Article Images (0)
×![]()
featured_play_listSave article to folder
X
Create new folder
Full Text
PDF
Table of Contents
Summary
Graphical Abstract
Highlights
Abstract
Introduction
Results
Discussion
Limitations Of The Study
STAR★Methods
Key Resources Table
Resource Availability
Lead Contact
Materials Availability
Data And Code Availability
Experimental Models And Study Participant Details
Cell Culture
Generation Of QC6352-Resistant Cells
Mouse Transgenic And Xenograft Models
Method Details
Generation Of CRISPR-Cas9-Mediated Knockout Cells
Compound Resource And QC6352 Synthesis
SDS-PAGE And Western Blot
Crystal Violet Staining
Prestoblue Assay
Flow Cytometry Analysis Of Apoptosis By Annexin V Staining
DNA Fiber Assay
Immunohistochemistry
SiRNA Transfection
RNA-Seq
ADRN-MES Scoring System
Chromatin Immunoprecipitation Sequencing (ChIP-Seq)
CUT&Tag
ChIP-Seq And CUT& Tag Data Processing
Peak Overlap And Annotation
Visualization
Motif Analysis
Differential Analysis
ChIPseqSpikeInFree Normalization For H3K9me3 And H3K36me3
Heatmap Generation For Chromatin Co-Localization Of KDM4 And ADRN CRC TFs
Assay For Transposase-Accessible Chromatin Using Sequencing (ATAC-Seq) And Analysis
Animal Experiments
Quantification And Statistical Analysis
Supplemental Information
Author Contributions
Declaration Of Interests
Footnotes
Summary
Neuroblastoma with MYCN amplification (MNA) is a high-risk disease that has a poor survival rate. Neuroblastoma displays cellular heterogeneity, including more differentiated (adrenergic) and more primitive (mesenchymal) cellular states. Here, we demonstrate that MYCN oncoprotein promotes a cellular state switch in mesenchymal cells to an adrenergic state, accompanied by induction of histone lysine demethylase 4 family members (KDM4A-C) that act in concert to control the expression of MYCN and adrenergic core regulatory circulatory (CRC) transcription factors. Pharmacologic inhibition of KDM4 blocks expression of MYCN and the adrenergic CRC transcriptome with genome-wide induction of transcriptionally repressive H3K9me3, resulting in potent anticancer activity against neuroblastomas with MNA by inducing neuroblastic differentiation and apoptosis. Furthermore, a short-term KDM4 inhibition in combination with conventional, cytotoxic chemotherapy results in complete tumor responses of xenografts with MNA. Thus, KDM4 blockade may serve as a transformative strategy to target the adrenergic CRC dependencies in MNA neuroblastomas.
Graphical abstract
Highlights
• MYCN promotes KDM4 induction and cellular state switch from mesenchymal to adrenergic • KDM4 inhibition induces H3K9me3 and blocks MYCN and adrenergic transcriptome • KDM4 inhibition leads to significant suppression of neuroblastoma growth • KDM4 inhibition in combination with chemotherapy results in complete tumor responses
Abstract
Abu-Zaid et al. show that KDM4 inhibition blocks MYCN function in high-risk neuroblastoma models. The combination of a selective KDM4 inhibitor with chemotherapy reduces tumor growth of neuroblastoma xenografts.
Introduction
Neuroblastoma is a childhood malignancy of the peripheral nervous system caused by a differentiation arrest of neural crest-derived progenitor cells that are transformed by MYCN or MYC oncoproteins. 1 , 2 , 3 , 4 MYCN is amplified in 50% of high-risk neuroblastomas that are typically aggressive entities, associated with rapid progression, metastasis, and disease relapse. 5 , 6 There exist two distinct types of neuroblastoma cells regulated by their respective core regulatory circuitry transcription factors (CRC TFs), namely adrenergic (i.e., PHOX2B, HAND2, and ASCL1) and neural crest-like or mesenchymal (i.e., AP-1, PRRX1, and NOTCH) lineage-specific TFs, respectively. 7 , 8 , 9 , 10 CRC TFs are defined as a small set of master TFs that determine cellular identity and collectively regulate both their own gene expression, thus forming an interconnected auto-regulatory loop and the majority of gene expression in the cell. 11 Prior studies have indicated that many neuroblastoma cell lines with MNA are adrenergic (ADRN), in contrast to most non-MNA cell lines that are mesenchymal type (MES). 7 , 8 Nevertheless, some non-MNA cell lines (i.e., SKNFI) were reported to express an ADRN signature. 12 The ADRN neuroblastoma cells exhibit strong lineage dependency on CRC TFs. 7 , 9 , 10 , 13 , 14 Thus, exploring the transcriptional dependencies conferred by CRC TFs may provide an opportunity for the development of targeted therapies for high-risk neuroblastomas. One feature of the CRC TFs is that their encoding genes bear enhancers that are heavily marked with transcriptionally active histone modifications, 11 indicating that the enhancers that drive high-rate transcription of CRC TFs could be susceptible to perturbation of epigenetic modifiers. 15 , 16 Several studies have reported that histone methyltransferases and lysine demethylases regulate the viability and differentiation of MNA neuroblastoma cells. 17 , 18 , 19 , 20 , 21 , 22 , 23 However, whether these histone methylation modulators are involved in regulating CRC TFs in neuroblastomas has not been well studied. In addition, whether the activities of CRC TFs can be switched off by modulating histone methylation status, and whether this can be translated into a therapeutic approach, have not been rigorously explored. The JmjC domain containing KDM4 demethylases are responsible for removing H3K9me3/me2, 24 thereby antagonizing the heterochromatin-mediated gene silencing. 25 The KDM4 subfamily consists of five members that are overexpressed in many types of tumors and are required for efficient cancer cell growth, 26 , 27 , 28 , 29 , 30 and regulate copy number gain of oncogenes, 31 , 32 ribosomal biogenesis and protein translation, 33 , 34 and alternative splicing of androgen receptors. 35 KDM4s are also required for efficient cancer cell transformation induced by oncogenic TFs. 28 , 29 , 30 We recently reported that KDM4 inhibition suppresses the functions of PAX3-FOXO1, a master TF and key oncogenic driver of alveolar rhabdomyosarcoma, and downregulates the myogenic transcriptome defined by CRC TFs. 36 We therefore surmised that similar blockade of KDM4 activity in neuroblastoma might interfere with the activities of CRC TFs by inducing H3K9me3-mediated repression of active gene transcription. Here, we integrate these concepts and demonstrate that MYCN enables the conversion of neuroblastoma cells from a MES to ADRN state by inducing ADRN CRC TFs and suppressing mesenchymal CRC TF expression. These findings are accompanied by the induction of KDM4 protein expression that is required for the expression and activity of MYCN and ADRN CRC TFs in neuroblastoma cells, which can be reversed by a selective KDM4 inhibitor. This study provides a proof-of-concept that pharmacologic inhibition of KDM4 may be an attractive strategy as a therapeutic approach for high-risk MNA neuroblastomas.
Results
MYCN transdifferentiates neuroblastoma cells from MES to ADRN state with induction of KDM4 members
RNA-seq analysis of the MYCN-transformed neural crest cells revealed an induction of adrenergic markers (i.e., DBH and PHOX2B ) and repression of PRRX1 , a CRC TF associated with the mesenchymal state ( Figure S1 A), suggesting that MYCN expression may induce a plastic cellular state transition. To further verify the capacity of MYCN in mediating a cellular state switch from the MES to ADRN, we transduced MYCN into SK-N-SH cells, a neuroblastoma cell line with mesenchymal state. 7 , 8 RNA-seq analysis with an ADRN-MES scoring algorithm for the reported ADRN and MES gene signatures 8 demonstrated that MYCN expression led to a shift of transcriptome from MES to ADRN ( Figure 1 A), consistent with the gene set enrichment analysis (GSEA), showing the upregulation of ADRN CRC TFs by MYCN ( Figure 1 B). Western blotting confirmed that MYCN overexpression induced the expression of adrenergic CRC TF proteins (PHOX2B, HAND2, and ASCL1) and the oncoprotein LIN28B, which functions as a promoter binding factor associated with the ADRN CRC loop, 37 and conversely reduces the expression of several proteins associated with the mesenchymal signature (PRRX1, FIBRONECTIN, VIMENTIN, and AXL) ( Figure 1 C), which was further verified in a second MES cell line GIMEN with MYCN induction by western blots and RNA-seq following GSEA ( Figures S1 B–S1D). These data further indicate that MYCN may alter the mesenchymal state of neuroblastoma cells to a more adrenergic state, at least partially. Nevertheless, considering the high heterogeneity of neuroblastoma, whether the ADRN/MES state in each cell could be equally affected by MYCN awaits further studies using scRNA-seq approaches. Figure 1 MYCN induces transdifferentiation of neuroblastoma cells from MES to ADRN state accompanied by induction of KDM4s that promote tumorigenesis (A) ADNR vs. MES gene score in SK-N-SH cells expressing mCherry vs. MYCN. (B) GSEA of RNA-seq data from SK-N-SH cells transduced with mCherry vs. MYCN using the ADRN CRC TF gene set. (C) Western blot analysis of parental SK-N-SH cells, and those transduced with MSCV-mCherry and MSCV-MYCN with the indicated antibodies. ∗NS, non-specific. (D) Western blot analysis of cell lysate of BE2C and SIMA cells transfected with MYCN siRNA using the indicated antibodies. (E) Crystal violet staining showing the colony formation of BE2C parental cells (n = 3) and with knockout of KDM4A (n = 3), KDM4B (n = 3), KDM4C (n = 3), and corresponding controls (n = 3 per group). (F–H) Xenograft tumor growth curve of BE2C cells with knockout of KDM4A, KDM4B, and KDM4C (n = 4 per group) and the corresponding controls (n = 4 per group) implanted into NSG mice. Data are mean ± SEM. Student’s t test ∗∗p < 0.05. (I) Western blot analysis of KDM4 knockout in BE2C cells used for experiments in (F–H) with the indicated antibodies. GSEA of RNA-seq of 26 neuroblastoma cell lines (19 with MNA, 7 without MNA) from the Cancer Cell Line Encyclopedia revealed that, in comparison with the non-MNA cells, the MNA cells were significantly enriched with histone lysine modifying genes ( Figures S1 E and S1F), including those responsible for methylating histones. These data support that epigenetic modifiers may act in concert and are particularly important in MNA cells. We therefore reasoned that MNA cells may be more vulnerable to disruption of the epigenetic balance mediated by both methyltransferases and demethylases. Previously, we showed that the histone demethylase families KDM4 and KDM6 regulate MYCN activity in neuroblastoma. 20 , 23 , 38 Intriguingly, distinct from KDM6B induction by MYCN, we identified that MYCN induced the expression of KDM4A-C at the protein levels but not at the transcriptional levels ( Figure 1 C; Table S1 ). Despite this, MYCN oncoprotein directly binds at the promoter regions of KDM4 in primary MNA tumors ( Figure S1 G), whose epigenetic landscapes have been extensively characterized. 39 In these tumors, MYCN binds at the promoter regions of KDM4A and KDM4C , where transcriptionally active marks (H3K4me3 and H3K27Ac) were also present. At the KDM4B genomic locus, MYCN and H3K27Ac showed a region of broad occupancy across the whole KDM4B gene (100 kb in length) ( Figure S1 G), indicative of a super-enhancer at the KDM4B locus. RNA-seq analysis of a large cohort of neuroblastomas 40 revealed that KDM4A-4C were expressed at much higher levels than KDM4D and KDM4E and that the expression of KDM4B was significantly higher than KDM4A and KDM4C in MNA and high-risk neuroblastomas ( Figures S1 H). Spearman correlation revealed that only KDM4B and KDM4D at the transcriptional level were positively correlated with MYCN , and KDM4A and KDM4C were negatively correlated with MYCN but positively correlated with MYC ( Figure S1 I). Western blotting further showed that KDM4 proteins were overall highly expressed in MNA xenografts compared with the non-MNA xenografts ( Figure S1 J). Although some non-MNA xenografts also expressed high levels of KDM4 proteins. Very surprisingly, C-MYC was not only highly expressed in the two out five of non-MNA tumors that expressed high levels of KDM4, but also was highly expressed in three out five of MNA tumors with different isoforms. While it remains to be studied for the biological functions of C-MYC in MNA tumors, these data may suggest that the patient-derived xenograft (PDX) tumors were highly heterogeneous and lacked a clear expression correlation of MNA and KDM4s. Knockdown of MYCN in BE2C and SIMA cells greatly reduced the expression of KDM4s and ADRN CRC TFs but showed minimal effect on the MES gene VIMENTIN ( Figure 1 D), further demonstrating that MYCN is an important regulator of ADRN CRC TFs and KDM4s. The induction of KDM4 protein expression by MYCN indicates that KDM4 may be important in MNA neuroblastoma. Indeed, genetic deletion of KDM4A-C inhibited colony formation of BE2C cells ( Figure 1 E), and further delayed the growth of BE2C xenografts ( Figures 1 F–1H), indicating that KDM4A-C are involved in regulation of MNA tumor growth. Notably, western blotting demonstrated that KDM4B knockout led to reduction of KDM4A and KDM4C ( Figure 1 I), while this effect was maintained in tumors for KDM4C but not for KDM4A ( Figure S1 K), suggesting that KDM4 family members may co-regulate each other’s expression in BE2C cells. However, in KELLY cells, knockout of KDM4B had negligible effect on expression of KDM4A and KDM4C using the same gRNA ( Figure S1 L), indicating that the effect of KDM4B deletion on KDM4A and KDM4C in BE2C cells was not due to off-target effects; rather, this coregulation is cell line dependent.
