Noradrenergic neuroblastoma is characterized by a core transcriptional regulatory circuitry (CRC) comprised of transcription factors (TF) such as PHOX2B, HAND2, and GATA3, which form a network with MYCN. At normal physiologic levels, MYCN mainly binds to promoters but when aberrantly upregulated as in neuroblastoma, MYCN also binds to enhancers. Here, we investigated how MYCN invades enhancers and whether CRC TFs play a role in this process. HAND2 was found to regulate chromatin accessibility and to assist MYCN binding to enhancers. Moreover, HAND2 cooperated with MYCN to compete with nucleosomes to regulate global gene transcription. The cooperative interaction between MYCN and HAND2 could be targeted with an Aurora A kinase inhibitor plus a histone deacetylase inhibitor, resulting in potent downregulation of both MYCN and the CRC TFs and suppression of MYCN-amplified neuroblastoma tumor growth. This study identifies cooperation between MYCN and HAND2 in neuroblastoma and demonstrates that simultaneously targeting MYCN and CRC TFs is an effective way to treat this aggressive pediatric tumor.

Significance:

HAND2 and MYCN compete with nucleosomes to regulate global gene transcription and to drive a malignant neuroblastoma phenotype.

Distinct combinations of transcription factors (TF) need to work together to specifically bind to DNA and precisely regulate gene transcription (1). Transcriptional dysregulation is a hallmark of many cancers where MYC family gene amplification and translocation disrupt transcriptional programs in normal development (2). Under physiologic conditions, MYC or MYCN TF mainly binds to promoter regions. However, in cancer cells, aberrantly elevated MYC or MYCN binds to both gene promoters with canonical E-box motifs and enhancers containing low-affinity noncanonical E-boxes (3–5). This enhancer invasion is hypothesized to be one of the mechanisms by which aberrantly elevated MYC or MYCN drives tumorigenesis.

MYCN amplification occurs in tumors from patients with neuroblastoma, a tumor arising from neural crest sympathoadrenal derivatives and marks high-risk disease (6, 7). Studies indicate neuroblastoma subverts a subset of neural crest lineage–specific TFs including PHOX2B, HAND2, GATA3, and ASCL1 to form a core transcriptional regulatory circuitry (CRC) in noradrenergic neuroblastoma (8–12). MYCN invades enhancers and decreases in MYCN expression depleted genome bound TWIST1 at enhancers (5). However, mechanisms by which MYCN invades enhancers are ill defined. Because the studies on gene transcription models indicate that small numbers of TFs work together to achieve specific DNA binding (1, 13), we hypothesize that the CRC TFs assist MYCN to bind enhancers to govern a noradrenergic neuroblastoma tumor phenotype.

Here, we demonstrate that the CRC TF HAND2 facilitates MYCN enhancer invasion. HAND2 regulates chromatin accessibility and cooperates with MYCN to control global gene expression. More importantly, our findings indicate the need to target CRC TFs simultaneously in combination with MYCN to increase the efficacy of cancer therapies. Indeed, the use of an Aurora A kinase inhibitor (AURKAi) plus a HDAC inhibitor (HDACi) to target both MYCN and the CRC TFs synergistically suppresses neuroblastoma growth in MYCN-amplified cell lines.

Cell culture

Human embryonic kidney cells (HEK293T) and Lenti-X HEK293 were obtained from ATCC and were maintained in DMEM. Human neuroblastoma cell lines IMR32, IMR5, KCNR, LAN5, LAN6, and SK-N-FI were obtained from the cell line bank of the Pediatric Oncology Branch of the NCI and have been genetically verified. All the neuroblastoma cell lines were maintained in RPMI1640 medium. All the cell culture medium was supplemented with 10% FBS, 100 μg/mL streptomycin, 100 U/mL penicillin, and 2 mmol/L l-glutamine. Cells were grown at 37°C with 5% CO2. All cell lines were frequently assayed for Mycoplasma using MycoAlert Kit (Lonza) to ensure they were free of Mycoplasma contamination. The cell lines used were within 12 passages after thawing.

Plasmids and stable clones

To generate Hs_3_siHAND2-resistant HAND2 constructs, HAND2 coding sequence with mutated nucleotides that cannot be recognized by Hs_3_siHAND2 and without changes in the amino acid code was synthesized by IDT company (Supplementary Table S1). The HAND2-mutant (HAND2mut) open reading frame was cloned into the doxycycline (Dox)-inducible pLVX-pTetOne-puro vector (Takara Bio) using In-Fusion HD (Takara Bio) following the manufacturer's manual. By using the same approach, MYCN open reading frame (Supplementary Table S1) was cloned into pLVX-pTetOne-puro vector. IMR32 cells were infected with lentiviral particles generated using either empty pLVX-TetOne-Puro vector or the pLVX-TetOne-Puro-HAND2mut vector, or pLVX-TetOne-Puro-MYCN vector followed by puromycin selection. The stable clones named as IMR32tetEV, IMR32tetHAND2mut, and IMR32tetMYCN, respectively. HAND2 and MYCN expression in IMR32tetHAND2mut could be induced with 0.025–0.1 μg/mL Dox treatment.

Real-time PCR

The RNeasy Plus Mini Kit (Qiagen) was used to collect the mRNA according to the manufacturer's protocol. Quantitative measurements of total β-actin and other genes’ levels were obtained using the BIO-RAD CFX Touch realtime PCR detection system and performed in triplicate. Ct values were standardized to β-actin levels. Representative data from biological replicates were shown in this study. Primer sequences used for real-time PCR are shown in Supplementary Table S1.

CRISPR-Cas9–mediated gene knockout

IMR32 cells were transduced with the lentivirus of Edit-R Dox-Inducible Lentiviral Cas9 (Dharmacon) followed by blasticidin selection (5 μg/mL). Guide RNA (gRNA) that target the exon of human HAND2 gene (sgHAND2; Supplementary Table S2) was designed, synthesized, and cloned into the pLentiGuide-Puro vector (GenScript). The Dox-inducible Cas9-expressing IMR32 cells were transduced with sgHAND2 or nontargeting control gRNA (sgCtrl) lentiviral particles (multiplicity of infection = 10) and selected with puromycin (0.5 μg/mL). IMR32Cas9-sgCtrl or IMR32Cas9-sgHAND2 cell line was further single clone selected from the pool and cells are maintained in complete RPMI1640 containing 0.5 μg/mL puromycin.

Transient transfection

Transient transfection was performed as described previously (14). siRNAs were purchased from Qiagen or GE Dharmacon company (Supplementary Table S2). siRNAs were transiently transfected into neuroblastoma cells using Nucleofector electroporation (Lonza): solution L and program C-005 for IMR32 and IMR5; solution V and program A-030 for the rest neuroblastoma cell lines.

Cell growth and neurite extension assay

To evaluate cell proliferation, neuroblastoma cells were plated in 96-well plates and the growth kinetics were monitored in IncuCyte ZOOM (Essen BioScience) using the integrated confluence algorithm as a surrogate for cell number. An alternative approach is to use CellTiter-Glo (Promega) to perform cell viability assay. Cell neurite length was measured using Essen IncuCyte ZOOM neurite analysis software. SynergyFinder online tool (RRID:SCR_019318) was used to study the synergistic effect of the combination treatment of neuroblastoma cells in vitro.

