proliferative myoblast cells. Interestingly, myoblast signature for Cluster 2 was significantly higher than Clusters 1, 3, 6, and 8 (Fig. 4C and Supplementary Table 4, Table 3). For the less proliferative myocyte signature, Cluster 6 had significantly higher myocytes signature compared to the rest of the clusters. Differential gene expression analysis showed Cluster 6 with high expression of skeletal muscle genes including multiple myosin light chain and heavy chain genes (Supplementary Table 4, Table 4–11). Clusters 3 and 8 which increased the most after PFI-90 treatment had reduced expression of myoblast genes indicating that PFI-90 treatment led to an increase in the number of cells with the lowest myoblast signature. When we examined the PAX3-FOXO1 signature ⁸, we found that Clusters 3 and 8 also had lower expression of PAX3-FOXO1 downstream targets. Heatmap hierarchical clustering using leading edge genes from myoblast signature and PAX3-FOXO1 signature both showed that Clusters 3 and 8 separated from the rest of the Clusters (Supplementary Fig. 8A). These findings demonstrated a strong correlation between PAX3-FOXO1 expression and maintenance of the myoblast signature. We next performed cell cycle analysis of the scRNA-seq clusters (Fig. 4D-E). Overall, there was an increase of cells in G1 and S while G2M was decreased after PFI-90 treatment indicating potential G2M arrest. When examining which clusters were associated with the decrease in G2M, we saw significant decrease in Cluster 2 after PFI-90 treatment (Fig. 4E). To better understand the cell cycle changes, we performed flow cytometry analysis at 24 hours after treatment with PFI-90 vs DMSO in SCMC cell line. We found profound accumulation of S phase cells while G2 and M was significantly decreased indicating a S phase blockade (Fig. 4F). These findings from scRNA-seq indicated that PFI-90 treatment resulted in decreased expression of PAX3-FOXO1 downstream targets concomitant with decreased myoblast signature and cell division in individual cells.

PFI-90 Increases H3K4 and H3K9 Methylation

To determine how inhibition of KDM3B results in abrogation of PAX3-FOXO1 action, we performed ChIP-seq analysis using antibodies against PAX3-FOXO1, H3K4me3, H3K9me2, H3K27me3, and H3K27ac. First, we performed ChromHMM analysis using our previously published ChIP-seq data to assign the chromatin into 16 states ^{8, 9}. We observed that H3K9me2 was increased throughout the entire genome after treatment with PFI-90 compared to DMSO, with the most pronounced increase at enhancer regions (Fig. 5A). ChromHMM also showed that H3K4me3 was increased mostly in the promoter region which is consistent with its function as a transcription initiation mark. Next, we looked at the transcription start sites (TSS) and PAX3-FOXO1 binding sites. We found that there was an increase in histone methylation at H3K9me2 both at the TSS and PAX3-FOXO1 binding sites in PFI-90 treated cells compared to DMSO (Fig. 5B). This was consistent with PFI-90 having highest inhibition of KDM3B which demethylates histones at H3K9. There was also an increase in H3K4me3 around TSS following treatment with PFI-90, consistent with PFI-90 inhibiting KDM1A and KDM5A which demethylate H3K4. We saw no changes in the methylation status of H3K27. These changes corresponded with PFI-90 selectivity for each of the respective histone marks.

Since PAX3-FOXO1 drives core regulatory transcription factors at super enhancer sites marked with H3K27ac, we looked to see if there were any changes to enhancer-related histone marks around the PAX3-FOXO1 binding sites. There was no change in H3K27ac at PAX3-FOXO1 sites with treatment of PFI-90 compared to DMSO (Fig. 5B). Super enhancer ROSE analysis also showed no changes in super enhancers when RH4 cells were treated with PFI-90 compared to DMSO (Supplementary Fig. 9A). PFI-90 also did not change the binding of PAX3-FOXO1 to target chromatin regions as assessed by ChromHMM despite increases in H3K9me2, consistent with a previous report of PAX3-FOXO1 having the ability to bind heterochromatin as a pioneering factor ²⁴. We also examined the protein level of PAX3-FOXO1 by Western since PAX3-FOXO1 binds super enhancers and found that PFI-90 did not directly decrease PAX3-FOXO1 protein levels (Supplementary Fig. 9B). These results are similar to PFI-63 and consistent with the ChIP-seq data (Supplementary Fig. 9C).