KDM4 proteins maintain the ADRN transcriptome of neuroblastoma cells with MYCN amplification
To determine the roles of KDM4 in MNA neuroblastoma, we depleted KDM4A-C in two MNA cell lines representing the ADRN state (BE2C and KELLY) and one non-MNA cell representing the MES state (SK-N-AS), and performed RNA-seq to investigate the effect of KDM4 on total gene expression. In BE2C cells, loss of KDM4A and KDM4B significantly downregulated the ADRN gene signature but upregulated the MES signature, while KDM4C depletion showed a more significant effect on MES upregulation ( Figure 2 A). Similar results were obtained in KELLY and SK-N-AS cells, in which depletion of KDM4 genes either downregulated an ADRN signature and/or upregulated a MES signature ( Figure S2 A). We specifically examined the MES CRC TFs including NOTCH3, AP-1 components (FOS, FOSL, JUN, and JUNB), which were significantly upregulated by KDM4 knockdown ( Figure S2 B). These data indicate that KDM4 proteins are dynamically involved in regulating the transcriptional state of neuroblastoma cells. Figure 2 KDM4 maintain the ADRN state of neuroblastoma cells and KDM4 inhibition downregulates oncogenes and the MYC pathway, and upregulates the type I interferon pathway (A) GSEA of RNA-seq data from BE2C cells with KDM4A, KDM4B, and KDM4C depletion using the reported gene signatures of ADRN and MES. (B) Heatmap showing the genomic binding sites of KDM4A, KDM4B, and KDM4C that are common among all KDM4s, shared by any two KDM4s, or occupied by individual KDM4s (left panel). HOMER Motif analysis results showing the transcription factor binding motifs at these loci of KDM4 binding (right panel). (C) Heatmap showing the chromatin co-localization of KDM4 and ADRN CRC TFs at promoter and enhancer regions of ADRN and MES gene signatures. (D) Volcano plot showing differential gene expression by RNA-seq analysis of cells with MNA and without MNA treated with 100 nM of QC6352 for 48 h. (E) GSEA pathway analysis showing the downregulated and upregulated pathways induced by QC6352 in MNA and non-MNA cells. (F) Venn diagram showing the genes downregulated (left) and upregulated (right) in BE2C, KELLY, and SK-N-AS cells by KDM4 knockdown and QC6352 treatment. (G) Venn diagram showing the genes downregulated (left) and upregulated (right) in BE2C, SIMA, KELLY, SK-N-AS, and SK-N-SH cells by QC6352. (H) Venn diagram showing the neuroblastoma fitness genes shared with the genes downregulated in MNA (BE2C, SIMA, and KELLY) and non-MNA (SK-N-AS and SK-N-SH) cells by QC6352. To understand the mechanism by which KDM4 proteins regulate ADRN/MES transcriptomes, we performed cleavage under targets and tagmentation (CUT&TAG) to map the chromatin binding sites of KDM4 proteins, followed by motif enrichment analysis for the TFs at KDM4 binding sites. While there were common binding sites among KDM4 proteins (6,546 for KDM4A, KDM4B, and KDM4C; 384 for KDM4A and KDM4B; 409 for KDM4B and KDM4C), a substantial number of binding sites were specific to individual KDM4 members (66,703 for KDM4A; 14,986 for KDM4B; 41,841 for KDM4C) ( Figures 2 B and S2 C). Hypergeometric optimization of motif enrichment (HOMER) Motif analysis, a motif discovery algorithm designed for regulatory element analysis in genomics applications, 41 revealed that both ADRN and MES CRC TF 42 motifs were enriched at the common binding sites or individual KDM4 binding sites ( Figures 2 B and S2 D). Nevertheless, more ADRN CRC TF motifs appeared to be enriched than MES CRC TF motifs ( Figure 2 B), supporting that KDM4 proteins are involved in regulating the ADRN and MES transcriptome in neuroblastoma cells. We then performed an analysis to examine the co-localization of KDM4 and ADRN CRC TFs at adrenergic and mesenchymal CRC gene promoters and super-enhancers. We used the published ChIP-seq data that included genomic binding of MYCN, HAND2, ISL1, PHOX2B, GATA3, and TBX2 in BE2C cells. 14 There was a strong co-localization of KDM4 with ADRN CRC TFs at the promoter regions of both ADRN and MES genes, while the co-localization of KDM4 and ADRN CRC TFs were weaker at the super-enhancer regions ( Figure 2 C). These data suggest that physical occupation of these ADRN CRC TFs and KDM4 at genomic loci may not solely determine the cellular fate. Nevertheless, these data support the hypothesis that KDM4, MYCN, and ADRN CRC TFs work in concert to regulate gene expression.
KDM4 inhibition represses expression of neuroblastoma oncogenes and adrenergic CRC transcription
To further determine the molecular effects of KDM4 inhibition on neuroblastoma cells (MNA: BE2C, SIMA, and KELLY; non-MNA: SK-N-AS and SK-N-SH), we treated cells with 100 nM QC6352, a selective KDM4 inhibitor, for 48 h and performed RNA-seq for differential gene expression analysis ( Figure 2 D), followed by GSEA ( Figures 2 E and S2 E). In the three MNA cell lines, QC6352 significantly reduced the mRNA expression of MYCN , in addition to several other ADRN-associated genes such as LIN28B , ALK , ASCL1 , and HAND2 ( Figures 2 D and S2 E). LIN28B is an oncogene in neuroblastoma. 43 , 44 , 45 LIN28B represses expression of let-7 miRNA and consequently elevates MYCN expression. 44 However, LIN28B also has a let-7-independent role in regulating ADRN CRC TF transcription by mediating protein-protein interactions with the sequence-specific zinc-finger TF ZNF143. 37 ALK is another neuroblastoma oncogene that is mutated or amplified in a subgroup of patients 46 and enhances MYCN activity. 47 HAND2 is a member of the adrenergic CRC TFs that maintain cells in adrenergic state, 7 , 14 as is ASCL1 . 10 These data revealed that KDM4 inhibition leads to repression of key neuroblastoma oncogenes, each of which is a neuroblastoma-lineage essential gene as demonstrated by DepMap data ( www.depmap.org ) 48 and, indeed, many low-throughput confirmatory experiments. 7 , 10 , 14 Interestingly, in the non-MNA cell lines that were less responsive to QC6352, adrenergic CRC TFs ( HAND2 , PHOX2B , TFAP2B , and ASCL1 ) were also downregulated by QC6352 treatment ( Figure 2 D). GSEA showed that QC6352 significantly reduced MYC and ADRN gene signatures but induced a type I interferon response in all tested cell lines ( Figure 2 E). This might emerge from activation of tumor-cell-intrinsic cGAS-STING signaling due to replication-stress-induced cytosolic DNA accumulation by KDM4 inhibition. 49 Consistent with KDM4 knockdown, QC6352 induced the expression of MES CRC TFs such as NOTCH3 and AP1 components ( FOS and FOSL1 ) ( Figure S2 F). Interestingly, QC6352 inhibited the expression of PRRX1 in non-MNA cells while MNA cells showed no detectable expression of PRRX1 . These data indicate that KDM4 inhibition has a complex outcome for MES CRC TFs. We then compared the differential genes induced by QC6325 with KDM4 knockdowns ( Figure 2 F). While the Venn diagram showed various overlapping genes whose expression was altered by QC6352 and individual KDM4 knockdown, each condition also induced unique gene expression and QC6352 had a broader effect. To further understand the differential impact of QC6352 on MNA vs. non-MNA cells, we analyzed the RNA-seq data from all cell lines treated with QC6352 by using Venn analysis ( Figure 2 G). While there were common genes downregulated (n = 353) and upregulated (n = 141) among all cell lines, QC6352 induced differential gene expression between MNA and non-MNA cells even more dramatically in each cell line, suggesting inter-tumor cell heterogeneity may also affect the QC6352 response. To examine whether the differential genes affected by QC6352 between MNA and non-MNA cells could be important for neuroblastoma, we compared the neuroblastoma dependency genes from genome-wide screening using the DepMap program with the common genes downregulated in MNA or non-MNA cells ( Figure 2 H). We found that 34 neuroblastoma dependency genes were downregulated by QC6352 in both MNA and non-MNA cells, including the ADRN CRC TF genes ASCL1 and HAND2 , as well as DBH ( Table S2 ), another commonly used ADRN marker. For MNA cells, QC6352 particularly downregulated the neuroblastoma oncogenes among the 43 neuroblastoma dependency genes, including ALK , MYCN , IGF2 , and LIN28B ( Table S2 ), all of which are known to be oncogenic drivers and essential to MNA cell survival. For non-MNA cells, QC6352 downregulated the expression of 37 neuroblastoma dependency genes, including two ADRN CRC TF genes ISL1 and PHOX2B ( Table S2 ). Immunoblotting confirmed that KDM4 inhibition led to reduction of KDM4 members, MYCN, and adrenergic CRC TFs (ASCL1, HAND2, and PHOX2B) ( Figure S2 G). Considering that these non-MNA cells are often a mixture of MES with ADRN, some ADRN genes were also downregulated by QC6352. We therefore further investigated the effect of QC6352 on GIMEN, which was thought to be a pure mesenchymal cell line. We performed RNA-seq and western blot analyses after GIMEN cells were treated with the QC6352 for 48 h. Interestingly, GIMEN cells express high levels of ADRN CRC TFs (ASCL1, LIN28B, and HAND2) in addition to MES markers (VIMENTIN, C-MYC, and PRRX1). These data raised one hypothesis that mesenchymal cells such as GIMEN may not be purely mesenchymal and also express adrenergic markers. Nevertheless, GSEA results showed that QC6352 inhibited the ADRN signature and appeared to further increase the expression of MES signature albeit not significantly ( Figure S2 H). Western blot analysis validated that QC6352 reduced the expression of ADRN TFs (ASCL1, LIN28B, and HAND2) and MES TFs (C-MYC and PRRX1) except NOTCH3 ( Figure S2 I). LIG4 and ATR genes, key components in DNA damage response, were also specifically downregulated by QC6352 in MNA cells ( Table S2 ). ATR inhibition displayed anti-tumor activity in neuroblastoma models. 50 , 51 Pathway enrichment analysis showed that QC6352 downregulates the neuroblastoma dependency genes involved in pre-mRNA splicing in both MNA and non-MNA cells, receptor tyrosine kinase signaling in MNA cells, and GPCR signaling in non-MNA cells ( Figure 2 H). Taken together, these data indicate that KDM4 inhibition results in downregulation of MYCN and multiple adrenergic CRC transcriptional factors.