Soft agar clonogenic assay

To assess the effects of the loss of HAND2 on anchorage-independent cell growth, 1 × 104 IMR32Cas9-sgCtrl or IMR32Cas9- sgHAND2 cells were cultured in 0.7% top agarose in media on a layer of 1.4% bottom agar/media to prevent the adhesion of cells to the culture plates. Medium was changed twice a week with or without 0.5 μg/mL Dox, and the number of visible colonies was counted after crystal violet staining after 3 weeks culture.

Protein isolation and Western blotting analysis

For assessment of protein levels, cells were lysed using RIPA buffer, and 10 μg of total protein was separated and electroblotted as described previously (15). Protein bands probed with diluted primary antibodies (Supplementary Table S3) were detected using a goat anti-rabbit or mouse IgG-HRP conjugated secondary antibody (200 μg/mL; Santa Cruz Biotechnology) and visualized using enhanced chemiluminescence (Amersham Biosciences).

RNA sequencing

Total RNA was isolated from IMR32 and LAN5 cells that have been transiently transfected with different siRNAs or siCtrl for 72 hours and subjected to RNA sequencing (RNA-seq) analysis as described previously (16). Each sample had three biological repeats. Statistical results of differentially expressed genes from Partek Flow, or Parteck Genomics Suite v7.17 or DESeq2 (RRID:SCR_000154) were analyzed using QIAGEN's Ingenuity Pathway Analysis (QIAGEN) and gene set enrichment analysis (GSEA). By default, the FDR less than 0.25 is significant in GSEA.

Assay for transposase-accessible chromatin using sequencing

Assay for transposase-accessible chromatin using sequencing (ATAC-seq) was performed as described previously (16). ATAC libraries were sequenced on an Illumina NextSeq machine (2 × 75 cycles). The peak sets for ATAC-seq were further analyzed using the deepTools2 suite (v3.3.0; ref. 17). By using bamCoverage, peaks were normalized to reads per kilobase per million reads normalized read numbers.

Chromatin immunoprecipitation sequencing

Chromatin immunoprecipitation sequencing (ChIP-seq) was performed using the ChIP-IT High Sensitivity kit (Active Motif, catalog no. 53040) as described previously (16). Briefly, formaldehyde (1%, 15 minutes) fixed cells were sheared to achieve chromatin fragmented to a range of 200–700 bp using an Active Motif EpiShear Probe Sonicator. IMR32 cells were sonicated at 25% amplitude, pulse for 20 seconds on and 30 seconds off for a total sonication “on” time of 16 minutes. Sheared chromatin samples were immunoprecipitated overnight at 4°C with antibodies targeting different proteins (Supplementary Table S3). To normalize ChIP-seq signal, we employed Active Motif ChIP-seq spike-in using Drosophila chromatin (Active Motif catalog no. 53083) and an antibody against Drosophila specific histone variant H2Av (Active Motif, catalog no. 61686) according to the manufacturer's instructions. ChIP-seq DNA libraries were prepared by Frederick National Laboratory for Cancer Research sequencing facility. Libraries were multiplexed and sequenced using TruSeq ChIP Samples Prep Kit (75 cycles), catalog no. IP-2-2-1012/1024 on an Illumina NextSeq machine.

ChIP-seq data processing

ChIP-seq data processing was performed as described previously (16). Peaks from ChIP-seq of MYCN, HAND2, PHOX2B, GATA3, H3K27ac, H3K4me3, H3K27me3, and RNA Pol II were selected at a stringent P value (P < 10−5 for PHOX2B and P < 10−7 for the rest of the targets). Peaks within 1,000 bp to the nearest transcription start site (TSS) were set as promoter. The distribution of peaks (as intronic, intergenic, exonic, etc.) was annotated using HOMER. Enrichment of known and de novo motifs were found using HOMER script “find Motifs Genome.pl” (RRID:SCR_010881). Reference genome normalization (RRPM, reference-adjusted reads per million mapped reads) was calculated with the ChIP-Rx method (18).

Monitoring of synergistic effects of drug combinations

The therapeutic effect of AURKAi alisertib (MedChemExpress, HY-10971) and HDACi LBH589 (MedchemExpress, HY-10224) in neuroblastoma cell lines was determined in a checkerboard fashion. Cell lines were seeded in two 96-well plates and incubated overnight. Each combination dose had two replications. Next day, cell lines were treated with different dose combination of alisertib and LBH589. Control cells were treated with DMSO. Each plate has their control cells. Cell viability was determined after 72 hours using the CellTiter-Glo luminescent assay (Promega, catalog no. G9242). Cell viabilities of DMSO-treated cells were set to 100%. Results were graphed with GraphPad Prism (RRID:SCR_002798) software. IncuCyte cell confluence assay was used for testing the impact of synergistic effects of drug combinations to neuroblastoma cell growth in real time. Representative data from biological replicates were shown in this study.

Xenograft tumor studies

To test the therapeutic effect of alisertib and LBH589 in vivo, 4–6 weeks old female athymic nude mice (Frederick Animal Facility, NCI) were orthotopically injected (injected through the adrenal fat pad into the adrenal gland) with 2.5 × 105 IMR5-GFP-Luc cells into the in 30 μL Matrigel. Two weeks later, the luminescence signal in the tumor cells was measured by in vivo imaging system imaging. When the luminescence signal >1 × 108, mice were grouped into four groups (10–11 mice/group): Group 1, treated with vehicle; Group 2, treated with 10 mg/kg alisertib (disserved with 10% DMSO, 40% PEG300, 5% TWEEN-80, 45% saline) by oral gavage twice a day for 7 consecutive days (cycle days 1–7) in a 21-day cycle; Group 3, treated with 3 mg/kg LBH589 (disserved with 15% DMSO and saline) by intraperitoneal injection three times (M‐W‐F) a week; Group 4, treated with 10 mg/kg alisertib and 3 mg/kg LBH589. After 6 weeks (two cycles) of drug treatment, mice were euthanized, and tumors were collected and weighed. For the pharmacodynamic study, when the luminescence signal reached 1 × 108 total flux (photons per second), mice were grouped into four groups, which were treated with vehicle, alisertib, LBH589, and alisertib + LBH589, respectively. Alisertib (10 mg/kg) was given at time 0, 8, and 24 hours, LBH589 (3 mg/kg) was given at time 0 and 24 hours. Mice were euthanized 4 hours after the final treatment and tumors were collected for Western blot analysis. All xenograft studies were approved by the NCI's Animal Care and Use Committee, and all animal care was in accordance with institutional guidelines.

Data availability

All the home generated RNA-seq, ChIP-seq, and ATAC-seq datasets can be found in the Gene Expression Omnibus (GEO) (RRID:SCR_005012) database. GEO accession number for data generated in this study is GSE183641, and the subseries that are linked to GSE183641 are GSE183609 and GSE183636 for RNA-seq; GSE184057, GSE184058, and GSE184059 for ChIP-seq; GSE184229 and GSE184232 for ATAC-seq. GEO accession number for publicly available ChIP-seq data derived from BE2C cells is GSE94822.