Given that H3K9me2 was increased most significantly at enhancer regions, we wanted to know the relative distribution of H3K9me2 at PAX3-FOXO1 binding sites compared to other transcription factors. Focusing on enhancer regions associated with genes based on RNA-seq transcript level using EDEN, we selected the top 10% with increased H3K9me2 after PFI-90 treatment and performed ChIP Enrichment Analysis (ChEA) via Enricher which compares gene lists with previously published ChIP-seq of transcription factors (Fig. 5C and Supplementary Table 5, Table 1)²⁵. PAX3-FOXO1 was the top significant transcription factor overlapping with H3K9me2 followed by ZNF217 which is one of the PAX3-FOXO1 core regulatory transcription factors ⁸. Next, we looked to see if the increase in H3K4me3 methylation at promoter regions was associated with enrichment of biological processes predicted by RNA-seq GSEA, namely myogenesis and apoptosis. We detected new H3K4me3 peaks after treatment with PFI-90 and performed gene ontology analysis which showed strong (8.9-fold) enrichment for muscle differentiation as well as 3.4 fold enrichment for apoptosis (Fig. 5D and Supplementary Table 5, Table 2) ²⁶. Therefore, these ChIP analysis results validated the bulk RNA-seq data; i.e., PFI-90 suppresses PAX3-FOXO1 function and upregulates myogenesis as well as apoptosis programs.

PFI-90 Decreases Chromatin Loop Strength and Decreases Topologically Associating Domains (TADs) at PAX3-FOXO1 Sites

Given that there were no changes to PAX3-FOXO1, H3K27me3, and H3K27ac binding and respective super enhancer calls following treatment with PFI-90, we asked if the increase in H3K9me2 and H3K4me3 is sufficient to change the chromatin structure in FP-RMS cells. We performed Hi-C and assessed if PFI-90 treatment was associated with decreased number of loops given that H3K9me2 is a repressive heterochromatin mark. We found that there were minimal changes in the number of loops (675 loss and 582 gain). Genes associated with these loops did not show any statistically significant association with gene sets. However, we found that although there was no difference in loop number, there was decrease in loop strength (Supplementary Fig. 10A-B). Next, we looked to see if TADs were changed by PFI-90 given the decrease in loop strength. Interestingly, there was significant decrease in TAD score in PFI-90 treated cells compared to DMSO (Supplementary Fig. 10C). Genes associated with the decreased TADs and loop strength were enriched for PAX3-FOXO1 target genes by GSEA gene sets (Supplementary Table 5, Table 3). Decrease in TADs and loop strength at PAX3-FOXO1 sites suggests that decrease in PAX3-FOXO1 target transcripts, observed by RNA-seq, may be related to the increase in H3K9me2 at PAX3-FOXO1 sites.

PFI-90 Decreased RNA Pol2 Binding and Phosphorylation

RNA Pol2 binding and phosphorylation of Ser5 results in transcription initiation and phosphorylation of Ser2 leads to elongation. We wanted to determine if PFI-90 treatment and the resulting increase in H3K9me2 was associated with disruption of RNA Pol2. We performed ChIP-seq analysis and found that PFI-90 treatment resulted in remarkable decrease in RNA Pol2 binding at TSS and phosphorylation of RNA Pol2 at both Ser5 and Ser2 (Fig. 6A). Furthermore, we saw decrease in enhancer bound RNA Pol2 and phosphorylation (Fig. 6B). Looking at MYOD1, there was significant decrease in RNA Pol2 binding and phosphorylation both at gene body and enhancers which was associated with increased H3K9me2 (Fig. 6C). Next, we determined the differential RNA Pol2 peaks between PFI-90 and DMSO and ranked the associated genes. GSEA analysis showed that PAX3-FOXO1 gene sets were enriched in downregulated peaks for Total Pol2 and Pol2 Ser5 consistent with RNA-seq findings of downregulated PAX3-FOXO1 target transcripts (Fig. 6D and Supplementary Table 5, Table 4–5).

PFI-90 Inhibits FP-RMS Tumor Growth In Vivo

Finally, we tested PFI-90 in two xenograft models of FP-RMS using the RH4 cell line. In a metastatic intravenous mouse model, PFI-90 was able to significantly delay tumor progression measured by luciferase signal (Fig. 7A, p = 0.0016). We also tested PFI-90 in an orthotopic model of FP-RMS injecting tumor cells intramuscularly into the gastrocnemius muscle. We again saw a significant delay in tumor progression measured by tumor volume (Fig. 7B, p = 0.0046). There was no significant change in weight during PFI-90 treatment (Supplementary Fig. 11). We also performed an additional in vivo study using a subcutaneous tumor model. After 5 days of daily treatment with PFI-90, tumors were collected for RNA-seq analysis. GSEA showed downregulation of PAX3-FOXO1 nuclear targets, and enrichment for myogenesis and apoptosis gene sets after PFI-90 treatment versus DMSO, consistent with our in vitro findings (Fig. 7C and Supplementary Table 6).