KDM4 inhibition disrupts the chromatin accessibility of adrenergic CRC transcriptional factors in MYCN -amplified cells
GSEA demonstrated that KDM4 inhibition reduced the ADRN transcriptional signatures in all cell lines tested ( Figure S2 ), consistent with the results from an ADRN-MES scoring algorithm showing that KDM4 inhibition led to a reduction of ADRN transcriptomic pattern ( Figures 3 A and 3B). To investigate how KDM4 inhibition affects expression of the key oncogenes and adrenergic CRC TFs, we performed ATAC-seq (assay for transposase-accessible chromatin using sequencing), 52 an assay for assessing the chromatin accessibility, after treatment of cell lines with DMSO or QC6352. HOMER Motif analysis of TF binding affected by QC6352 revealed that inhibiting KDM4 significantly reduced the chromatin accessibility of genes bearing ASCL1, GATA3, ISL1, MYCN, PHOX2A binding motifs ( Figures 3 C, 3D, and S3 A–S3C), and other TFs that are involved in developmental processes ( Figure S3 E). Functional enrichment analysis of protein-protein interaction networks revealed that these TFs constituted an interaction network ( Figures 3 E and S3 D) that determines super-enhancers and cell identity, and regulates gene transcription and regulatory circuitry in development and cancer ( Figure S3 E). Consistent with the RNA-seq and ATAC-seq data, QC6352 reduced chromatin accessibility at the gene loci of MYCN , HAND2 , and PHOX2B that encode adrenergic CRC TFs, and their targets such as ALK ( Figure 3 F). In summary, these data indicate that KDM4 inhibition results in suppressed chromatin accessibility at loci critical for the adrenergic gene expression program. Figure 3 QC6352 impacts the chromatin accessibility of ADRN CRC TFs in MYCN -amplified cells (A and B) Adrenergic or mesenchymal median score in response to DMSO or 100 nM of QC6352 treatment for 48 h in adrenergic dominant cell lines (BE2C, SIMA, and KELLY) (A) and mesenchymal dominant cell lines (SK-N-AS and SK-N-SH) (B). KDM4 inhibition results in enhanced MES scoring with reduced ADRN signatures in all tested cell lines (red arrow indicates direction of signature change). (C) HOMER Motif analysis showing the predicted adrenergic CRC TF binding motifs in which genome-wide chromatin accessibility is impacted by QC6352 treatment of SIMA cells. (D) PHOX2A, ASCL1, and MYCN motif densities around ATAC-seq open chromatin regions (±1,000 bp) by categories in SIMA cells. Motif density was determined by the HOMER program and normalized to that in a background sequence of equal length. (E) STRING network analysis of transcriptional factors predicted to bind the regions with reduced chromatin accessibility by QC6352 treatment of SIMA cells. The expression threshold for genes encoding TFs is set as log2CPM > 1. CPM, counts per million mapped reads. Red color indicates ADRN CRC TFs. Blue color indicates MES CRC TFs. (F) IGV snapshots show chromatin accessibility at MYCN , HAND2 , PHOX2B , and ALK loci is reduced by QC6352 in SIMA and BE2C cells. (G) Density heatmaps for CUT&TAG of KDM4A-C genomic binding in BE2C cells, and ChIP-seq density heatmaps of the indicated histone marks in BE2C cells after 48 h treatment with control or 200 nM QC6352, ranked by MYCN read intensity within ± 5 kb of peak summits. MYCN ChIP-seq data were retrieved from GEO SuperSeries GSE94824. (H) The indicated epigenetic marks upregulated and downregulated by QC6352 treatment of BE2C cells for 48 h followed by ChIP-seq or CUT&TAG analyses. (I) The global distribution of H3K9me3, H3K36me3, H3K27Ac, H3K4me1, and H3K27me3 peaks at different regions of annotated genes. 3′ UTR, 3′ untranslated region; 5′ UTR, 5′ untranslated region; TSS, transcription start site. (J) Snapshot of using the IGV program displaying the KDM4s, H3K9me3, H3K36me3, H3K27me3, H3K27Ac, and H3K4me1 peaks at the PHOX2B and ASCL1 genomic loci.
KDM4 inhibition alters the epigenetic landscape of MYCN -amplified cells
We next investigated whether KDM4 inhibition caused epigenetic changes in oncogenic transcription by examining the KDM4 substrates H3K9me3 and H3K36me3, as well as markers of active (H3K27ac and H3K4me1) and repressed (H3K27me3) enhancers and promoters, by performing ChIP-seq analysis in BE2C cells after treatment with 200 nM of QC6352 for 48 h. In accordance with the KDM4 inhibiting activity, QC6352 induced a global induction of H3K9me3 peaks (up 32,942 vs. down 4) that were largely aligned with KDM4 and MYCN binding sites ( Figures 3 G and 3H). These data also validated that QC6352 resulted in suppression of KDM4 protein activity, since the demethylase activity of KDM4 removes H3K9me3. Genomic annotation of H3K9me3 binding peaks induced by QC6352 showed that the major regions impacted by KDM4 inhibition were intergenic (38%) and intronic (51.8%) regions, with a smaller fraction of others that were promoters (4.2%) ( Figure 3 I). While KDM4 inhibition also induced global reprogramming of H3K36me3, the upregulated peaks were just slightly higher than the downregulated peaks (up 6,641 vs. down 5,243) ( Figure 3 H). KDM4 members have distinct binding modes to H3K9me3 and H3K36me3, and in vitro kinetics studies showed that KDM4A-4C preferentially catalyzes the demethylation of H3K9me3 over H3K36me3. 53 These factors may explain the differential response of H3K9me3 and H3K36me3 to QC6352 in BE2C cells. The H3K36me3 peaks altered by QC6352 mainly occurred at intronic regions (up 80% vs. down 83%) ( Figure 3 I), consistent with the feature of gene body occupation of H3K36me3. The enhancer marks H3K27Ac and H3K4me1 were also broadly changed by KDM4 inhibition, with 5,040 up vs. 4,472 down for H3K27Ac and 9,239 up vs. 6,357 down for H3K4me1, respectively ( Figure 3 H). While the percentages of H3K27Ac at genomic regions (intergenic, intron, promoter) were comparable between the upregulated and downregulated peaks, H3K4me1 showed differences at intergenic and promoter regions ( Figure 3 I). The H3K27me3 peaks upregulated and downregulated by QC6352 were 2,878 and 1,131, respectively ( Figure 3 H), and the percentages of H3K27me3 at intergenic and promoters were remarkably altered by QC6352 between the upregulated and downregulated peaks. These data indicate that KDM4 inhibition not only has a direct impact on KDM4 substrates, but also may indirectly reprogram other epigenetic marks. Focused analysis of the genomic loci of two ADRN CRC TF genes, PHOX2B and ASCL1 , demonstrated a broad induction of H3K9me3 by QC6352 ( Figure 3 J), in addition to an increase of H3K27Ac and H3K4me1 at these loci, suggesting that H3K9me3 might be involved in inhibiting gene transcription at these loci. Then we specifically examined the MES CRC TF gene NOTCH3 and the AP-1 gene FOS , both of which were induced by KDM4 knockdown and QC6352 treatment. While the H3K9me3 showed minimal changes, H3K36me3 across the gene bodies and H3K27Ac at the promoter regions of NOTCH3 and FOS were greatly induced ( Figures S3 F and S3G). H3K36me3 and H3K27Ac serve as active gene transcription marks. These data indicate that KDM4 inhibition induces the transcriptional upregulation of MES CRC TFs. To further understand the importance of KDM4 in maintaining an adrenergic state, we generated QC6352-resistant BE2C cells by long-term culture of BE2C cells in high concentrations of QC6352 (2.5 μM), and performed RNA-seq and western blot analyses as well as ATAC-seq. We surmised that the MNA cells may undergo cellular fitness by acquiring a mesenchymal state when they develop resistance to KDM4 inhibition. Indeed, GSEA showed a great induction of retinoic acid-induced signature and MES signature in QC6352-resistant cells ( Figure 4 A). A heatmap showed the differential expression of ADRN/MES CRC TFs and relevant markers in parental and QC6352-resistant cells ( Figure 4 B). ATAC-seq revealed that the transcriptional factor binding motifs for MES CRC TFs (JUN, FOSL2, etc.) were highly enriched in QC6352-resistant BE2C cells ( Figures 4 C and 4D). We then used western blot to further validate that QC6352-resistant cells had a great reduction of ADRN markers (MYCN, LIN28B, ALK, PHOX2B, HAND2, and ASCL1) but increased MES markers (VIMENTIN and FN1) ( Figure 4 E), in line with the RNA-seq results. Collectively, these data have further substantiated our hypothesis that the MYCN-KDM4 axis plays an important role in regulating ADRN and MES cellular state. Figure 4 Mesenchymal state of BE2C cells after development of resistance to QC6352 (A) GSEA after RNA-seq for BE2C parental vs. QC6352 resistant cells showing the induction of signatures of all- trans retinoic acid and MES after BE2C cells develop resistance to QC6352. (B) Heatmap showing the differential expression of ADRN and MES CRC TFs and markers in BE2C parental vs. QC6352 resistant cells. (C) Volcano blot showing the transcriptional factor motif enrichment ranked by MD score in BE2C parental vs. QC6352 resistant cells. (D) Heatmap showing the motif displacement distribution of MES CRC TFs (FOSL1 and JUN) and ADRN CRC TFs (NFYA and ATF4). Increasing yellow indicates increasing motif frequency, MD score, and the number of this motif within 1.5 kb of an ATAC-seq peak in BE2C parental vs. QC6352-resistant cells. (E) Western blot analysis for ADRN and MES markers in BE2C parental vs. QC6352 resistant cells with the indicated antibodies.