Statistical analysis

The statistical analyses used throughout this article are specified in the appropriate result paragraphs and Materials and Methods. Additional statistical analyses were performed using standard two-tailed Student t test, one-way ANOVA, and the software GraphPad Prism 8.1.0.

CRC TF HAND2 regulates the expression of MYCN targets and is essential in neuroblastoma

We analyzed the published ChIP-seq data of neuroblastoma CRC TFs and MYCN in the SK-N-BE(2)C (BE(2)C) cell line (MYCN amplified; ref. 10) through the Intervene tool (19), and identified both unique and overlapping binding sites among these TFs (Fig. 1A). For example, signal tracks showed different combinatorial colocalization among these TFs on genomic loci near the GATA2 gene locus (Fig. 1B).

Figure 1.

HAND2 regulates the expression of MYCN targets and is essential in neuroblastoma. A, Intersection plot shows the overlap of ChIP-seq peaks of CRC TFs and MYCN determined by ChIP-seq data analysis. B, ChIP-seq signal tracks show the overlapped binding sites of CRC TFs and MYCN next to GATA2 gene locus. C, The knockdown of each of the CRC TFs for 72 hours detected by Western blot assay. D, The knockdown of CRC TFs results in a decrease of cell number shown by the IncuCyte cell confluence assay. E, The regulation of a panel MYCN direct target genes by MYCN and the CRC TFs is detected by real-time PCR and shown in the heatmap after silencing each of these TFs for 72 hours. F, The knockdown of HAND2 in different neuroblastoma cell lines for 72 hours detected by Western blot assay. G, The effect of silencing of HAND2 on neuroblastoma cell proliferation detected by CellTiter-Glo assay. H, Depletion of HAND2 through Dox (0.5 μg/mL)-inducible Crispr/Cas9 system results in a decrease of anchorage-independent cell growth shown by the reduced colony formation in soft agar.

Figure 1.

HAND2 regulates the expression of MYCN targets and is essential in neuroblastoma. A, Intersection plot shows the overlap of ChIP-seq peaks of CRC TFs and MYCN determined by ChIP-seq data analysis. B, ChIP-seq signal tracks show the overlapped binding sites of CRC TFs and MYCN next to GATA2 gene locus. C, The knockdown of each of the CRC TFs for 72 hours detected by Western blot assay. D, The knockdown of CRC TFs results in a decrease of cell number shown by the IncuCyte cell confluence assay. E, The regulation of a panel MYCN direct target genes by MYCN and the CRC TFs is detected by real-time PCR and shown in the heatmap after silencing each of these TFs for 72 hours. F, The knockdown of HAND2 in different neuroblastoma cell lines for 72 hours detected by Western blot assay. G, The effect of silencing of HAND2 on neuroblastoma cell proliferation detected by CellTiter-Glo assay. H, Depletion of HAND2 through Dox (0.5 μg/mL)-inducible Crispr/Cas9 system results in a decrease of anchorage-independent cell growth shown by the reduced colony formation in soft agar.

Close modal

To determine the biological function of the CRC TFs in neuroblastoma, we mined the Project Achilles genome-wide CRISPR-Cas9 screen data. We found that compared with other pediatric cancer cell lines, neuroblastoma cell lines in particular are preferentially dependent on these TFs (Supplementary Fig. S1A), which is consistent with the essential role of CRC TFs previously identified in neuroblastoma (8–11). We silenced each of these TFs using two different siRNAs (Fig. 1C; Supplementary Fig. S1B) and found that the more effective knockdown resulted in a greater decrease in cell number (Fig. 1D; Supplementary Fig. S1C). Usually, silencing of one CRC member resulted in a decreased expression of other CRC members (Supplementary Fig. S1D).

We picked five well-known MYCN direct target genes and performed real-time PCR to investigate how MYCN targets were regulated by each of these CRC TFs. Results showed the transcriptional activity of HAND2 most closely phenocopied loss of MYCN (Fig. 1E). Silencing of HAND2 significantly reduced cell viability in all the other tested neuroblastoma cell lines albeit to varying levels (Fig. 1F and G). To confirm that this biological function was not due to an off-target effect of siHAND2_3, we performed a HAND2 rescue experiment by overexpressing a siHAND2_3-resistant HAND2-mutant construct (see Materials and Methods). The overexpression of this HAND2-mutant construct attenuated both the siHAND2_3-induced upregulation of GAP43 mRNA levels and repression of cell proliferation (Supplementary Fig. S1E and S1F), demonstrating the specificity of siHAND2_3. We generated CRISPR-Cas9–inducible IMR32 stable clones, with a control (sgCtrl) or a single-guide RNA that targeted HAND2 gene locus (sgHAND2). Dox treatment decreased HAND2 protein levels (Supplementary Fig. S1G), which was accompanied by decreased cell proliferation and colony formation in soft agar (Fig. 1H; Supplementary Fig. S1H), indicating a role of HAND2 in neuroblastoma tumorigenicity.

In neuroblastoma patient data, we did not observe a consistent association between HAND2 expression levels and the overall survival in two different neuroblastoma patient datasets (Supplementary Fig. S1I). However, in MYCN-amplified neuroblastoma patient data, we found that high expression of HAND2 associates with poor overall survival (Supplementary Fig. S1J), suggesting a cooperative role for HAND2 and MYCN.

HAND2 regulates chromosome accessibility and facilitates MYCN enhancer invasion

To determine whether HAND2 assists MYCN binding to DNA, we silenced HAND2 in IMR32 cells and performed ChIP-seq and ATAC-seq. ChIP-seq data analyses showed that 70% of the MYCN binding sites overlapping with HAND2 binding sites are at the active enhancer regions (overlapping with H3K27ac binding sites but outside of the promoter; Fig. 2A). However, almost 50% of the MYCN binding sites that did not overlap with HAND2 binding sites (MYCN unique peaks) were at promoter regions (Fig. 2B). HOMER (20) de novo motif scan identified low-affinity E-box-like sequences (CANNTG) in the peaks MYCN shared with HAND2, while the MYCN unique peaks contain a high-affinity canonical E-box (CACGTG; Fig. 2A and B). Genomic Regions Enrichment of Annotations Tool (GREAT; ref. 21) gene ontology (GO) analysis revealed genes associated with MYCN and HAND2 overlapping sites are enriched in the regulation of sympathetic nervous system development, while the genes associated with unique MYCN peaks are enriched in RNA processing (Fig. 2C and D).

Figure 2.

HAND2 assists MYCN to bind to DNA. A, Venn diagram shows the distribution of MYCN binding sites that overlapped with HAND2 binding sites in IMR32 cells (top). Motif scan shows that the MYCN binding motif is a noncanonical E-box. B, Venn diagram shows the distribution of MYCN unique binding sites in IMR32 cells. Motif scan shows that the MYCN binding motif is a canonical E-box. C, GO analysis of MYCN and HAND2 overlapped peaks–associated genes. D, GO analysis of MYCN unique peaks–associated genes. E, Composite profile of ChIP-seq data shows the average ChIP-seq signal of MYCN and ATAC-seq signal after the knockdown of HAND2 for 72 hours in IMR32 cells. F, Signal tracks show that the knockdown of HAND2 results in a decrease of MYCN signal and ATAC-seq signal within the CARMN gene locus. G, Composite plots show that the overexpression of HAND2 in IMR32 cells for 36 hours results in an increased average ChIP-seq signal of MYCN that is accompanied by an increase of average ATAC-seq signal. H, Signal tracks show that the overexpression of HAND2 results in an increased MYCN signal and ATAC-seq signal within the CARMN gene locus. I, Composite plots show a decreased average GATA3 ChIP-seq signal but no change of PHOX2B signal after knocking down HAND2 in IMR32 cells for 72 hours. J and K, Composite plots show the influence of the knockdown of PHOX2B or GATA3 in IMR32 cells for 72 hours on the average ChIP-seq signal of MYCN and ATAC-seq signal. RRPM, spike-in normalized, reference-adjusted reads per million mapped reads; RPKM, reads per kilobase per million mapped reads.