We show in our study that the novel compounds, PFI-63 and PFI-90, inhibit multiple KDMs with highest selectivity for KDM3B. To our knowledge, this is also the first report that FP-RMS are sensitive to KDM3B inhibition. PFI-63 and PFI-90 downregulated PAX3-FOXO1 target transcripts without changing PAX3-FOXO1 protein or PAX3-FOXO1 binding levels at enhancers. Interestingly, despite H3K9me2 being a repressive mark and significantly increased after PFI-90 treatment, deposition of the activating H3K27ac mark was not affected indicating that enhancer/super-enhancer designations were unchanged. Moreover, increase in H3K9me2 from PFI-90 treatment did not result in changes to number of loops as would be expected given that H3K9me2 is a marker of heterochromatin. It is possible that the 24-hour time point was too early to detect changes or that the increase in H3K4me3 may be contributing to the stability of chromatin structure limiting compartment changes. However, we did see a decrease in loop strength and TADs at PAX3-FOXO1 sites with PFI-90 treatment. These findings may propose a remarkable biological property of KDM3B; that inhibition of KDM3B and the resulting increase in H3K9me2 can dial down transcription despite maintenance of super enhancer loops and binding of oncogenic transcription factors. In line with this hypothesis, PFI-90 treatment resulted in significant decrease in RNA Pol2 binding and phosphorylation indicating that H3K9me2 increase may be sufficient in dialing down of transcription.

Discovery of KDM3B as a potential target for therapy in FR-RMS extends the findings with other KDM inhibitors reported to be effective in RMS, including the KDM1A inhibitor GSK690 and QC6352, and the pan-KDM inhibitor JIB-04 ^{27, 28, 29}. To date, only one other group has reported the identification of potential KDM3B inhibitors (JDI-4 and JDI-12) ³⁰. However, the ability of JDI-4 and JDI-12 to inhibit other KDMs has not been tested; considering the high homology between KDM family members, it is difficult to conclude KDM3 specificity without direct comparison. For instance, another well-known KDM inhibitor, GSK-J4, which was initially reported to be selective for KDM6A and KDM6B, was found to similarly inhibit KDM4C and KDM5B, further stressing the need for testing of multiple KDM family members when characterizing KDM inhibitors ³¹.

We also observed an increase in H3K4me3 due to PFI-90 inhibition of KDM1A and KDM5A. Analysis of H3K4me3 ChIP-seq showed that peaks induced by PFI-90 were at genes involved in myogenesis and apoptosis. Interestingly, although both KDM1A and KDM5A target H3K4 methylation, RNA-seq transcript profile for CRISPRi knockdown of KDM1A showed enrichment of myogenic differentiation and apoptosis gene sets while knockdown of KDM5A did not. Negative enrichment of some of the PAX3-FOXO1 gene sets was also observed when KDM1A was knocked down (Fig. 3D), consistent with the fact that KDM1A can remove methyl groups from H3K9 in certain contexts ³². These results demonstrated that KDM1A is a relevant target of PFI-90’s biological action.

We propose a model of PFI-90 action that involves the downregulation of PAX3-FOXO1 target gene expression by increasing H3K9me2 mainly through the inhibition of KDM3B with potential contribution from KDM1A inhibition (Fig. 7D). Given that FP-RMS is “addicted” to the PAX3-FOXO1-driven super enhancer circuitry, it seems likely that this core regulatory circuitry is disproportionately sensitive to the increase in the repressive H3K9me2 mark leading to decrease in RNA Pol2 at PAX3-FOXO1 sites. PFI-90 activation of the myogenic differentiation program and apoptosis is likely via increases in H3K4 methylation predominantly by inhibition of KDM1A (Fig. 6D). Overall, H3K9me2 and H3K4me3 increase leads to transcriptional rewiring resulting in sensitivity of FP-RMS and other transcriptionally driven cancers to PFI-90. In regards to multi-KDM inhibition, due to the functional redundancy of KDM family members for H3K substrates, simultaneously targeting numerous KDMs is expected to show increased efficacy than selectively targeting an individual KDM ²⁸. As KDM3B has been implicated in taxane-platin-resistant small cell lung cancer, acute lymphoblastic leukemia, and hepatocellular carcinoma, PFI-90 emerges as a lead compound for therapeutic development against multiple cancer types ^{33, 34, 35}. Also, because PFI-90 exerted potent growth inhibitory activity on Ewing’s Sarcoma and osteosarcoma cell lines (Supplementary Fig. 1D), KDM3B inhibition could be effective in other pediatric cancers especially transcription driven cancers. In conclusion, PFI-90 represents a novel compound that inhibits multiple histone demethylases with highest selectivity to KDM3B that can be further developed for the therapy of highly malignant FP-RMS and other transcription driven cancers.

Cell Lines

Cell lines were tested for mycoplasma and identities were ensured by RNA-seq and genotyping. RMS cell lines RH4 and RH30, were kind gifts from Dr. Peter Houghton, SCMC from Dr. Janet Shipley, RD and CTR from Dr. Lee Helman. Osteosarcoma cell lines OSA and HU09 and Ewing’s Sarcoma cell lines TC-32 and A673 were from Dr. Paul Meltzer.

Small Molecules

High throughput screening of the 62,643 compounds from the NCI Molecular Targets Program was carried out in 384-well plates as previously described ⁷. PFI-63 (PubChem CID 912764) and PFI-90 (PubChem CID 12499590) were purchased from Enamine (EN300-6731773, EN300-262365 respectively).

In Vitro Dose Response

Dose responses were performed by quantifying percent cell confluence from phase contrast images using Incucyte ZOOM in 384-well plate format. Dose response was achieved using a range between 30 µM to 0.17 nM in triplicates. IC50 values were calculated for each time point using GraphPad 8.