Pharmacologic inhibition of KDM4 selectively affects neuroblastoma cells with MYCN amplification
Pharmacologic inhibition of KDM4 may be an effective strategy in MNA neuroblastoma cells whose survival is dependent on ADRN CRC TFs. To test the feasibility of translating KDM4 inhibition for the treatment of MNA neuroblastoma, we first treated a panel of neuroblastoma cell lines ( Table S3 ) with QC6352. A PrestoBlue assay simultaneously measuring cytostatic (growth inhibition) and cytotoxic (cell viability reduction) effect after 120 h treatment revealed that the MNA neuroblastoma cells were significantly more sensitive to QC6352 than the cells without MNA ( Figure 5 A), with a mean GI 50 (16 vs. 122 nM) and mean LC 50 (3.5 vs. 8.5 μM), respectively. These results were consistent with the colony formation assay for these cell lines treated with 100 nM of QC6352 for 10 days ( Figures 5 B and 5C). We noticed that three MNA cell lines, LAN5, NGP, and COG-N-496, had higher LC 50 values, although they showed comparable GI 50 values with other ADRN cell lines. While we did not have any ADRN/MES information for COG-N-496, LAN5, and NGP expressed an ADRN signature. 12 We therefore tentatively attributed the high LC 50 values of these cell lines to their doubling time (NGP ∼50–70 h, LAN5 ∼100 h). However, these cell lines could be outliers that are less responsive to other drugs as well and not just KDM4 inhibitors. In general, the MNA cells that were more sensitive to QC6352 expressed higher levels of KDM4 proteins ( Figure S4 A). CHLA20, a non-MNA cell line that expresses high levels of KDM4 proteins and MYC, rather than MYCN, was also sensitive to QC6352 treatment and was an outlier among the non-MNA cell lines ( Figures 5 A–5C). However, the non-cancerous human fibroblast cell lines (BJ and HS68), and murine neural crest cells, the cells of origin of neuroblastoma (O9-1) were resistant to QC6352, even at a concentration up to 10 μM ( Figures 5 B and 5C). It has been reported that KDM4 and p53 may regulate each other’s function. 54 , 55 , 56 However, p53 status did not show a significant correlation with QC6352 sensitivity in neuroblastoma cells ( Figure S4 B). To further determine the role of MYCN in response to QC6352, we used SHEP-N21 cell line with a Tet-off system in which MYCN expression could be rapidly silenced by administration of doxycycline. A PrestoBlue assay showed that GI 50 was increased from 39.16 to 234.62 nM when MYCN expression was silenced ( Figure 5 D), in line with the colony formation assay results showing that cells responded to QC6352 in a dose-dependent manner in the presence of MYCN but were resistant to QC6352 treatment when MYCN was silenced ( Figures 5 E and 5F). Figure 5 KDM4 inhibition demonstrates selectivity against neuroblastoma cells with MYCN amplification (A) PrestoBlue assay shows the cytostatic (GI 50 ) and cytotoxic (LC 50 ) effect of QC6352 on MYCN-amplified (MNA, n = 10) and non-MYCN amplified cells (non-MNA, n = 7). Student’s t test. (B) Colony formation assay stained by crystal violet after neuroblastoma cells and noncancerous cells were treated with 100 nM and 10 μM of QC6352, respectively, for 10 days. (C) Quantification of cell density from (B). Data are mean ± SEM. p values calculated by Student’s t test. ns, not significant; ∗p < 0.01, ∗∗p < 0.001, ∗∗∗p < 0.0001. (D) PrestoBlue assay showing the GI 50 of SHEP-N21 cells treated with QC6352 in the presence or absence of doxycycline administration. Data are mean ± SEM. (E) Colony formation assay of SHEP-N21 cells treated with QC6352 in the presence or absence of doxycycline administration. (F) Quantification of cell density from (E) by ImageJ. Data are mean ± SEM. p values calculated by Student’s t test ∗p < 0.01. (G) Tumor growth curves after 9 day treatment with vehicle (n = 8) or 50 mg/kg of QC6352 (n = 7), once daily, in SIMA xenografts implanted into NSG mice. Data are mean ± SEM. p values at each time point on day 8 afterward were 0.0048, 0.0012, 0.0012, 0.0016, and 0.0023, respectively (indicated as ∗∗, Wilcoxon rank-sum test, two-sided, p values adjusted using the Benjamini-Hochberg method). (H) Tumor growth curves after 9 day treatment with vehicle (n = 6) or 50 mg/kg of QC6352 (n = 8), once daily, in SK-N-AS xenografts implanted into NSG mice. Data are mean ± SEM. p values at each time point on day 9 afterward were 0.0450, 0.0527, 0.1214, 0.1709, and 0.1214, respectively. (I) SJNB-14 PDX tumor growth curve in CB17 SCID mice treated with vehicle (n = 7) and 25 mg/kg of QC6352 (n = 7), twice daily, for 3 weeks. Data are mean ± SEM. Adjusted p values for comparison of two groups at each time point from day 7 afterward were 0.0035, 0.0035, 0.0093, and 0.0667, respectively (indicated as ∗∗). (J) Kaplan-Meier survival curve for SJNB-14 PDX model treated with QC6352. (K) Transgenic TH-MYCN/ALK F1178L mouse model treated with 25 mg/kg of QC6352, twice daily, for 3 weeks. Tumor volume was monitored by ultrasound imaging. Control n = 8, QC6352 n = 8. Data are mean ± SEM. p values at each time point on week 1 and week 2 were 0.0866 and 0.0149, respectively. Due to number loss (5 out of 8) of mice that died of tumor progression on week 3 in vehicle group, statistical calculation was excluded (indicated as ∗∗, Wilcoxon rank-sum test, two-sided, p values adjusted using the Benjamini-Hochberg method). (L) Kaplan-Meier analysis of survival rate of transgenic TH-MYCN/ALK F1178L mice treated with 25 mg/kg of QC6352, twice daily, for 3 weeks. We then tested the efficacy of QC6352 in vivo in two neuroblastoma xenograft models: SIMA and SK-N-AS. Based on reported pharmacokinetics studies, 57 , 58 mice were dosed with 50 mg/kg orally, once daily when tumor volumes reached >100 mm 3 . Treatment was stopped after a continuous 9-day dosing when SIMA tumors underwent complete regression ( Figure 5 G). However, QC6352 was much less effective against the SK-N-AS xenografts ( Figure 5 H). These data were consistent with our in vitro studies. We further tested QC6352 for a 3-week dosing schedule (25 mg/kg, twice daily) in three additional neuroblastoma models: (1) NB-1691, an MNA xenograft model that is resistant to most current therapeutic regimens, (2) SJNB-14 (PDX 14), a PDX model with MNA, derived from a relapsed patient, and (3) an immunocompetent transgenic model ( TH-Mycn/Alk F1178L ), an autochthonous neuroblastoma model driven by Mycn and Alk F1178L , in which the F1178L mutant in Alk is equivalent to human F1174L, the predominant ALK mutation in human neuroblastoma. 59 , 60 , 61 , 62 ALK F1174L potentiates the oncogenic activity of MYCN in neuroblastoma. 63 KDM4 inhibition significantly delayed tumor progression and extended animal survival in all three MYCN -driven neuroblastoma models without body weight loss ( Figures 5 I–5L and S4 C–S4F). Western blot analysis of SIMA tumors revealed that QC6352 caused a global elevation of H3K9me3 and H3K36me3, reduced the expression of MYCN, LIN28B, HAND2, and KDM4, but induced the MES marker VIMENTIN ( Figure S5 G). These data indicate that inhibiting KDM4 has a profound therapeutic effect on MNA neuroblastomas, with a lesser effect on non-MNA neuroblastomas.
KDM4 inhibition induces differentiation and apoptosis in MYCN -amplified cells
The effects of KDM4 inhibition by QC6352 on cell proliferation and survival was a slow process. During this process, QC6352 induced remarkable cell morphology changes, such as neurite outgrowth, in cell lines with MNA ( Figure 6 A). These findings are characteristic of neuroblastoma cell differentiation, mimicking the phenotype of isotretinoin therapy. 64 , 65 , 66 Accordingly, hematoxylin and eosin staining of SIMA xenografts revealed that KDM4 inhibition by QC6352 resulted in a phenotype of tumor cell differentiation, as characterized by scattered tumor cells surrounded by large amounts of pink stromal neuropil, the product of neuritic processes ( Figure 6 B). To test if the morphology would shift back to the pre-treatment morphology, we performed the experiment by inducing differentiation first, then withdrew QC6352, with 13- cis -retinoic acid as a control. Indeed, we observed that the morphology shifted back after washout of QC6352 and 13- cis -retinoic acid ( Figure S5 A), indicating phenotypic plasticity. Immunohistochemistry staining for the cleaved caspase-3, an apoptotic marker, was also significantly induced by QC6352 ( Figures 6 C and 6D), indicating that KDM4 inhibition leads to apoptosis of neuroblastoma cells. To determine if KDM4 inhibition altered the cellular states in vivo , we performed RNA-seq analysis of SIMA xenografts treated with QC6352 vs. vehicle, and GSEA results showed a significant inhibition of adrenergic signature genes ( Figure 6 E). GSEA of RNA-seq data showed that QC6352 induced a differentiation gene signature in three MNA cell lines (BE2C, KELLY, and SIMA) similar to that of BE2C cells treated with all- trans -retinoic acid 67 ( Figure 6 F), but to a much less degree in the two non-MNA cell lines (SK-N-AS and SK-N-SH) ( Figure 6 G). Interestingly, the retinol metabolism pathway was significantly induced by QC6352 ( Figure S5 B), indicating that KDM4 inhibition may induce an intrinsic signal resembling the exogenous retinoic acid treatment. KDM4 inhibition also induced significant upregulation of apoptosis gene signatures in MNA cell lines in comparison with the non-MNA cell lines ( Figures 6 H and 6I). The apoptosis gene expression signatures correlated with markers of apoptosis, induction of cleaved PARP by immunoblot ( Figure 6 J), and annexin V by flow cytometry analysis ( Figures 6 K and S5 C). We also found that QC6352 greatly reduced the expression of SLC7A11 in MNA cells in comparison with non-MNA cells ( Figure S5 D). As MYCN-driven cysteine addiction depends on the system Xc-cystine/glutamate antiporter (SLC7A11) for ROS detoxification, 68 , 69 the greater reduction of SLC7A11 in MNA cells could further contribute to the anticancer effect of QC6352. These data may explain why MNA cells are more sensitive to KDM4 inhibition, although induction of the interferon pathway by QC6352 may further increase the anticancer effect as our previous studies have demonstrated that administration of type I interferon β inhibits neuroblastoma cell proliferation and tumor growth. 70 , 71 , 72 , 73 A recent study indicates that inhibition of KDM4 leads to replication stress. 49 The MNA cells thus might be more sensitive to QC6352 due to a higher replication stress. We therefore performed DNA fiber assay to assess this possibility. However, we did not observe remarkable differences between BE2C and KELLY (MNA) vs. SK-N-AS (non-MYCN) cells, although all cells showed similar reduction in rates of DNA replication in response to QC6352 treatment ( Figure S5 E) Figure 6 KDM4 inhibition induces differentiation and apoptosis in MYCN- amplified cells (A) QC6352 induces neurite outgrowth. BE2C and SIMA cells treated with 200 nM of QC6352 for 5 days. Photos taken under a light microscope (10×). (B) Hematoxylin and eosin (H&E) staining shows that QC6352 induces neuroblastoma differentiation in SIMA xenografts as indicated by the large amounts of pink neuropil (blue arrow area). Scale bars, 100 μm. (C and D) Immunohistochemical staining of cleaved caspase-3 in SIMA xenografts treated with QC6352. Brown, cleaved caspase-3; blue, nuclei. The bottom panel showing quantification of cell numbers with positive cleaved caspase-3 per view (vehicle n = 13, QC6352 n = 16).Data are mean ± SEM. Student’s t test. Scale bars, 200 μm. (E) GSEA of RNA-seq data from SIMA xenografts treated with QC6352 using ADRN gene signature. (F–I) GSEA of RNA-seq data from indicated cell lines treated with QC6352 showing differentiation signature and apoptosis signaling pathways. (J) Neuroblastoma cells were treated with DMSO and 100 nM of QC6352 for 3 (KELLY) or 5 (SIMA, BE2C, SK-N-SH, SK-N-FI, and SK-N-AS) days and probed with the indicated antibodies. (K) Comparison of apoptosis induced by QC6352 in MNA and non-MNA cells. The graph shows annexin V-FITC- and PI-negative cells (live cells), annexin V-FITC-positive cells (early apoptosis), annexin V-FITC- and PI-positive cells (late apoptotic cells) quantified by fluorescence-activated cell sorting analysis. Data are mean ± SEM. p value calculated by Student’s t test. MYCN -amplified (BE2C, KELLY, and SIMA) and non- MYCN -amplified cells (SK-N-AS, SK-N-SH, and SK-N-FI) were treated with DMSO or 100 nM QC6352 for 72 h.