Figure 2.

HAND2 assists MYCN to bind to DNA. A, Venn diagram shows the distribution of MYCN binding sites that overlapped with HAND2 binding sites in IMR32 cells (top). Motif scan shows that the MYCN binding motif is a noncanonical E-box. B, Venn diagram shows the distribution of MYCN unique binding sites in IMR32 cells. Motif scan shows that the MYCN binding motif is a canonical E-box. C, GO analysis of MYCN and HAND2 overlapped peaks–associated genes. D, GO analysis of MYCN unique peaks–associated genes. E, Composite profile of ChIP-seq data shows the average ChIP-seq signal of MYCN and ATAC-seq signal after the knockdown of HAND2 for 72 hours in IMR32 cells. F, Signal tracks show that the knockdown of HAND2 results in a decrease of MYCN signal and ATAC-seq signal within the CARMN gene locus. G, Composite plots show that the overexpression of HAND2 in IMR32 cells for 36 hours results in an increased average ChIP-seq signal of MYCN that is accompanied by an increase of average ATAC-seq signal. H, Signal tracks show that the overexpression of HAND2 results in an increased MYCN signal and ATAC-seq signal within the CARMN gene locus. I, Composite plots show a decreased average GATA3 ChIP-seq signal but no change of PHOX2B signal after knocking down HAND2 in IMR32 cells for 72 hours. J and K, Composite plots show the influence of the knockdown of PHOX2B or GATA3 in IMR32 cells for 72 hours on the average ChIP-seq signal of MYCN and ATAC-seq signal. RRPM, spike-in normalized, reference-adjusted reads per million mapped reads; RPKM, reads per kilobase per million mapped reads.

Close modal

We found that approximately 34% MYCN binding sites at the active enhancers did not overlap with HAND2 binding sites (Fig. 2B), indicating that HAND2 binds to a subset of MYCN-bound enhancers. GO analysis showed that the active enhancers bound by both MYCN and HAND2 associate with genes involved in noradrenergic neuron differentiation (Supplementary Fig. S2A, left), while the active enhancers only bound by MYCN associate with chromatin organization and modification (Supplementary Fig. S2A, right). Motif scan showed enhancers bound by both MYCN and HAND2 are enriched with low-affinity binding noncanonical E-box, PHOX2B and HAND2 binding motifs (Supplementary Fig. S2B, left), while MYCN unique enhancers are enriched with high-affinity canonical E-box but not CRC TF binding motifs (Supplementary Fig. S2B, right). These analyses indicate that the enhancers colocalized by HAND2 and MYCN are selectively more noradrenergic than MYCN-specific enhancers.

We next investigated the influence of HAND2 on MYCN binding to DNA by knocking down HAND2 in IMR32 cells (Supplementary Fig. S2C). At the HAND2 and MYCN overlapping binding sites, loss of HAND2 caused a 40% decrease (summit of the composite plot) in the average MYCN ChIP-seq signal, as well as a 30% decrease in the average ATAC-seq signal, reflecting a decrease of chromatin accessibility (Fig. 2E). A representative example within the CARMN gene locus is shown by signal tracks (Fig. 2F). The depletion of HAND2 was not always accompanied by decreased MYCN binding. Thus, we stratified the ChIP-seq peaks into Group I, those with decreased MYCN ChIP-seq signals (>40% decrease); or Group II, those with no change in MYCN signals after HAND2 silencing (within ±1.1-fold changes). Analysis of the resulting peaks (heatmap) showed that the loss of HAND2 decreased MYCN and ATAC-seq signals in Group I regions (Supplementary Fig. S2D). In contrast after HAND2 knockdown, the Group II regions were not associated with changes in MYCN and ATAC-seq signals (Supplementary Fig. S2D). These results indicate that the removal of both HAND2 and MYCN from DNA is associated with a decrease in chromatin accessibility.

In HAND2 gain-of-function studies, induction of HAND2 for 36 hours in IMR32 cells did not alter MYCN levels (Supplementary Fig. S2E). The ChIP-seq signal of many of the HAND2 peaks did not increase after the overexpression of HAND2 possibly due to the saturation of endogenous HAND2 on these genomic loci. Thus, we focused on those HAND2 and MYCN common peaks whose HAND2 ChIP-seq signal increased (>3-fold). On these, a gain of HAND2 expression increased the average MYCN ChIP-seq and ATAC-seq signals (Fig. 2G and H).

We further examined MYCN binding on DNA at 24 hours after knocking down HAND2 in IMR32 cells when there was no change in MYCN protein levels (Supplementary Fig. S2F). By focusing on all the HAND2 and MYCN overlapping binding sites, the loss of HAND2 resulted in a 5% decrease in the average MYCN ChIP-seq signal at the summit (Supplementary Fig. S2G, left). However, for the HAND2 and MYCN overlapping peaks with >4-fold decrease of HAND2 ChIP-seq signal, there was a 25% decrease of MYCN ChIP-seq signal at the summit (Supplementary Fig. S2G, middle). The loss of HAND2 for 24 hours did not affect the MYCN ChIP-seq signal of those MYCN unique peaks that did not overlap with HAND2 binding sites (Supplementary Fig. S2G, right). IMR32tetHAND2 cells treated with Dox for 6 hours resulted in increases in HAND2 protein levels without affecting MYCN expression (Supplementary Fig. S2H). Yet the overexpression of HAND2 for 6 hours resulted in a significant increase of the average MYCN ChIP-seq signal at HAND2 peak center with HAND2 signals increasing by 3- to 4-fold (Supplementary Fig. S2I and J). Representative signal tracks close to TBX4 or UQCRFS1 gene locus demonstrated this influence of HAND2 silencing on MYCN binding to DNA in a time-dependent manner (Supplementary Fig. S2K and S2L).

Using a Dox-inducible MYCN overexpression IMR32 stable cell line, we found that over expressing MYCN did not rescue the decrease in cell proliferation induced by HAND2 silencing (Supplementary Fig. S2M and S2N). Next, we performed MYCN ChIP-seq analysis in MYCN-overexpressed IMR32 cells before and after the silencing of HAND2 and found that HAND2 depletion caused a 30% decrease in the average MYCN ChIP-seq signal at the summit (Supplementary Fig. S2O). Thus, HAND2 is required to recruit MYCN to chromatin and the overexpression of MYCN cannot fully rescue the effect of silencing HAND2.