In Vitro Enzyme Inhibition Assay

Enzyme inhibition assay for KDM2B, KDM3A, KDM3B, KDM4B, KDM5A, KDM5B, KDM6A, KDM6B, HDAC1, HDAC2, HDAC3, PRMT5 were performed by Eurofins (St Charles, MO). KDM1 inhibition was performed using the LSD1 Activity Quantification Assay Kit (Abcam, ab113460).

RNA-seq

RNA-seq was performed as previously described ³⁶. Briefly, total RNA collected using RNeasy kit (Qiagen, 74034). Poly-A selected RNA libraries were prepared and sequenced on an Illumina HiSeq2000. Reads were aligned to hg19 using STAR version 2.5.3a, and gene expression was calculated as Transcripts per Million mapped reads (TPM) using RSEM version 1.3.1 and the UCSC reference. For variance normalization, DESeq2 package was utilized. GSEA analysis was performed on log2-fold change ranked list calculated by comparison with control using published gene sets.

scRNA-seq

RH4 cells were treated with PFI-90 at 1 µM for 16 hours and collected. 10X Genomics platform was used for scRNA-seq. Library was sequenced on Illumina NextSeq2000. Fastq files were aligned using cellragr V6 using hg19. Barcodes filtration was performed based off the distribution of data (median +/- 3 deviations) for high percentage of mitochondria, low number of genes and low number of cells. Individual sample normalization was done using SCTransform through Seurat V3 setting the variable.features.rv.th to 1.3 ³⁷. Doublets were detected using DoubletFinder V2 with a doublet estimate of 3%. Cell cycle annotations was determined using the CellCycleScoring method in Seurat. Sample integration, clustering, dimension reduction and differential expression was performed using the Seurat V3 Package. Data integration was performed using the default parameters followed by clustering using the Leiden algorithm ³⁸ and a resolution of 0.2 and projected using Uniform Manifold Approximation and Projection (UMAP) ³⁹. Differential expression was performed using the FindMarkers function in Seurat with the option of using MAST ⁴⁰ for method which uses a hurdle model for handling the zero inflated single cell data. GSEA analysis was performed by using the fold change from the differential expression analysis and fed into fGSEA ⁴¹.

ChIP-seq

ChIP-seq was performed as previously described ⁴². Briefly, cells were fixed using formaldehyde (1% final concentration) for 14 min and sheared for 27 cycles 30s on/30s off (Active Motif, 53052). Samples were immunoprecipitated overnight at 4°C with antibodies and purified by agarose beads (Active Motif, 53040). Refer to antibody table for details including the in-house anti-PAX3-FOXO1 antibody (Supplementary Table 7) ⁴³. ChIP-seq reads were aligned to hg19 using BWA version 0.7.17. Samples were normalized using Drosophila spike-in to reads per million mapped dm3. Peaks were called using MACS2 (version 2.1.1.20160309, https://github.com/taoliu/MACS) using “narrow” mode for all targets except for H3K27me3 and H3K9me2 for which “broad” was used. Mapping artifacts were removed (locations black-listed by ENCODE, https://sites.google.com/site/anshulkundaje/projects/blacklists). The spike-in normalized read count for merged peaksets was calculated using deepTools’ “multiBigwigSummary”. The nearest genes were selected with bedtools’ “closest”. The pre-ranked log2 fold changes were calculated and GSEA performed by “gsea-cli.sh GSEAPreranked -scoring_scheme weighted” command with FDR cutoff 0.25.

ChromHMM

The chromatin states were characterized using chromHMM ⁴⁴ which can learn and infer the combinatorial patterns of histone marks using Hidden Markov Model algorithm. The read count for each state was calculate with deepTools’ “multiBamSummary” command ⁴⁵. The percentage of genome coverage for each state was calculated with in-house bash script.

ATAC-seq

ATAC-seq was performed as previously described ⁸. Briefly, cells were treated with DMSO or PFI-63 and collected. 50,000 cells were counted and nuclei isolated. Tagmentation was performed (Illumina, 20034198) and purified using MinElute kit (Qiagen, 28206). Library was amplified (NEB, M0541L) for 16 cycles. Library was purified using AMPure XP beads (Beckman, A63880) and sequenced using NextSeq500. The spike-in normalized read count for promoter regions of each gene was calculated using deepTools’ “multiBigwigSummary”. Then we used GSEA’s ¹¹ “gsea-cli.sh GSEAPreranked -scoring_scheme weighted” command on ranked gene list to generate gene set enrichment list with FDR cutoff 0.25.