QC6352 in combination with vincristine/irinotecan leads to a complete response in MYCN -amplified tumors
A significant challenge in the clinical management of neuroblastoma with MNA is to achieve durable responses and overcome chemotherapy resistance. 74 We expect that KDM4B inhibition will improve the effect of chemotherapy by disrupting MYCN transcription and blocking adrenergic CRC TF activity in maintaining active gene transcription. KDM4 inhibition may also hinder DNA repair, since KDM4 has been shown to be involved in DNA damage response. 75 , 76 GSEA demonstrated that QC6352 treatment repressed expression of the genes involved in mitotic phase cell cycle checkpoint and homologous DNA repair ( Figure S6 ). We therefore hypothesized that KDM4 inhibition would worsen the cytotoxic effects of drugs that disrupt mitosis such as vincristine that prevents microtubule polymerization, causing mitotic failure, or those that cause DNA double-strand breaks such as irinotecan that forms a complex with DNA-topoisomerase I and thus interferes with the moving replication fork. Indeed, adding QC6352 significantly improved the efficacy of vincristine/irinotecan (VCR/IRN), a drug combination used in relapsed neuroblastoma treatment, in two MNA (BE2C and KELLY) and one non-MNA (SK-N-AS) xenograft models ( Figure 7 ). In the BE2C model, all mice had a complete response (CR) to the combination therapy, while nearly all mice treated with chemotherapy, QC6352 alone, or vehicle control had progressive disease ( Figures 7 A and 7B; Table S4 ). Notably, the BE2C cell line was derived from a bone marrow biopsy taken from a child with disseminated neuroblastoma after repeated courses of chemotherapy and radiotherapy; therefore, it is an aggressive model that is resistant to most current therapies. Chemotherapy-QC6352 combination therapy also significantly extended time to relapse in BE2C tumor models without overt body weight loss ( Figures 7 C and 7D; Table S5 ). In the KELLY model, combination therapy also showed uniform complete response, extended mouse survival without body weight loss, although chemotherapy alone showed 22% complete response and 33% stable disease ( Figures 7 E–7H). SK-N-AS is a MES, non-MNA cell line, that showed resistance to VCR/IRN chemotherapy ( Figure 7 I). However, despite this resistance to VCR/IRN alone, combination therapy with QC6352 achieved 14% complete response and 28% stable disease ( Figure 7 J) and significantly extended time to relapse ( Figure 7 K), without changes in mouse body weight ( Figure 7 L). Taken together, QC6352 in combination with conventional chemotherapeutic agents used as standard of care in relapsed disease greatly improved their efficacy in preclinical neuroblastoma models. Figure 7 QC6352 in combination with vincristine/irinotecan leads to 100% complete response in MYCN -amplified tumors (A) Tumor growth curve of BE2C xenografts in NSG mice treated with vehicle (n = 7), QC6352 (n = 10), VCR/IRN (n = 9), or QC6352/VCR/IRN (n = 10) for 3 weeks. Data are mean ± SEM. Wilcoxon rank-sum test (two-sided) was used to compare outcome of the two groups at each time point. p values were adjusted using the Benjamini-Hochberg method. (B) Waterfall plot of response to treatment of BE2C xenografts. The vehicle control group was not included as mice had to be euthanized early due to rapid tumor progression. (C) Kaplan-Meier survival curves of BE2C xenograft models. Log rank (Mantel-Cox) method was used for p value calculation. (D) Mouse body weight of BE2C models during treatment course in three models. Data are mean ± SEM. (E–H) Tumor growth curve (E), waterfall plot of response (F), Kaplan-Meier survival curves (G), and mouse body weight (H) of KELLY xenografts in NSG mice treated with vehicle (n = 8), QC6352 (n = 9), VCR/IRN (n = 9), or QC6352/VCR/IRN (n = 10) for 3 weeks. Data are mean ± SEM. (I–L) SK-N-AS xenografts treated with vehicle (n = 5), QC6352 (n = 5), VCR/IRN (n = 6), or QC6352/VCR/IRN (n = 7) for 3 weeks, as indicated in (E–H). Data are mean ± SEM.
Discussion
MYCN and adrenergic CRC TFs form a forward-loop circuitry to drive each other’s expression. 11 , 14 In this study, we demonstrate that KDM4 inhibition was selectively effective against MYCN -amplified neuroblastoma by epigenetically repressing the transcription of adrenergic CRC TFs, resulting in the altering of histone modifications at their genomic loci. The engagement of KDM4 in this process suggests that this subfamily of histone demethylases plays an important role in maintaining the cell identity of MNA neuroblastoma. Here, we establish that QC6352, the best-in-class KDM4 inhibitor with high KDM4 selectivity and oral bioavailability, 58 epigenetically represses the transcription of MYCN and genes encoding adrenergic CRC TFs, and exhibits selective and potent antiproliferative activity as a single agent and in a combination regimen with chemotherapy in MYCN -amplified disease models. The mechanism of action of QC6352 against MNA cells likely lies in its KDM4 inhibitory activity, which results in broad induction of the heterochromatin mark H3K9me3 and subsequent reprogramming of the epigenetic landscape of neuroblastoma cells, leading to downregulation of chromatin accessibility at the genomic loci of MYCN and adrenergic CRC TFs and their downstream target genes. However, the histone demethylase-independent activity cannot be excluded. Despite lower potency against the non-MNA neuroblastoma cells (SK-N-SH and SK-N-AS), QC6352 also represses the expression of adrenergic CRC TFs and c-MYC in these models. Since epigenetic modifiers facilitate the activities of master transcriptional factors, our rationale is that MYCN “hijacks” KDM4 in MNA cells. However, we cannot exclude other possibilities in non-MNA cells that may “hijack” KDM4 through other transcriptional factors.
Limitations of the study
The mechanism accounting for the discrepancy of mRNA and protein expression for some ADRN/MES genes and the KDM4 members that were induced by MYCN was not investigated. A follow-up study to understand how MYCN regulates protein translation of KDM4 and ADRN CRC TFs may reveal a new mechanism of MYCN in neuroblastoma and provide additional therapeutic opportunities. We used a ChIPseqSpikeInFree computational algorithm to investigate the changes in histone marks, which may cause potential bias to understand the mechanism of KDM4 inhibition without an experimental spike-in normalization. In the future, we will further validate the mechanism by using ChIP-Rx to extract discernible differences with or without KDM4. We only treated the animals for 3 weeks in our combination therapy study. Whether a sustained complete response can be achieved by a longer-term treatment awaits further studies. However, it is possible that a small set of cells may escape or be resistant to longer-term treatment with the KDM4 inhibitor.
STAR★Methods
Key resources table
Resource availability
Lead contact
Further information and requests for resources and reagents should be directed to the lead contact, Jun Yang ( Jun.Yang2@stjude.org ).
Materials availability
All plasmids and cell lines generated in this study are available from the lead contact Dr. Jun Yang with a completed material transfer agreement.
Data and code availability
The RNA-seq data, ATAC-seq data, ChIP-seq data, CUT&Tag data have been deposited in GEO database. Accession numbers are listed in the key resources table . This paper does not report original code. Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.
Experimental models and study participant details
Cell culture
IMR32, LAN5, NGP, SKNDZ, KELLY, SIMA, SK-N-BE2, BE2C, NB1691, SK-N-AS, SK-N-SH, CHP212, SK-N-FI, LAN6 and SHEP-21N were cultured in 1X Roswell Park Memorial Institute (RPMI) 1640 (Corning, 15-040-CV) supplemented with 10% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801), and 1% Penicillin/Streptomycin (Gibco, 15140122). COGN496, CHLA20, and CHLA42 were cultured in 1X Iscove’s modified Dulbecco’s medium (Gibco, 12440053) supplemented with 20% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801), 1% Penicillin/Streptomycin (Gibco, 15140122), and 1X Insulin-Transferrin-Selenium (Gibco, 41400045). HS68 and BJ cells were cultured in high-glucose Dulbecco’s Modified Eagle Medium (15-013-CV, Corning), supplemented with 100 U/mL penicillin/streptomycin (15140122, Gibco), 2 mmol/L L-glutamine (A2916801, Gibco) and 10% fetal bovine serum (F2442, Sigma-Aldrich). O9-1 cells were cultured in complete mouse embryonic stem (ES) media with 15% fetal bovine serum (FBS) and mouse leukemia inhibitory factor (LIF) (Millipore, ES-101B). All cells were maintained at 37°C in an atmosphere of 5% CO 2 . All human-derived cell lines were validated by short tandem repeat (STR) profiling using PowerPlex 16 HS System (Promega) once a month. Once a month a polymerase chain reaction (PCR)-based method was used to screen for mycoplasma employing the LookOut Mycoplasma PCR Detection Kit (MP0035, Sigma-Aldrich) and JumpStart Taq DNA Polymerase (D9307, Sigma-Aldrich) to ensure cells were free of mycoplasma contamination.