HAND2, PHOX2B, and GATA3 are broadly identified as CRC TFs in tested neuroblastoma cell lines (8–10). Next, we investigated how HAND2 silencing affects the PHOX2B and GATA3 binding to MYCN regulated genes. The silencing of HAND2 for 72 hours resulted in a >20% decrease in GATA3 protein levels but did not alter PHOX2B expression (Supplementary Fig. S2P). Composite plots at the HAND2 and MYCN overlapping binding sites showed decreases in average GATA3 but not PHOX2B ChIP-seq signals after silencing HAND2 (Fig. 2I). Decreases in GATA3 ChIP-seq signal might partially be due to decreases in GATA3 protein levels caused by HAND2 silencing (Supplementary Fig. S2P). Nevertheless, the depletion of HAND2 results in decreases in DNA-bound MYCN and GATA3 but not PHOX2B, suggesting a complicated cooperative gene regulatory network.

To investigate whether PHOX2B and GATA3 are required for MYCN to bind to DNA, we silenced their expression in IMR32 cells and performed ChIP-seq and ATAC-seq experiments. Western blot results showed that PHOX2B silencing for 72 hours did not affect MYCN protein levels (Supplementary Fig. S2Q) while GATA3 silencing resulted in an increase in MYCN protein levels (Supplementary Fig. S2R). The composite plots showed that the average ChIP-seq signal of MYCN did not change after PHOX2B silencing (Fig. 2J), while the GATA3 silencing increased MYCN ChIP-seq signal (Fig. 2K), which might be due to increases in MYCN protein after GATA3 silencing (Supplementary Fig. S2R). Our results indicate that CRC TF HAND2, but not PHOX2B or GATA3 is required for MYCN to bind to DNA.

Next, we investigated how MYCN silencing affects the CRC TF–DNA interactions. Knockdown of MYCN for 72 hours increased GATA3 protein levels but not the other TFs (Supplementary Fig. S2S). Composite plots showed that MYCN silencing led to decreases in MYCN ChIP-seq signal but not HAND2 binding (Supplementary Fig. S2T). By focusing on GATA3 and PHOX2B, we found that depletion of MYCN resulted in increases in GATA3 and PHOX2B ChIP-seq signals (Supplementary Fig. S2U and S2V).

To determine whether HAND2 facilitates MYCN DNA binding and increases chromatin accessibility is cell line specific, we analyzed LAN5 neuroblastoma cells. HAND2 silencing for 72 hours resulted in a 17% decrease in MYCN protein (Supplementary Fig. S2W) with a 30% decrease in the average MYCN ChIP-seq signal at the summit of the composite plots (Supplementary Fig. S2X). Only a 5% decrease in the average ATAC-seq signal was observed (Supplementary Fig. S2X). In IMR32 cells only when both HAND2 and MYCN were absent from DNA, there was an obvious decrease of chromatin accessibility as indicated by decreases in ATAC-seq signal (Supplementary Fig. S2D). Thus, we specifically focused on those HAND2 and MYCN overlapped binding sites with a >1.5-fold decrease of MYCN ChIP-seq signal after HAND2 depletion in LAN5 cells. Remarkably, composite plots showed an approximately 50% decrease in ATAC-seq signal at summit when both HAND2 and MYCN were absent from DNA upon HAND2 silencing (Supplementary Fig. S2Y). These results indicate that HAND2 facilitates MYCN enhancer binding, and the cooperative-binding of these TFs on DNA is associated with increased chromatin accessibility.

HAND2 regulates the expression of a subset of MYCN target genes

To investigate the global gene expression regulated by HAND2 or MYCN, we performed RNA-seq in HAND2- or MYCN-silenced IMR32 cells (Supplementary Fig. S2C). GSEA of the RNA-seq data showed that loss of HAND2 resulted in a significant negative enrichment of MYC targets and positive enrichment of neuronal gene signatures (Fig. 3A). Around 44% of the genes regulated by MYCN were also regulated by HAND2 (Fig. 3B; Supplementary Table S4). In LAN5 cells, HAND2 silencing (Supplementary Fig. S2W) resulted in a significant negative enrichment of MYC targets and positive enrichment of neuronal gene signatures (Fig. 3C). Around 33% of the genes regulated by MYCN were also regulated by HAND2 (Fig. 3D; Supplementary Table S5). HAND2 silencing resulted in a positive enrichment of neuronal genes and increased functional neuronal differentiation as shown by neurite extension in both IMR32 cells and LAN5 cells (Supplementary Fig. S3A and S3B). We found that among the 216 genes regulated by both MYCN and HAND2 in LAN5 cells, 46 neuronal-enriched genes were also regulated by both MYCN and HAND2 in IMR32 cells (Supplementary Table S5).

Figure 3.

HAND2 regulates the expression of a large subset of MYCN targets. A, GSEA shows a negative enrichment of MYC signature genes and a positive enrichment of neuronal genes after the silencing of HAND2 in IMR32 cells for 72 hours. B, Venn diagram shows that around 45% genes regulated by MYCN is also regulated by HAND2 in IMR32 cells. C, GSEA shows a negative enrichment of MYC signature genes and a positive enrichment of neuronal genes after knockdown of HAND2 in LAN5 cells for 72 hours. D, Venn diagram shows that around 33% of genes regulated by MYCN are also regulated by HAND2 in LAN5 cells.

Figure 3.

HAND2 regulates the expression of a large subset of MYCN targets. A, GSEA shows a negative enrichment of MYC signature genes and a positive enrichment of neuronal genes after the silencing of HAND2 in IMR32 cells for 72 hours. B, Venn diagram shows that around 45% genes regulated by MYCN is also regulated by HAND2 in IMR32 cells. C, GSEA shows a negative enrichment of MYC signature genes and a positive enrichment of neuronal genes after knockdown of HAND2 in LAN5 cells for 72 hours. D, Venn diagram shows that around 33% of genes regulated by MYCN are also regulated by HAND2 in LAN5 cells.

Close modal

Loss of HAND2 affects regional epigenetic modifications

To investigate how HAND2 regulates gene transcription at the epigenetic levels, we silenced HAND2 in IMR32 cells and performed ATAC-seq and ChIP-seq for histone marks, RNA Pol II, RNA Pol II Ser2P (S2), and BRD4. Analysis of HAND2 ChIP-seq identified 25,549 high-confidence HAND2 binding sites in the IMR32 cells that aligned with or without those transcriptional marks shown in the heatmaps (Fig. 4A). GREAT analysis of HAND2 binding sites showed that the majority of HAND2 binding sites were distal (>5 kb) to the transcription start site TSS (Fig. 4B). Peak distribution analysis showed that 46% of HAND2 binding sites were at the active enhancer regions (overlapped with H3K27ac peaks; Fig. 4C). Homer de novo motif scanning of HAND2 peaks identified two highly enriched known HAND2 binding motifs (Fig. 4D) identified in neuroblastoma (8) or in human embryonic stem cell–derived mesoderm cells (GSE61475). GO analysis showed HAND2 peak–associated genes were enriched in sympathetic nervous system development (Fig. 4E).

Figure 4.