Hi-C

RH4 cells were treated with PFI-90 at 1 µM for 24 hours and collected. The Dovetail Micro-C kit (Dovetail, 21006) was used to generate libraries. In short, MNase is used to digest the chromatin after fixation. Proximity ligation is performed and after reverse crosslinking, library is synthesized and purified using streptavidin-based bead purification. 3 separate libraries of technical replicates were sequenced to the total depth of 800 million reads. The raw reads were mapped to hg19 genome using a bash script omni-c_qc.sh from Dovetail Github (https://github.com/dovetail-genomics). The replicates were then merged with Pairtools’ “merge”. The active and inactive compartments were assigned by FAN-C ⁴⁶ with “fanc compartments” at 500kb resolution. Eigenvectors was corrected by Spearman correlation with H3K4me1. Then we looked for the flipped compartments with a cutoff of loading difference 0.02. The flipped compartments were intersected with known coding genes ≥ 1 TPM in either DMSO or PFI90 RNAseq. Enrichment analysis was performed by overlapping the flipped gene list with NCI v33 gene sets using R package “GeneOverlap” ⁴⁷. For loop calling, we used the FAN-C command “fanc loops” with 10kb resolution. Loops were further filtered with “rh-filter” and “remove-singlets” options. The lost and gained loops were determined using bedtools ⁴⁸ command “pairToPair”. To identify the loop associated genes, we used “pairToBed” command in bedtools. For TAD analysis, we first used HiCExplorer’s ⁴⁹ hicFindTADs command with FDR cutoff 0.01 at 5kb resolution. Then we compared DMSO vs PFI-90 by using HiCExplorer’s “hicDifferentialTAD” command. We further filtered the rejected TAD regions by adjusting intra-TAD p-values with cutoff 0.05. Then called loops in differential TADs were intersected with transcription start site regions of the coding genes. GSEA performed by overlapping with NCI v33 gene sets using R package “GeneOverlap”. We performed Aggregate Peak Analysis on loops of interests using Juicer’s “apa” command. The contact matrices were normalized using the interactions identified in spike-in mouse reads.

Tiled KDM3B CRISPR Knockout Screen

sgRNA guides were designed throughout all KDM3B exons resulting in 479 sgRNA. Lentivirus library was produced by transfecting HEK293 cells (Thermo, L3000015) and virus collected at 24, 48, and 72 hours and pooled. PEG-it (SBI, LV810A-1) was used to concentrate the virus. RH4 cells were transduced at MOI of 0.2 and Puromycin (Thermo, A1113803) selected for 7 days. Cells were collected every 5 days for sequencing. DNA was isolated using DNA isolation kit (Qiagen, 80204). gRNA inserts were amplified using primers harboring Illumina TruSeq adapters with i5 and i7 barcodes, and the resulting libraries were sequenced on NextSeq500 ⁵⁰. MAGeCK was used for pairwise comparison to the plasmid pool. Log fold depletion of guides were calculated and walking average of 4 guides were plotted for the entire KDM3B gene.

Western Blot Analysis

Whole protein cell lysates were obtained by using RIPA buffer while total histone cell extracts were obtained by using EpiQuick Total Histone Extraction Kit (Epigentek, OP-0006). Protein concentrations were estimated by Pierce BCA Protein Assay Kit (Thermo, 23225). Refer to antibody table for details (Supplementary Table 7). Membranes were developed using ECL reagent (GE Healthcare, RPN2106).

Apoptosis Quantification

RH4 and RH30 cells were treated with PFI-63 and PFI-90 for 24 hours. Caspase-Glo 3/7 kit (Promega, G8091) was used, and luminescence measured for quantitation of apoptosis.

Flow Cytometry

SCMC cells were treated with PFI-90 at 1µM for 24 hours vs DMSO. Cells were fixed and permeabilized before stained using antibody against Ki67 and pH3 followed by secondary Alexa488 antibody and Hoechst (Supplementary Table 7). Cells were analyzed on BD FACSymphony A5.

Protein Purification

Enzymatic domain of KDM3B was synthesized using plasmid expression vector (Addgene, 53627). Protein was expressed in E. Coli and purified using His-trap column followed by sephacryl S200 SEC column. Full length KDM3B plasmid was synthesized by GeneUniversal (Newark, DE). Trichoplusia ni (TniFNL) cells were used to produce the KDM3B protein and was isolated using HisTrap FF.

KDM3B Enzyme Activity Assay

Full length KDM3B was tested for enzymatic function using histone tail with K9 dimethylation (EpigenTek, R-1026-200). Enzyme activity was performed in 50 mM HEPES pH 7.5, 1 mM TCEP, 100 µM 2-OG, 100 µM Ascorbate, and 50 µM Iron Sulfate for 3 hours at room temp. End product was analyzed on X500B Q-TOF (SCIEX) mass spectrometer coupled to an Exion UHPLC.

NMR

NMR spectra were recorded on Bruker AVANCE III 500 MHz spectrometer equipped with a Bruker TCI cryogenically cooled probe. Samples were prepared in 90% aqueous buffer containing 10% D₂O for sample locking. The buffer used was per Leung et al ⁵¹, and composed of 50 mM Tris-d11, pH 7.5, and 80 uM zinc sulfate monohydrate. PFI-90 was at 300 µM and KDM3B at 15 µM. Suppression of the water peak was through excitation sculpting for the WaterLOGSY experiments using the pulse program ephogsyno.2 or through presaturation for the CPMG experiments (pulse program cpmgpr1d). A reference ¹H spectrum was collected for PFI-90 peaks and annotated based on supporting ¹³C and HSQC spectra. For WaterLOGSY (WL) experiments, spectra were acquired with 256–512 scans ⁵², while for CPMG experiments, spectra were acquired with 64 scans.