Generation of QC6352-resistant cells
The generation of QC6352-resistant BE2C cells was established by exposing the parental BE2C cells to QC6352 in sequential, increasing concentrations of QC6352 (0.1, 0.2, 0.5, 1, 2, and 2.5 μM; each concentration was incubated for 5–7 days). The final concentration of QC6352 was 2.5 μM.
Mouse transgenic and xenograft models
The St. Jude Children’s Research Hospital Institutional Animal Care and Use Committee approved all studies performed. Subcutaneous xenografts were established in CB17 severe combined immunodeficient mice (CB17 scid, Taconic) or NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NOD scid gamma (NSG, JAX) mice by implanting 5 x10 6 cells in Matrigel. Transgenic TH-Mycn/Alk F1178L mouse model has been published. 84 , 85 We used both genders of CB17 SCID or NSG or transgenic TH-MYCN/ALK F1178L mice to exclude sex bias. Tumor xenograft measurements were done weekly using electronic calipers, and volumes calculated as width π/6 × d3 where d is the mean of two diameters taken at right angles. Ultrasound imaging was applied to monitor tumor growth in TH-Mycn / Alk F1178L mice. Technicians in the St. Jude Center for In Vivo Imaging and Therapeutics performed ultrasound scanning on mice weekly using a VEVO-2100, and determined tumor volumes using VevoLAB 3.0.0 software. All ultrasound data were acquired in a blinded fashion.
Method details
Generation of CRISPR-Cas9-mediated knockout cells
The sgRNA oligos (KDM4A: Forward 5′-CACCGGACTTCATAGAAGGTCTCGG-3’; Reverse 5′-AAACCCGAGACCTTCTATGAAGTCC-3’; KDM4B: Forward 5′-CACCGGCTCCCTGCGACTCTATGT-3’; Reverse 5′-AAACACATAGAGTCGCAGGGAGCC-3’; KDM4C: Forward 5′-CACCGCCCGGTGCAAAAGAAATGCG-3’; Reverse 5′- AAACCGCATTTCTTTTGCACCGGGC-3′) were resuspended in nuclease-free water to a final concentration of 100 μM. Next, in a 1.5 mL microcentrifuge tube, the sgRNA oligos were mixed in a 1:1 ratio and placed on 95°C heating block and left for 5 min. Then, the heating block was turned off and the 1.5 mL microcentrifuge tube was left to cool down to room temperature while it was still placed on the heating block for nearly 50 min. In a 1.5 mL microcentrifuge tube, the following reagents were mixed: 100 ng of pPB-US-ECasE (Addgene, 83961), 1 μL of annealed KDM4 sgRNA oligos, 1 μL of BbsI (New England BioLabs, R0539S), 2 μL of NEBuffer 2.1 (New England BioLabs, catalog#B7202), 1 μL of T7 DNA Ligase (New England BioLabs, M0318S), 1 μL of 10 mM ATP (New England BioLabs, P0756S), and 0–14 μL of nuclease-free water to make a final volume of 20 μL. The 1.5 mL microcentrifuge tube was heated for 37°C on a heating block for 2 h. Bacterial transformation was done using One Shot Stbl3 Chemically Competent E. coli (Invitrogen, C7373-03) as per the manufacture’s guidelines. Next day, 10 single clones were picked up and subjected to Sanger sequencing to validate successful cloning of the KDM4 sgRNA oligos into pPB-US-ECasE plasmid, using the primers hU6-F (5′-GAGGGCCTATTTCCCATGATT-3′) and LKO.1 5’ (5′-GACTATCATATGCTTACCGT-3′). The verified plasmids were maxipreped using QIAGEN’s kit according to instruction. On day 1, 1 x 10 5 cells were seeded in a single well of a 6-well plate. On day 2, stable transient transfection was done by mixing Tube 1 with Tube 2. Tube 1 was prepared by mixing 125 μL of Opti-MEM (ThermoFisher Scientific, catalog#31985062) with 0.5 μg of the cloned plasmid and 0.2 μg of Super PiggyBac Transposase Expression Vector (System Biosciences, PB210PA-1). Tube 2 was prepared by mixing 125 μL of Opti-MEM (ThermoFisher Scientific, 31985062) with 1.4 μL of PEIpro transfection reagent (Polyplus, 101000033) in a 1:2 ratio (plasmids:PEIpro reagent). Next, Tube 2 was added to Tube 1, mixed, and allowed to stand at room temperature for 15 min. Next, the total mixture (250 μL) was added in a drop-wise fashion to the cells. Please note the addition of Super PiggyBac Transposase Expression Vector allows for stable transfection. On day 3, the transfection medium was discarded, and fresh medium was added. On day 4 and day 5, the cells were observed. On day 5, the cells were selected with puromycin (2 μg/mL) until the non-transduced cells were completely dead. Then, the transfected cells were allowed to proliferate over time. Then, the transfected cells were treated with either ethanol (ETOH, control cells) or 500 nM of 4-Hydroxytamoxifen (4OHT, knockout cells) for 5 days. ETOH- and 4OHT-treated cells were FACS-sorted into single cells in a 96-well plate and allowed to proliferate over time. Ten growing clones were transferred to 24-well plates and then 12-well plates for further cellular proliferation. Next, the clones were validated for knockout efficiency using western blot, and the 100% KDM4A/KDM4B/KDM4C KO clones were used further in the experiments.
Compound resource and QC6352 synthesis
Vincristine (HY-N0488) and Irinotecan (HY-16562A) were purchased from MedChem Express (MCE). QC6352 was synthesized using the reported method 36 , 58 with minor modifications. Briefly, the cross-coupling of [(1R)-6-Bromo-1,2,3,4-tetrahydronaphthalen-1-yl] methanamine (compound 1) with 3-fluoro-isonicotinonitrile produced intermediate 2. The Buchwald−Hartwig coupling of intermediate 2 with N-methyl-aniline followed by alkaline hydrolysis gave the final product QC6352 (compound 4, with a 37% efficiency over 3 steps). Reagents and conditions: (a) 3-fluoro-isonicotinonitrile, N, N-Diisopropylethylamine, DMF, 120°C, 3 h, 55%; (b) N-methyl-aniline, RuPhos Pd G3, Cs 2 CO 3 , toluene, microwave, 100°C, 3 h, 76%; (c) NaOH, MeOH, 88%. The compound was purified by gradient flash column chromatography (Biotage Isolera Four, Sweden), eluted with 0–100% EtOAc in hexane.
SDS-PAGE and western blot
Cells were washed twice with ice-cold phosphate-buffered saline (PBS) and directly lysed on ice with 2X sample loading buffer (0.1 M Tris HCl [pH 6.8], 200 mM dithiothreitol [DTT], 0.01% bromophenol blue, 4% sodium dodecyl sulfate [SDS] and 20% glycerol). On ice, cell lysates were sonicated once with a 5 s bursts at 40% amplitude output (Sonics, VIBRA CELL) followed by 25 min heating at 95°C. After the cell lysates were briefly centrifuged at 13,000 × g at room temperature for 2 mins, 10–20 μL of the cell lysates were separated on 4–15% Mini-PROTEAN TGX Stain-Free Protein Gels from Bio-Rad and transferred to methanol-soaked polyvinylidene difluoride (PVDF) membranes (Millipore). Membranes were blocked in PBS buffer supplemented with 0.1% TWEEN 20 and 5% skim milk (PBS-T) and incubated for 1 h at room temperature under gentle horizontal shaking. Membranes were incubated overnight at 4°C with the primary antibodies under gentle horizontal shaking. The next day, membranes were washed 3 times (for 5 min) with PBS-T at room temperature under gentle horizontal shaking. Protected from light, membranes were then incubated with goat anti-mouse or goat anti-rabbit HRP-conjugated secondary antibodies (1:5,000) for 1 h at room temperature under gentle horizontal shaking, followed by three 5-minite washes with PBS-T at room temperature. Lastly, membranes were incubated for 1 min at room temperature with SuperSignal West Pico PLUS Chemiluminescent Substrate (34580, Thermo Fisher Scientific) and the bound antigen-antibody complexes were visualized using Odyssey Fc Imaging System (LI-COR Corp., Lincoln, NE).
Crystal violet staining
After removing media, cells were washed with Dulbecco’s phosphate buffered saline without calcium or magnesium (DPBS, Lonza) and treated with 4% Formaldehyde in PBS (PFA) for 20 min. Once PFA was removed, cells were stained with 0.1% crystal violet stain for 1 h.
Prestoblue assay
Cell viability was determined by PrestoBlue Cell Viability Reagent from Thermo Fischer Scientific according to the manufacturer’s instructions. PrestoBlue reagent is a cell-permeable and resazurin-based solution that serves as a cell viability indicator. It utilizes the mitochondrial reducing power of living cells to convert resazurin (blue) to resorufin (red). This colorimetric change can then be detected using absorbance or fluorescence measurements. In brief, cells were seeded into 96-well plates (CulturPlate-96 from PerkinElmer) at an initial density of 3,000 cells per well and incubated at 37°C with 5% CO 2 for 24 h. Cells were treated with different final concentrations of QC6352 (10000, 500, 250, 50, 25, 12.5, 5, 2.5, 0.5, 0.05 and 0 nM) for 5 days. The QC6352 stock (10 mM) was prepared in DMSO, and all working dilutions were prepared in culture media. Each concentration was tested in 8 replicates. The total volume per well was 100 μL. After completion of the treatment period, 10% of PrestoBlue reagent (10 μL) was added to each well and plates were incubated at 37°C with 5% CO 2 for 30 min. Then, fluorescence (560 nm excitation/590 nm emission) was measured using a BioTek Synergy H1 microplate reader. Calculations of GI 50 and LC 50 viability values were done through a pre-defined Microsoft Excel spreadsheet (Supplementary File). Viability values were calculated as a percentage of live cells with respect to the control sample (0 nM QC6352). The negative control was culture medium without cells.
Flow cytometry analysis of apoptosis by annexin V staining
Cells were seeded at a density of 100, 000 cells in each well in 6 well plates. Next day, cells were treated with DMSO or QC6352 for 72 h. Cells were trypsinized with 0.05% TPVG solution for 4 min and centrifuged with benchtop centrifuge at 1000 rpm for 5 min. Apoptosis was detected by dual staining of Annexin V and PI using apoptosis assay kit (TONBO biosciences, CA, USA) according to manufacturer’s instructions. Annexin V-FITC positive cells were collected using log amplification, and 10,000 events were recorded, and data was analyzed using BD FACSDiva Software.