Loss of HAND2 affects regional epigenetic modification. A, Heatmaps of ChIP-seq data at the ranked HAND2 peak center before (siCtrl) and after (siHAND2) knocking down of HAND2 for 72 hours in IMR32 cells. B, HAND2 peak distribution determined by prediction tool GREAT. C, Venn diagram shows that around 46% of HAND2 binding sites are at active enhancer regions (overlapping with H3K27ac binding sites). D, Motif scan shows the enrichment of two known HAND2 binding motifs. E, HAND2 binding sites–associated genes identified by GREAT GO analysis. F, The silencing of HAND2 affects SEs establishment. G, Signal tracks show that the knockdown of HAND2 decreased signal of H3K27ac and RNA-seq at the HAND1 gene locus. H and I, Composite plots show the changes of the average ChIP-seq signals of the indicated proteins after the knockdown of HAND2 for 72 hours in IMR32 cells.

Figure 4.

Loss of HAND2 affects regional epigenetic modification. A, Heatmaps of ChIP-seq data at the ranked HAND2 peak center before (siCtrl) and after (siHAND2) knocking down of HAND2 for 72 hours in IMR32 cells. B, HAND2 peak distribution determined by prediction tool GREAT. C, Venn diagram shows that around 46% of HAND2 binding sites are at active enhancer regions (overlapping with H3K27ac binding sites). D, Motif scan shows the enrichment of two known HAND2 binding motifs. E, HAND2 binding sites–associated genes identified by GREAT GO analysis. F, The silencing of HAND2 affects SEs establishment. G, Signal tracks show that the knockdown of HAND2 decreased signal of H3K27ac and RNA-seq at the HAND1 gene locus. H and I, Composite plots show the changes of the average ChIP-seq signals of the indicated proteins after the knockdown of HAND2 for 72 hours in IMR32 cells.

Close modal

ChIP-seq and ATAC-seq heatmaps demonstrated that most of the ranked HAND2-centered peaks identified in siCtrl IMR32 cells aligned with the H3K27ac, RNA Pol II peaks, and ATAC-seq peaks (Fig. 4A), indicating that most HAND2 binding sites are within accessible chromatin regions. No H3K27me3 signal and weak H3K4me3 signal were detected at the HAND2 peak center (Fig. 4A), both of which were known to be restricted from enhancer regions. Silencing of HAND2 resulted in an almost complete loss of HAND2 signal that was accompanied by decreases in H3K27ac signal at the HAND2 peak center, while no obvious decrease of average BRD4 ChIP-seq signal was observed (Fig. 4A). Coimmunoprecipitation using anti-HAND2 antibody indicated that HAND2 did not pull down BRD4 (Supplementary Fig. S4A). These results suggest that BRD4 does not mediate HAND2 function at facilitating MYCN enhancer invasion or increasing chromatin accessibility.

As a component of the CRC, HAND2 is driven by super-enhancers (SE) and binds to the SEs of genes important for cell identity (8, 10). However, whether HAND2 plays a role in the establishment of SEs has not been investigated. We identified 409 SEs (associated with 355 genes) in siCtrl and 439 SEs (associated with 382 genes) in siHAND2 cells (Fig. 4F). The loss of HAND2 resulted in a loss of 87 SE-associated genes and an acquisition of 114 SE-associated genes with 267 SE-associated genes relatively unchanged (Supplementary Table S6). Representative ChIP-seq signal tracks showed that HAND2 silencing resulted in a dramatic decrease of H3K27ac and RNA-seq signals at the HAND1 gene locus (Fig. 4G), and a dramatic increase of H3K27ac and RNA-seq signals at the CNTN2 gene locus (Supplementary Fig. S4B).

To investigate how HAND2 regulates gene transcription, we focused on genes with HAND2 binding sites that were transcriptionally regulated by HAND2. In combination with HAND2 ChIP-seq and RNA-seq results, we identified 605 directly downregulated genes (associate with 2,065 peaks) and 464 directly upregulated genes (associate with 1,890 peaks) after HAND2 silencing (Supplementary Table S7). By focusing on the directly “down-regulated genes” after HAND2 silencing, the composite plots showed that the average H3K27ac signal decreased upon HAND2 silencing (Fig. 4H). Decreased H3K27ac but not H3K4me3 and H3K27me3 signals occurred at TSS (Fig. 4H), and decreased signals for RNA Pol II Ser2P but not RNA Pol II occurred both within the gene bodies and in the regions downstream of the polyadenylation signals (Fig. 4H; Supplementary Fig. S4C). Pathway analysis of these “downregulated genes” indicated that HAND2 silencing activated PTEN signaling (Supplementary Fig. S4D). For genes directly “upregulated” after HAND2 silencing, the composite plots showed no change for H3K27ac signal at HAND2 peak centers, but increased H3K27ac and H3K4me3 signals, and decreased H3K27me3 signal were observed at TSS (Fig. 4I). Increased RNA Pol II and RNA Pol II Ser2P signals in the region downstream of the polyadenylation signals were observed (Fig. 4I; Supplementary Fig. S4E). Pathway analysis of genes upregulated after HAND2 silencing (HAND2 repressed genes) showed that these genes were associated with activated neuronal development–related signaling (Supplementary Fig. S4F). Our results suggest that HAND2 activates gene transcription mainly by increasing enhancer activity, while HAND2 represses gene transcription mainly by decreasing promoter activity.

HAND2 and MYCN work together to cooperatively regulate chromatin accessibility

The “cooperative” TF-DNA binding model proposes that the TFs recognize adjacent binding sites and work together to compete with a nucleosome to access and cooperatively bind DNA (22–24). If HAND2 and MYCN cooperate in this way, the loss of both HAND2 and MYCN should have a dramatic effect on nucleosome occupancy. To test this both HAND2 and MYCN were silenced (Supplementary Fig. S5A), and ATAC-seq for chromatin accessibility was assessed. By focusing on the MYCN and HAND2 common binding sites, we stratified the ChIP-seq peaks into Group I, those that had a decreased MYCN ChIP-seq signal (>1.5-fold change), or Group II in which there was no change of ChIP-seq signal (<1.1-fold change) in the MYCN signal after the silencing of HAND2. Analysis of the resulting peaks (heatmap) showed that the loss of HAND2 decreased MYCN binding and was accompanied by a decrease in the ATAC-seq signal in Group I regions (Fig. 5A, left). In contrast after HAND2 knockdown, the Group II regions were not associated with changes in MYCN binding or average ATAC-seq signal (shown in the heatmap; Fig. 5A, left). The same results were also shown in composite plots (Fig. 5A, right). The knockdown of both HAND2 and MYCN led to greater decreases in chromatin accessibility shown by the decreased ATAC-seq signal in both heatmaps and composite plots (Fig. 5A). We specifically focused on those HAND2 and MYCN common peaks in which there were no changes in MYCN binding and ATAC-seq signal after silencing HAND2 alone (Fig. 5A, Group II). In these genomic regions, we found that the knockdown of either HAND2 or MYCN alone had no effect on chromatin accessibility. It was only the loss of both HAND2 and MYCN that decreased chromatin accessibility as detected by the decreased ATAC-seq signal (Fig. 5A).

Figure 5.