SPR

Full length KDM3B and PFI-90 were assessed using Biacore T200 system (GE Healthcare). Neutravidin was covalently immobilized on a CM5 chip. Avi-tagged KDM3B protein was added to FC2 to a final response of ~ 800 RU. PFI-90 was prepared in degassed, filtered HBS-P+ (GE Healthcare) buffer with 3% DMSO. Single cycle kinetic experiments were performed using five injections (30 µL min^− 1) of increasing concentrations (20–2000 nM) passed over the sensor chip for 150-s association, followed by a 420-s dissociation.

CRISPRi Knockdown of KDMs

Plasmid construct for lentivirus containing dCas9-KRAB with target sgRNA was synthesized against KDM1A, KDM3B, KDM4B, and KDM5A. sgRNA sequences are as follows. KDM1A: caccgccacggagcgacagagcgag. KDM3B: caccgcagcggcgagcggagatccg KDM4B: caccgcggtcgcca gcaaccgagcg. KDM5A: caccgcgtctttgagccgagttggg. Lentivirus was produced by transfecting HEK293 cells and pooling virus collection at 24, 48, and 72 hours. CRISPRi knockdown cells were selected using Puromycin.

In Vivo Xenograft Model of FR-RMS

All animal experiments were approved by Institutional Animal Care and Use Committee. In the orthotopic IM model, 7-week-old NSG mice were injected with 5 million RH4 cells with Matrigel into the gastrocnemius muscle. Treatment with PFI-90 started 5 days later at 20 mg/kg body weight dosed daily i.p. for 5 days with 2-day rest. Dose was well tolerated so PFI-90 was increased to 25 mg/kg on day 28 of treatment until the end of study. In the IV metastatic model, 12-week-old NSG mice were injected with 1 million RH4 cells expressing luciferase. Treatment with PFI-90 was started 3 days later at 25 mg/kg body weight given daily i.p. for 5 days with 2 days rest. For tumor analysis, 5 million RH4 cells were injected subcutaneously. PFI-90 treatment started after 12 days at 25 mg/kg dose i.p. daily for 5 days at which time tumors were collected and snap frozen. RNA was processed with RNA-seq as above

Statistical Analysis

Statistical analysis performed for bulk RNA-seq, scRNA-seq, ChIP-seq, and HiC are described in their respective sections above. GraphPad 8.4.3 was used for IC50 calculations (four parameter logistics regression), caspase quantitation and flow cytometry (t-test), and in vivo mouse studies, (analysis of variance).

Data and Materials Availability

All sequencing files are available on GEO XXX.

All analysis R Codes are available on Github XXX.

ACKNOWLEDGEMENTS:

This project was supported by the Children’s Cancer Foundation and in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the US Government.

AUTHOR CONTRIBUTIONS:

Conceptualization: YK, BEG, JFS, RGH

Methodology: YK, RS, MP, AA, GMW, BZS, JJB, JS, XW, HC, VG, RGH

Investigation: YK, BEG, RS, JFS, AA, SP, GMW, BZS, JJB, JTK, YKS, CW, MO, LMJ, RGH

Formal analysis: YK, BEG, RS, AA, SP, GMW, DM, JJB, SRS, JRE, HC, VG, MO, LMJ, JSW, BRO, RGH

Visualization: YK, RS, AA, HC

Funding acquisition: YK, JK

Resources: AA, GMW, JS, RC, SRS, DE, JJ, LP, MO, JBM, BRO

Project administration: JK

Supervision: RGH, JK

Writing – original draft: YK, RS, DM, JSW, RGH, JK

Writing – review & editing: YK, BEG, RGH, JK

COMPETING INTERESTS:

Patents: Compounds PFI-63 and PFI-90 under US provisional Patent number 62/936,722. Patent title of “Small Molecule Inhibitors of Histone Demethylases for Treating Rhabdomyosarcoma (RMS) and Other Cancers”.

Authors declare that they have no competing interests.