DNA fiber assay
The cells were treated with QC6352 (200nM) and DMSO for 5 days, were labeled with Cldu (20μM) for 20 min followed by Idu (200μM) for 20 min. After labeling, cells were centrifuged and resuspended in ice-cold PBS. Nearly 2500 cells (2.5μL) were spotted onto glass slides, cells were air dried for 5 min and were lysed with lysis buffer (7.5μL) for 8–10 min. Subsequently, slides are tilted to allow the DNA fibers to spread. The slides were then fixed with methanol and acetic acid (3:1) for 15 min followed by denaturation using 2.5 M HCL for 60–80 min. After blocking with 5% BSA, the slides were incubated with anti-BrdU (1:40; mouse, BD Biosciences; 347580) and anti-BrdU (1:80; Rat, Abcam; AB6326) for overnight at 4°C. After two washes with PBS, the slides were incubated with secondary antibodies (Cy2 and Cy3; 1:500; Jackson ImmunoResearch). DNA fibers were visualized using a con focal microscope with 63x magnification and fiber length were measured using ImageJ software. 100–300 DNA fibers were used to analyze the data.
Immunohistochemistry
Xenografts were fixed in 10% neutral buffered formalin, embedded in paraffin, sectioned at 4 μM, stained with hematoxylin and eosin and reviewed by light microscopy using an upright Nikon Eclipse Ni microscope (Nikon Instruments, Inc.). was performed on 4 μm thick tissue sections mounted on positively charged glass slides (Superfrost Plus; Thermo Fisher Scientific, Waltham, MA), and dried at 60°C for 20 min. The procedures for immunohistochemistry were performed using a Ventana DISCOVERY ULTRA autostainer (Roche). Analysis of cleaved caspase 3 IHC . Immunohistochemistry for cleaved caspase 3 (BioCare medical, CP229C, 1:500) was performed using a Ventana DISCOVERY ULTRA automated stainer (Ventana Medical Systems, Inc., Tucson, AZ). Heat-induced epitope retrieval (HIER) was carried out for 32 min at 37°C using Cell conditioning media 1 (CC1, 950-124, Ventana Medical Systems, Inc, Tucson, AZ) followed by incubation with the primary antibody for 60 min and then visualization with OmniMap anti-rabbit HRP (760–4311, Ventana Medical Systems, Inc, Tucson, AZ) and the DISCOVERY ChromoMap DAB kit (760-159, Ventana Medical Systems, Inc, Tucson, AZ). Positive cells in different views under microscope were counted as individuals for unpaired Student’s t test.
siRNA transfection
was performed using Lipofectamine RNAiMAX Transfection Reagent (13778-075, Thermo Fischer Scientific) according to the manufacturer’s instructions using the reverse transfection method. In brief, final concentration of 5 nM–10 nM of siRNA was mixed with 500 OptiMEM and 7 μL of RNAiMAX Transfection Reagent and left at room temperature for 10 min. Then, the mixture was added to well in a drop-wise fashion. siRNA oligos targeting KDM4A-C were purchased from Thermo Fisher Scientific: siKDM4A (cat# s18636), siKDM4B (cat#s22868 and s22867), siKDM4C (cat#s22990), siMYCN (ca#s9135). Non-targeting control#2 siRNA (Dharmacon) was used as a control.
RNA-seq
Total RNA from cells and tumor tissues were performed using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Paired-end sequencing was performed using the High-Seq platform with 100bp read length. Reads were aligned to the human GRCh37-lite using SJCRH’s Strongarm pipeline. Counts per gene were obtained using htseq-count version 0.6.1 with Gencode vM5 level 1and 2 gene annotations. Counts were normalized with VOOM and analyzed with LIMMA within the R statistical environment. Significance was defined as having a false discovery rate (FDR) < 0.05. VOOM normalized counts were analyzed with Gene Set Enrichment Analysis (GSEA).
ADRN-MES scoring system
Relative adrenergic (ADRN) and mesenchymal (MES) scoring was performed with modifications as in Van Groningen et al. 8 Briefly, a gene set signature corresponding to adrenergic or mesenchymal genes was extracted from Van Groningen et al. 8 RNA-seq data corresponding to these signature genes was extracted from BE2C, SIMA KELLY, SK-N-SH or SK-N-AS cells treated with DMSO or QC6352 compound. Genes were inverse rank ordered, with median ADRN or MES scores calculated individually. Untreated cell line RNA-seq data was extracted from the Cancer Cell Line Encyclopedia, available at www.depmap.org . Median values were plotted using Graphpad Prism 9.0.
Chromatin immunoprecipitation sequencing (ChIP-seq)
1 x 10 7 of BE2C cells treated with DMSO control or 200 nM of QC6352 for 48 h were cross-linked with freshly prepared 1% formaldehyde (final concentration) for 10 min and quenched with 125 mM glycine (final concentration) for 5 min at room temperature. Cells were washed once with Dulbecco’s Phosphate-Buffered Saline (DPBS). 5–8 mL of ice-cold cell lysis buffer (5 mM PIPES pH 8.0, 85 mM KCl and 0.5% NP-40 supplemented with 1 tablet protease inhibitor cocktail) were added to cells. The cells were recovered with a cell scraper and transferred to 50-mL conical tubes on ice. Cell pellets were collected by centrifugation at 2,000 rpm for 5 min at 4°C. The pellets were resuspended in cell lysis buffer and incubated on ice for 10 min. Cells were passed 20 times through a 20-gauge needle. The nuclei were collected by centrifugation at 2,000 rpm at 4°C for 5 min. The nuclear pellets were resuspended and lysed with 250 μL RIPA buffer (1X PBS, 1% NP-40, 0.5% sodium deoxycholate and 0.1% SDS with 1 tablet protease inhibitor cocktail). The lysates were sonicated for at least 30 min (30 s on/30 s off) using Bioruptor Pico (Diagenode, Inc., Denville, NJ) and centrifuged at 14,000 rpm in a microfuge for 15 min at 4°C. The supernatant was collected, and the sonicated DNA fragments were examined by electrophoresis on a 1% agarose gel. The sonicated samples with enrichment of fragments between 100 and 500 bp range were used for the chromatin immunoprecipitation (ChIP) after removing 50 μL from each sample as input. The primary antibodies H3K36me3 (RevMab, 31-1051-00), H3K4me1 (Abcam, ab8895) and H3K27ac (Abcam, ab4729) were coupled with 25 μL magnetic beads (Dynabeads M-280 Sheep Anti-Rabbit IgG, Invitrogen 11203D) while H3K9me3 (Absolute, 309M3-B) antibodies were coupled with 100μL of Dynabeads 280 streptavidin overnight in the cold room at 4°C, washed three times with PBS/BSA (1X PBS/5 mg/mL BSA (fraction V)) at 4°C, then incubated with sonicated DNA chromatin samples overnight in the cold room at 4°C. The beads containing immuno-bound chromatin were collected by placing the microfuge tube on a magnet rack. The beads were extensively washed with LiCl washing buffer (100 mM Tris pH 7.5/500 mM LiCl/1% NP-40/1% sodium deoxycholate) 5 times and TE buffer (10 mM Tris-HCl pH 7.5 and 0.1 mM Na2EDTA) once. Bound chromatin was recovered by IP elution buffer (1% SDS and 0.1 M NaHCO3) and reverse-crosslinked at 65°C overnight. DNAs were purified using the Mini-Elute PCR Purification Kit (Qiagen, Valencia, CA) after treatment with RNase A and proteinase K. ChIP enriched DNAs were submitted for library preparation and sequencing. ChIP-seq library preparation and sequencing were carried out by the Hartwell Center at St Jude Children’s Research Hospital, Memphis, TN, USA. Briefly, 5–10 ng of DNA was used to prepare libraries using the NEBNext ChIP-Seq Library Prep Reagent Set for Illumina with NEBNext Q5 Hot Start HiFi PCR Master Mix according to the manufacturer’s instructions (New England Biolabs). Completed ChIP-seq libraries were analyzed for insert size distribution using a 2100 BioAnalyzer High Sensitivity kit (Agilent) or Caliper LabChip GX DNA High Sensitivity Reagent Kit (PerkinElmer). All libraries were quantified using the Quant-iT PicoGreen dsDNA assay (Life Technologies) and Kapa Library Quantification kit (Kapa Biosystems). Fifty-cycle single-end sequencing was performed using an Illumlina HiSeq 2500 or HiSeq 4000.
CUT&Tag
for KDM4A, KDM4B, KDM4C and H3K27me3 in BE2C were prepared by following the protocol as described previously 86 (Kaya-Okur et al. 2019 and https://www.protocols.io/view/bench-top-cut-amp-tag-bcuhiwt6?step=1 ) with minor modifications. Approximately 500,000 BE2C cells were washed with wash buffer (20 mM HEPES pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1× Protease inhibitor cocktail). Nuclei were isolated with cold NE1 buffer (20 mM HEPES–KOH, pH 7.9; 10 mM KCl 0.1%; Triton X-100; 20% Glycerol, 0.5 mM Spermidine; 1x Protease Inhibitor) for 10 min on ice. Nuclei were collected by 600 x g centrifuge and resuspended in 1mL washing buffer containing with 10 μL of activated concanavalin A-coated beads (Bangs laboratories, BP531) at RT for 10 min. Bead-bound nuclei were collected by placing tube on magnet stand and removing clear liquid. The nuclei bound with beads were resuspended in 50 μL Dig-150 buffer (20 mM HEPES pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1× Protease inhibitor cocktail; 0.05% Digitonin; 2 mM EDTA) and incubated with a 1:50 dilution of antibody against KDM4A (Sigma, HPA007610), KDM4B (Active motif, 61221) and KDM4C (Sigma, HPA069357), and H3K27me3 (1:100)(CST, #9733) overnight at 4°C. The unbound primary antibodies were removed by placing the tube on the magnet stand and withdrawing the liquid. The primary antibody bound nuclei beads were mixed with Dig-150 buffer 100uL containing guinea pig anti-Rabbit IgG antibody (Antibodies, ABIN101961) in a 1:100 dilution for 1 h at RT. Beads bound nuclei were washed using the magnet stand 3× for 5 min in 1 mL Dig-150 buffer to remove unbound antibodies. A 1:250 dilution of pA-Tn5 adapter complex was prepared in Dig-300 buffer (20 mM HEPES, pH 7.5, 300 mM NaCl, 0.5 mM Spermidine, 0.05% Digitonin, 1× Protease inhibitor cocktail). After removing the liquid on the magnet stand, 100 μL mixture of pA-Tn5 and Dig- 300 buffer were added to the nuclei bound beads with gentle vortex and incubated at RT for 1 h. After 3 × 5 min in 1 mL Dig-300 buffer to remove unbound pA-Tn5 protein, nuclei were resuspended in 250 μL Tagmentation buffer (10 mM MgCl 2 in Dig-300 buffer) and incubated at 37°C for 1 h. 10 μL of 0.5 M EDTA, 3 μL of 10% SDS and 2.5 μL of 20 mg/mL Proteinase K were added to stop tagmentation by incubating at 55°C for 1 h. DNA was then precipitated by phenol/chloroform/isoamylalcohol followed by ethanol precipitation with glycogen and then dissolved in water. Sequencing libraries were prepared using NEBNext HiFi 2× PCR Master Mix (NEB, M0541L) according to the manufacturer’s instructions. The PCR products were cleaned up with SPRIselect beads and quantified using Qubit dsDNA HS assay kit (Agilent Technologies). The libraries were sequenced on a HiSeq2500 with paired-end 50-bp reads (Illumina).