HAND2 and MYCN cooperatively regulate chromatin accessibility and gene transcription. A, Heatmaps (left) and composite profiles (right) of ChIP-seq and ATAC-seq results before and after the knockdown of HAND2 (siHAND2), MYCN (siMYCN), or both (siH+M) in IMR32 cells (72 hours). B, Heatmap shows that the knockdown of both HAND2 and MYCN using siRNAs (siH+M) results in greater effect on the downregulation or upregulation of gene expression than knocking down each one alone (siH or siM) in IMR32 cells (72 hours). Ingenuity canonical pathway analysis shows the associated pathways of these downregulated and upregulated genes. C, GSEA shows that the knockdown of both HAND2 and MYCN results in a negative enrichment of MYC target genes and genes involved in G2–M checkpoint. D, GSEA shows that the knockdown of both HAND2 and MYCN results in a positive enrichment of neuron markers and neuronal differentiation genes. E and F, The silencing of both of HAND2 and MYCN results in a more significant downregulation of genes that is required for G2–M progression and upregulation of neuronal differentiation genes based on the RNA-seq results. CPM, counts per million. G and H, The knockdown of both HAND2 and MYCN results in a significant increase of neurite length shown by the cell image and the IncuCyte neurite-length assay.

Figure 5.

HAND2 and MYCN cooperatively regulate chromatin accessibility and gene transcription. A, Heatmaps (left) and composite profiles (right) of ChIP-seq and ATAC-seq results before and after the knockdown of HAND2 (siHAND2), MYCN (siMYCN), or both (siH+M) in IMR32 cells (72 hours). B, Heatmap shows that the knockdown of both HAND2 and MYCN using siRNAs (siH+M) results in greater effect on the downregulation or upregulation of gene expression than knocking down each one alone (siH or siM) in IMR32 cells (72 hours). Ingenuity canonical pathway analysis shows the associated pathways of these downregulated and upregulated genes. C, GSEA shows that the knockdown of both HAND2 and MYCN results in a negative enrichment of MYC target genes and genes involved in G2–M checkpoint. D, GSEA shows that the knockdown of both HAND2 and MYCN results in a positive enrichment of neuron markers and neuronal differentiation genes. E and F, The silencing of both of HAND2 and MYCN results in a more significant downregulation of genes that is required for G2–M progression and upregulation of neuronal differentiation genes based on the RNA-seq results. CPM, counts per million. G and H, The knockdown of both HAND2 and MYCN results in a significant increase of neurite length shown by the cell image and the IncuCyte neurite-length assay.

Close modal

GO analysis of Group I MYCN peak (Fig. 5A)-associated genes showed that these genes are dominantly enriched in nervous system development, while Group II MYCN peak (Fig. 5A)-associated genes are enriched in both nervous system development and heart development (Supplementary Fig. S5B and S5C). Both HAND2 and MYCN have been reported to be essential in regulating heart development (25, 26). These results indicate that although these TFs bind to heart development–associated genes in neuroblastoma, the depletion of HAND2 only affects MYCN binding on neuronal genes but not cardiac genes. Motif scan showed some difference in the motifs enriched in Group I and II (Supplementary Fig. S5D and S5E). HAND2 is known to bind to the noncanonical E-box, as well as a thymine (T) nucleotide enriched motif. In Group I but not Group II, we observed an enrichment of T nucleotide enriched HAND2 binding motif, which suggests that this special HAND2-DNA binding facilitates MYCN to bind to DNA sites of Group I.

Taken together, these results show that HAND2 cooperates with MYCN to increase chromatin accessibility of neuronal genes in neuroblastoma most probably by competing with nucleosomes to access and bind DNA.

HAND2 and MYCN cooperatively regulate gene transcription

To determine the effect of the loss of both HAND2 and MYCN on global gene transcriptional regulation, we performed RNA-seq analysis. The silencing of both HAND2 and MYCN in IMR32 cells led to greater changes in gene expression when compared with the silencing of either HAND2 or MYCN alone (Supplementary Fig. S5F; Fig. 5B left; Supplementary Table S8). Pathway analysis showed that the knockdown of both HAND2 and MYCN decreased transcription of the “mitotic signaling” pathway and activation of transcription in the “G2–M checkpoint regulation” (Fig. 5B, right). GSEA showed that the combined knockdown of both HAND2 and MYCN resulted in a negative enrichment of MYC signature genes, and a positive enrichment of neuronal markers (Fig. 5CF). Consistent with the dramatic epigenome and transcriptome changes resulting from the silencing of both HAND2 and MYCN expression, the loss of both TFs compared with loss of either one alone led to increased functional neuronal differentiation (Fig. 5G and H).

Targeting both MYCN and the CRC TFs is effective in suppressing neuroblastoma tumor growth

We found that MYCN and HAND2 cooperatively bind to DNA to regulate gene transcription and determine a malignant neuroblastoma phenotype, suggesting that targeting both MYCN and HAND2 simultaneously using small molecules will be more effective in suppressing neuroblastoma growth. Thus, we investigated the effect of pharmacologically targeting both MYCN and CRC TFs including HAND2. As a proof of concept, here we selected the AURKAi alisertib to target MYCN, and the HDACi LBH589 to target CRC TFs. AURKA interacts with both MYCN and the SCF ubiquitin ligase to stabilize MYCN protein, and the AURKAi treatment increases the degradation of MYCN in neuroblastoma cells (27, 28). Because SEs often drive oncogenes in cancers and CRC components are found to be essential in cancers, we chose a HDACi that was known to disrupt SE-driven CRCs (10, 29–32).

We found that the treatment of neuroblastoma cells with alisertib and LBH589 was more effective in reducing the viable cell number in MYCN-amplified cell lines compared with MYCN single-copy neuroblastoma cell lines (Fig. 6A; Supplementary Fig. S6A). IC50 assays clearly showed that MYCN-amplified cell lines were more sensitive to the LBH589 and alisertib, with IC50 values averaging greater than 4× lower than those found in MYCN single-copy cell lines (Supplementary Fig. S6B). These results indicated potent and selective effects of both HDACi and AUKRAi in MYCN-dysregulated cells. Importantly, the combination treatment using alisertib and LBH589 synergistically reduced viable neuroblastoma cell numbers with average Bliss synergy scores greater than 10 across a range of doses in IMR32, IMR5, or KCNR cells (Fig. 6B). Cell confluence assays confirmed the synergistic effects of alisertib + LBH589 treatment on cell proliferation in MYCN-amplified neuroblastoma cell lines (Fig. 6C). As expected, 6 hours of HDACi LBH589 treatment resulted in a decrease of the SE-driven CRC components HAND2, PHOX2B, and GATA3 at the mRNA levels in all the tested neuroblastoma cell lines (Supplementary Fig. S6C). As previously reported the inhibition of AURKA destabilized MYCN (27, 28), we found that a 48-hour alisertib treatment resulted in a 30%–50% decrease of MYCN at protein levels in neuroblastoma cells compared with control (Fig. 6D). Although the effect was TF and cell-line dependent, LBH589 treatment decreased both MYCN and CRC TF protein levels compared with the control, with decreases approaching 50% in most cell lines evaluated (Fig. 6D). The combination of alisertib and LBH589 treatment caused an approximately 80% decrease in MYCN and the CRC TF protein levels after 48 hours treatment in all the tested cell lines (Fig. 6D), indicating that this combined treatment was far more potent at inhibiting these targets than single-agent treatment.

Figure 6.