Shern JF, Yohe ME, Khan J. Pediatric Rhabdomyosarcoma. Crit Rev Oncog 20, 227–243 (2015).
Parham DM, et al. Correlation between histology and PAX/FKHR fusion status in alveolar rhabdomyosarcoma: a report from the Children's Oncology Group. Am J Surg Pathol 31, 895–901 (2007).
Sorensen PHB, et al. PAX3-FKHR and PAX7-FKHR gene fusions are prognostic indicators in alveolar rhabdomyosarcoma: A report from the children's oncology group. J Clin Oncol 20, 2672–2679 (2002).
Rudzinski ER, et al. Histology, fusion status, and outcome in metastatic rhabdomyosarcoma: A report from the Children's Oncology Group. Pediatr Blood Cancer 64, (2017).
Pappo AS, et al. Survival after relapse in children and adolescents with rhabdomyosarcoma: A report from the intergroup rhabdomyosarcoma study group. J Clin Oncol 17, 3487–3493 (1999).
Grobner SN, et al. The landscape of genomic alterations across childhood cancers. Nature 555, 321–327 (2018).
Gryder BE, et al. Chemical genomics reveals histone deacetylases are required for core regulatory transcription. Nat Commun 10, 3004 (2019).
Gryder BE, et al. Histone hyperacetylation disrupts core gene regulatory architecture in rhabdomyosarcoma. Nat Genet 51, 1714–1722 (2019).
Gryder BE, et al. PAX3-FOXO1 Establishes Myogenic Super Enhancers and Confers BET Bromodomain Vulnerability. Cancer Discov 7, 884–899 (2017).
Amstutz R, et al. Phosphorylation regulates transcriptional activity of PAX3/FKHR and reveals novel therapeutic possibilities. Cancer Res 68, 3767–3776 (2008).
Subramanian A, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences 102, 15545–15550 (2005).
Mootha VK, et al. PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34, 267–273 (2003).
Ebauer M, Wachtel M, Niggli FK, Schaefer BW. Comparative expression profiling identifies an in vivo target gene signature with TFAP2B as a mediator of the survival function of PAX3/FKHR. Oncogene 26, 7267–7281 (2007).
Alley MC, et al. Feasibility of drug screening with panels of human tumor cell lines using a microculture tetrazolium assay. Cancer Res 48, 589–601 (1988).
Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 1, 417–425 (2015).
Hookway E. The role of the lysine demethylases KDM5 and KDM6 in bone malignancies.). University of Oxford (2017).
Wang L, et al. A small molecule modulates Jumonji histone demethylase activity and selectively inhibits cancer growth. Nat Commun 4, 2035 (2013).
Keiser MJ, Roth BL, Armbruster BN, Ernsberger P, Irwin JJ, Shoichet BK. Relating protein pharmacology by ligand chemistry. Nat Biotechnol 25, 197–206 (2007).
Hoffmann I, et al. The role of histone demethylases in cancer therapy. Mol Oncol 6, 683–703 (2012).
Pishas KI, et al. Therapeutic Targeting of KDM1A/LSD1 in Ewing Sarcoma with SP-2509 Engages the Endoplasmic Reticulum Stress Response. Mol Cancer Ther 17, 1902–1916 (2018).
Vollmar M, Johansson, C., Gileadi, C., Goubin, S., Szykowska, A., Krojer, T., Crawley, L., von Delft, F., Arrowsmith, C.H., Bountra, C., Edwards, A., Oppermann, U. Crystal structure of JmjC domain of human histone 3 Lysine-specific demethylase 3B (KDM3B).) (2013).
Sanson KR, et al. Optimized libraries for CRISPR-Cas9 genetic screens with multiple modalities. Nat Commun 9, 5416 (2018).
Patel AG, et al. The myogenesis program drives clonal selection and drug resistance in rhabdomyosarcoma. Dev Cell 57, 1226–1240 e1228 (2022).
Sunkel BD, et al. Evidence of pioneer factor activity of an oncogenic fusion transcription factor. iScience 24, 102867 (2021).
Xie Z, et al. Gene Set Knowledge Discovery with Enrichr. Current Protocols 1, e90 (2021).
Ashburner M, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25, 25–29 (2000).
Haydn T, Metzger E, Schuele R, Fulda S. Concomitant epigenetic targeting of LSD1 and HDAC synergistically induces mitochondrial apoptosis in rhabdomyosarcoma cells. Cell Death Dis 8, e2879 (2017).
Walters ZS, et al. Role for the Histone Demethylase KDM4B in Rhabdomyosarcoma via CDK6 and CCNA2: Compensation by KDM4A and Apoptotic Response of Targeting Both KDM4B and KDM4A. Cancers (Basel) 13, (2021).
Singh S, et al. Targeting KDM4 for treating PAX3-FOXO1-driven alveolar rhabdomyosarcoma. Sci Transl Med 14, eabq2096 (2022).
Xu X, et al. Small molecular modulators of JMJD1C preferentially inhibit growth of leukemia cells. Int J Cancer 146, 400–412 (2020).
Heinemann B, et al. Inhibition of demethylases by GSK-J1/J4. Nature 514, E1-E2 (2014).
Metzger E, et al. LSD1 demethylates repressive histone marks to promote androgen-receptor-dependent transcription. Nature 437, 436–439 (2005).
Dalvi MP, et al. Taxane-Platin-Resistant Lung Cancers Co-develop Hypersensitivity to JumonjiC Demethylase Inhibitors. Cell Rep 19, 1669–1684 (2017).
Kim JY, et al. KDM3B Is the H3K9 Demethylase Involved in Transcriptional Activation of lmo2 in Leukemia. Mol Cell Biol 32, 2917–2933 (2012).
An MJ, et al. Histone demethylase KDM3B regulates the transcriptional network of cell-cycle genes in hepatocarcinoma HepG2 cells. Biochem Biophys Res Commun 508, 576–582 (2019).
Shern JF, et al. Comprehensive genomic analysis of rhabdomyosarcoma reveals a landscape of alterations affecting a common genetic axis in fusion-positive and fusion-negative tumors. Cancer Discov 4, 216–231 (2014).
Stuart T, et al. Comprehensive Integration of Single-Cell Data. Cell 177, 1888–1902 e1821 (2019).
Traag VA, Waltman L, van Eck NJ. From Louvain to Leiden: guaranteeing well-connected communities. Sci Rep 9, 5233 (2019).
McInnes L, Healy J, Melville J. Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:180203426, (2018).
McDavid A, Finak G, Yajima M. MAST: Model-based Analysis of Single Cell Transcriptomics, R package version 1.18.0.). Bioconductor (2021).
Korotkevich G, Sukhov V, Budin N, Shpak B, Artyomov MN, Sergushichev A. Fast gene set enrichment analysis. bioRxiv, 060012 (2021).
Yohe M, et al. MEK inhibition induces MYOG and remodels super-enhancers in RAS-driven rhabdomyosarcoma. Sci Transl Med 10, (2018).
Azorsa DO, et al. Immunohistochemical detection of PAX-FOXO1 fusion proteins in alveolar rhabdomyosarcoma using breakpoint specific monoclonal antibodies. Mod Pathol 34, 748–757 (2021).
Ernst J, Kellis M. ChromHMM: automating chromatin-state discovery and characterization. Nat Methods 9, 215–216 (2012).
Ramirez F, Dundar F, Diehl S, Gruning BA, Manke T. deepTools: a flexible platform for exploring deep-sequencing data. Nucleic Acids Res 42, W187-191 (2014).
Kruse K, Hug CB, Vaquerizas JM. FAN-C: a feature-rich framework for the analysis and visualisation of chromosome conformation capture data. Genome Biol 21, 303 (2020).
Shen L. GeneOverlap: An R package to test and visualize gene overlaps (2016).
Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
Wolff J, et al. Galaxy HiCExplorer 3: a web server for reproducible Hi-C, capture Hi-C and single-cell Hi-C data analysis, quality control and visualization. Nucleic Acids Res 48, W177-W184 (2020).
Hart T, et al. High-Resolution CRISPR Screens Reveal Fitness Genes and Genotype-Specific Cancer Liabilities. Cell 163, 1515–1526 (2015).
Leung IK, et al. Reporter ligand NMR screening method for 2-oxoglutarate oxygenase inhibitors. Journal of medicinal chemistry 56, 547–555 (2013).
Calabrese DR, Connelly CM, Schneekloth Jr JS. Ligand-observed NMR techniques to probe RNA-small molecule interactions. In: Methods in enzymology). Elsevier (2019).