ChIP-seq and CUT& tag data processing
Mapping reads and peak calling. The ChIP-seq raw reads were aligned to the human reference genome (hg19) using BWA (version 0.7.12; BWA aln+samse for ChIP-seq and BWA aln+sampe for CUT&Tag data). Duplicate reads were marked and removed by Picard (version 1.65). For CUT&Tag, only properly paired uniquely mapped reads were extracted by samtools (version 1.3.1 parameters used were -q 1 -f 2 -F 1804) for calling peaks and generating bigwig file. Narrow peaks were called by MACS2 (version 2.2.7.1) with parameters of “ -t chip_file -c input_file -q 0.05 -f BED --nomodel --extsize fragment_size --keep-dup all” ChIP-seq (where fragment_size was estimated by SPP version1.11) or “ -t cut_tag_file -q 0.05 -f BED --keep-dup all” for CUT&Tag data). For ChIP-seq broad peaks (H3K9me3, H3K36me3, H3K27me3, and H3K4me1), we used SICER (version 1.1, with parameters of redundancy threshold 1, window size 200bp, effective genome fraction 0.86, gap size 600bp, FDR 0.00001 with fragment size defined above).
Peak overlap and annotation
Peak regions were defined to be the union of peak intervals in replicates from control or treated cells respectively. For peak overlap analysis, mergeBed (BEDtools version 2.25.0) was used to combine overlapping regions from multiple peak sets into a new region and then a custom script was used to summarize common or distinct peaks and visualize in a Venn diagram. Promoter regions was defined as the regions 1.0 kb upstream and 1.0 kb downstream of the transcription start sites based on the human RefSeq annotation (hg19). Genomic feature annotation of peaks was done by annotatePeaks.pl, a program from the HOMER suite (v4.8.3, http://homer.salk.edu/homer/ ).
Visualization
We used genomeCoverageBed (BEDtools 2.25.0) to produce genome-wide coverage in BEDGRAPH file and then converted it to a bigwig file by bedGraphToBigWig. The bigwig files were scaled to 15 million reads to allow comparison across samples. To show average of several replicates as a single track in the browser, the bigwig files were merged to a single average bigwig file using UCSC tools bigWigtoBedGraph, bigWigMerge and bedGraphToBigWig. The Integrated Genomics Viewer (IGV 2.3.82) was used for visual exploration of data.
Motif analysis
The HOMER software was used to perform de novo motif discovery as well as a check the enrichment of known motifs in a set of given peaks. Motif density histograms were created using HOMER for target regions. Background regions were generated by selecting DNA sequences of equal length at 10 kb downstream of the target regions. The motif density at target regions was normalized to that at the control regions.
Differential analysis
ChIP-seq raw read counts were reported for each region/each sample using BEDtools 2.25.0. Raw read counts were Voom normalized and statistically contrasted using the R (version 3.5.1) packages limma and edgeR (version 3.16.5) for CPM calculation and differential analysis. An empirical Bayes fit was applied to contrast treated samples to control samples and to generate log fold changes, p values and false discovery rates for each peak region.
ChIPseqSpikeInFree normalization for H3K9me3 and H3K36me3
ChIPseqSpikeInFree(version 1.2.3) 82 was used to calculate scaling factor (S) for every sample (S i ). The effective ChIP-seq library size for sample i was then calculated as N i ∗ S i , where N i is the original library size. The effective library size was then used to normalize the read count from sample i during the downstream differential analysis and heatmap visualization.
Heatmap generation for chromatin co-localization of KDM4 and ADRN CRC TFs
The coverage areas of the MES or ADRN gene promoters (TSS +/− 2000bp) or super enhancers (center +/− 5000 bp) on the chromosome were displayed and visualized. The “computeMatrix” and “plotHeatmap” functions in the deepTools package 87 were used to generate a heatmap displaying the in-house ChIPseq dataset (H3K27Ac, KDM4A, KDM4B and KDM4C) and ATAC-seq data and public GSE94824 ChIP-seq dataset(MYCN, HAND2 ISL1, TBX2 and GATA3) in BE2C cells. Each row in the heatmap is a genomic region, and gradient colors represent different amounts of enriched reads. Super enhancers were called by ROSE 88 based in-house H3K27Ac ChIP-seq data in BE2C cells using default parameters. A note for Figure 2 B: The overlap analysis (shared binding sites) was done by two steps: (1) First merge nearby original peaks (from all KDM4A, B,C respectively) within distance <100 bp into one merged peakset. (2) “Shared peaks” are defined as those having at least 1 base overlap between two merged peaksets. This method works very well when peak sets have distinct genomic distributions like H3K9me3,H3K27Ac because they have minimal number of nearby peaks. However, the results from this method could be confusing if there are many nearby (within heatmap view window size) but not-physically overlapped peaks. In our case, KDM4A and KDM4B shared ∼40% peaks but many other “4A- or 4B- specific” binding sites likely localize nearby. Because the heatmap was generated using regions of peak center +/− 2kb, these nearby but non-overlapping peaks within 4kb windows will show signal in heatmap. Sometimes the long tails with relatively weaker signals in heatmap could be exclusively present in one set but not another.
Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) and analysis
Library preparations for ATAC-seq were based on a published protocol with minor modifications. 52 , 89 Briefly, cells (100,000 per sample) were harvested and washed with 150 μL ice-cold Dulbecco’s Phosphate-Buffered Saline (DPBS) containing protease inhibitor (PI). Nuclei were collected by centrifugation at 500 g for 10 min at 4°C after cell pellets were resuspended in lysis buffer (10 mM Tris-Cl pH 7.4, 10 mM NaCl, and 3 mM MgCl2 containing 0.1% NP-40 and PI). Nuclei were incubated with Tn5 transposon enzyme in transposase reaction mix buffer (Illumina) for 30 min at 37°C. DNAs were purified from the transposition sample by using Mini-Elute PCR purification kit (Qiagen, 28004) and measured by Qubit. Polymerase chain reaction (PCR) was performed to amplify with High-Fidelity 2X PCR Master Mix [72°C/5min +98°C/30 s + 12 × (98°C/10 s + 63°C/30 s + 72°C/60 s) + 72°C/5 min]. The libraries were purified using the Mini-Elute PCR purification kit (Qiagen). ATAC-seq libraries were pair-end sequenced on HiSeq4000 (Illumina) in the Hartwell Center at St Jude Children’s Research Hospital, Memphis, TN, USA. The ATAC-seq raw reads were aligned to the human reference genome (hg19) using BWA 90 and then marked duplicated reads with Picard (version 1.65), with only high-quality reads kept by samtools (version 1.3.1, parameter ‘‘-q 1 -F 1024’’). 91 Reads mapping to mitochondrial DNA were excluded from the analysis. All mapped reads were offset by +4 bp for the + strand and −5 bp for the – strand. 89 Peaks were called for each sample using MACS2 77 with parameters “-q 0.01 –nomodel –extsize 200 –shift 100”. Peaks were merged for the same cell types using BEDtools. 79 Peak annotation was performed using HOMER. 41 All sequencing tracks were viewed using the Integrated Genomic Viewer (IGV 2.3.82). 80 We downloaded 680 PWMs of human Transcription factor (TF) motifs from the HOCOMOCO database. 92 Then FIMO 93 was used to scan for occurrences of each motif in the hg19 reference genome with a p value cutoff of 10 −6 and to output all the putative binding sites across the genome for each TF motif. Motif density histograms were created using HOMER for target regions. Background regions were generated by selecting DNA sequences of equal length at 10 kb downstream of the target regions. The motif density at target regions was normalized to that at the control regions.
Animal experiments
The St. Jude Children’s Research Hospital Institutional Animal Care and Use Committee approved all studies performed. Mice were randomly assigned to experimental groups at 4–6 weeks of age. A 3-week treatment commenced when the implanted tumor volume reached 200 mm 3 QC6352 (25 mg/kg) was administered via oral gavage (PO), twice daily (BID), for 5 days on and 2 days off schedule. VCR (0.38 mg/kg) was given via intraperitoneal injection (IP) once weekly. IRN (1.25 mg/kg) was given via IP once daily, for 5 days on and 2 days off schedule. Tumor volume and body weight were measured twice weekly. Mice were euthanized when the tumor volume reached 20% of the body weight or the mice became moribund. Tumor response: For individual mice, progressive disease (PD) was defined as < 50% regression from initial volume during the study period and >25% increase in initial volume at the end of study period. Stable disease (SD) was defined as < 50% regression from initial volume during the study period and ≤25% increase in initial volume at the end of the study. Partial response (PR) was defined as a tumor volume regression ≥50% for at least one time point but with measurable tumor (≥0.10 cm3). Complete response (CR) was defined as a disappearance of measurable tumor mass (<0.10 cm3) for at least one time point.
Quantification and statistical analysis
All data in this study are displayed as the mean ± SEM. Comparison between two groups was determined using Student’s t test. Wilcoxon rank-sum test (two-sided) was used to compare the tumor volumes between two groups at every time point. p-values across multiple time points were adjusted for multiple comparison using the Benjamini-Hochberg method. Statistical analyses were performed using R version 4.0.2. Kaplan-Meier survival was analyzed using log rank (Mantel-Cox) method in Prism program.
Supplemental information
Document S1. Figures S1–S6 and Tables S1–S5 Document S2. Article plus supplemental information
Author contributions
S.S. performed animal experiments. A.A.-Z. performed in vitro studies of QC6352. H.J., W.P., B.X., A.D., N.A.M.S., and I.M.D. performed bioinformatic analyses. J.F. performed ATAC-seq, ChIP-seq, and CUT&TAG with help from B.X., W.R., and X.C. Q.W. synthesized QC6352. L.-A.V.D.V., P.J.M., and P.G.T. provided transgenic TH-MYCN/ALK F1178L mice. H.T. and L.Y. performed pathological analysis. Y.G. and Y.L. performed statistical analyses. P.P. performed DNA fiber assay. A.M.D., T.C., Z.R., and J.Y. secured the funding and supervised the project. K.W.F., M.E.H., and A.J.M. helped with ideas and data analysis. J.Y. conceived the project and wrote the manuscript with help from all authors.
Declaration of interests
The authors declare no competing interests.
Footnotes
Supplemental information can be found online at https://doi.org/10.1016/j.xcrm.2024.101468 .
Published: March 19, 2024
Article Images (0)
×![]()
Bioz is the world’s most comprehensive AI search engine for scientific experimentation. The Bioz search engine offers researchers billions of data-driven product, technique, and protocol recommendations. Bioz taps into the latest advances in Artificial Intelligence (AI) – including Natural Language Processing (NLP), Machine Learning (ML) and Generative AI to mine and structure hundreds of millions of pages of complex and unstructured scientific papers. By harnessing the power of AI, Bioz provides researchers with an unprecedented amount of summarized scientific experimentation knowledge right at their fingertips, ultimately helping to speed up drug discovery and the rate of success in finding cures for diseases. Bioz is used by over 15 Million researchers from over 17,000 different universities and companies in 196 countries.