There is NO Competing Interest.

SupplementaryTable1.xlsx
Supplementary Table 1
SupplementaryTable2.xlsx
Supplementary Table 2
SupplementaryTable3.xlsx
Supplementary Table 3
SupplementaryTable4.xlsx
Supplementary Table 4
SupplementaryTable5.xlsx
Supplementary Table 5
SupplementaryTable6.xlsx
Supplementary Table 6
SupplementaryTable7.xlsx
Supplementary Table 7
SupplementalFigures.docx
Supplementary Figures

Download PDF

Journal Publication

published 24 Feb, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

KDM3B inhibitors disrupt PAX3-FOXO1 oncogenic activity in fusion positive rhabdomyosarcoma.

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Small Molecule Screen Identifies Inhibitor of PAX3-FOXO1 Activity

PFI-63 Targets Histone Lysine Demethylases

Identification of Chemical Analog of PFI-63 with improved drug-like properties

Biophysical Interaction Between PFI-90 and KDM3B

KDM3B is Essential for Survival in FP-RMS

KDM Knockdown Recapitulates the PFI-90-induced Transcriptome

scRNA-seq Shows PFI-90-Induced Decrease PAX3-FOXO1 Signature, Myoblast Signature, and G2/M Population

PFI-90 Increases H3K4 and H3K9 Methylation

PFI-90 Decreases Chromatin Loop Strength and Decreases Topologically Associating Domains (TADs) at PAX3-FOXO1 Sites

PFI-90 Decreased RNA Pol2 Binding and Phosphorylation

PFI-90 Inhibits FP-RMS Tumor Growth In Vivo

Discussion

Materials And Methods

Cell Lines

Small Molecules

In Vitro Dose Response

In Vitro Enzyme Inhibition Assay

RNA-seq

scRNA-seq

ChIP-seq

ChromHMM

ATAC-seq

Hi-C

Tiled KDM3B CRISPR Knockout Screen

Western Blot Analysis

Apoptosis Quantification

Flow Cytometry

Protein Purification

KDM3B Enzyme Activity Assay

NMR

SPR

CRISPRi Knockdown of KDMs

In Vivo Xenograft Model of FR-RMS

Statistical Analysis

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1