Targeting both MYCN and CRC TFs. A, Heatmaps show the percentage of cell viability after different dose of HDACi LBH589 and AURKAi alisertib treatment in neuroblastoma cell lines. Cell viability was measured by CellTiter-Glo Cell viability assay after 72 hours drug treatment. B, SynergyFinder online tool was used for Bliss synergistic analysis to evaluate the synergistic effect of the combination treatment in MYCN-amplified cell lines shown in A. C, IncuCyte cell confluence assays show the synergistic effect of the alisertib (Ali) + LBH589 (LBH) treatment on cell proliferation (% confluency) over time. Red arrow, time point of adding compounds. D, Western blot analysis shows the protein levels of MYCN and CRC TFs in neuroblastoma cells treated with LBH589 (LBH, 7.5 nmol/L for IMR32, 7.5 nmol/L for IMR5, and 5 nmol/L for KCNR), alisertib (Ali, 2 nmol/L for IMR32, 5 nmol/L for IMR5, and 5 nmol/L for KCNR) alone or in combination for 24 and 48 hours. E, Schematic diagram shows the strategy of drug treatment in orthotopic IMR5-GFP-Luc–implanted xenografts. IVIS, in vivo imaging system. F, Tumor weight measurement shows a significant decrease of the tumor weight of the drug treatment groups compared with the tumor weight of the vehicle treatment group. The P value indicated is calculated in one-way ANOVA. (E, Created with BioRender.com.)

Figure 6.

Targeting both MYCN and CRC TFs. A, Heatmaps show the percentage of cell viability after different dose of HDACi LBH589 and AURKAi alisertib treatment in neuroblastoma cell lines. Cell viability was measured by CellTiter-Glo Cell viability assay after 72 hours drug treatment. B, SynergyFinder online tool was used for Bliss synergistic analysis to evaluate the synergistic effect of the combination treatment in MYCN-amplified cell lines shown in A. C, IncuCyte cell confluence assays show the synergistic effect of the alisertib (Ali) + LBH589 (LBH) treatment on cell proliferation (% confluency) over time. Red arrow, time point of adding compounds. D, Western blot analysis shows the protein levels of MYCN and CRC TFs in neuroblastoma cells treated with LBH589 (LBH, 7.5 nmol/L for IMR32, 7.5 nmol/L for IMR5, and 5 nmol/L for KCNR), alisertib (Ali, 2 nmol/L for IMR32, 5 nmol/L for IMR5, and 5 nmol/L for KCNR) alone or in combination for 24 and 48 hours. E, Schematic diagram shows the strategy of drug treatment in orthotopic IMR5-GFP-Luc–implanted xenografts. IVIS, in vivo imaging system. F, Tumor weight measurement shows a significant decrease of the tumor weight of the drug treatment groups compared with the tumor weight of the vehicle treatment group. The P value indicated is calculated in one-way ANOVA. (E, Created with BioRender.com.)

Close modal

Next, we evaluated the activity of alisertib combined with LBH589 in preclinical neuroblastoma xenograft models. Our pilot study showed that the tumor incidence of IMR32 implanted athymic nude mice was 67.6% (23/34) with a tumor size variance of 147.4 ± 151.8 (mm3) after 2 months. Thus, the IMR32 cell line is not a reliable in vivo model for reproducibility issues. For this reason, we used IMR32 subclone IMR5 (33), as these two cell lines had a similar response to the treatment of alisertib plus LBH589 in vitro (Fig. 6AD). IMR5 cells stably expressing GFP-Luc were injected at the orthotopic site of the nude mouse. Animals were randomized according to their bioluminescence signal and received therapy with two treatment cycles of drug treatment (Fig. 6E; see Materials and Methods). Although protein levels of MYCN and CRC components in collected tumors were heterogeneous, the overall levels of these proteins were lower in the drug treatment groups compared with the vehicle treatment group (Supplementary Fig. S6D). Increased H3K27ac signal was observed in the groups of mice treated with either LBH589 alone or LBH589 + alisertib (Supplementary Fig. S6D). Animals across all groups did not exhibit any signs of overt toxicity (Supplementary Fig. S6E). Importantly, tumor weight (Fig. 6F) measurements revealed that the combination of alisertib and LBH589 treatment was superior to either single agent alone at suppressing tumor growth. Our results demonstrated that the combination therapy of alisertib + LBH589 was effective in MYCN-amplified neuroblastoma.

Combinatorial control of gene expression by a small group of TFs is critical to establish and maintain the cancer cell identity (13, 24, 34). In this study, we identify that the CRC TF HAND2 assists MYCN in enhancer binding, and that these two TFs compete with nucleosomes to cooperatively regulate global gene expression. By targeting both MYCN and the neuroblastoma CRC, we find that a combination of AURKAi and HDACi treatment of neuroblastoma cells more potently reduces the protein levels of both MYCN and CRC TFs than the treatment of neuroblastoma cells with each drug alone, which results in a more effective inhibition of neuroblastoma tumor growth (Fig. 7).

Figure 7.

The cooperation between HAND2 and MYCN makes them ideal therapeutic targets. (Created with BioRender.com.)

Figure 7.

The cooperation between HAND2 and MYCN makes them ideal therapeutic targets. (Created with BioRender.com.)

Close modal

Our study shows that HAND2 selectively influences MYCN-enhancer binding as the loss of HAND2 results in a decrease in genome bound MYCN while the overexpression of HAND2 leads to an increase in genome bound MYCN. HAND2, PHOX2B, and GATA3 are highly expressed in neuroblastoma and form a CRC to cooperate with MYCN to determine a noradrenergic neuroblastoma phenotype (10). Here we find that the loss of HAND2 but not PHOX2B or GATA3 decreases MYCN enhancer binding at the genome-wide level. We find that HAND2 assists MYCN in binding to a majority of MYCN-bound enhancers, but other TFs might be required for MYCN to bind to the other genomic loci. CRISPR-Cas9 screen of neuroblastoma cell lines indicates that HAND2 is essential in both MYCN single copy and MYCN-amplified neuroblastoma cell lines (Supplementary Fig. S1A). One reason is that MYCN protein expression might be high in some of the neuroblastoma cell lines without MYCN amplification (35). Another explanation is that HAND2 has MYCN-independent function as greater than 50% of HAND2 regulated genes are not regulated by MYCN (Fig. 3B and D). Nevertheless, the simultaneous silencing of both HAND2 and MYCN in neuroblastoma cells results in greater changes in gene expression when compared with the silencing of either HAND2 or MYCN alone. This further demonstrates that these two TFs cooperatively regulate global gene expression.

Our results suggest that HAND2 facilitates MYCN enhancer invasion via a “cooperative” TF-DNA binding model. In this TF-DNA binding model, TFs recognize adjacent binding sites within regulatory regions to compete with nucleosomes to gain access and bind DNA simultaneously or sequentially (22–24, 36, 37). Our ATAC-seq results show that the silencing of both HAND2 and MYCN results in a more dramatic decrease of chromatin accessibility than the silencing of either TF alone. These results indicate that HAND2 and MYCN cooperate to compete with nucleosomes and achieve specific DNA binding. This cooperativity may arise with the assistance of ATP-dependent nucleosome remodeling factors. For example, initially bound TF recruits a remodeler to disassemble or move a nucleosome to facilitate binding of a second TF toward its binding site (20). We did not identify chromatin remodelers that are regulated by HAND2 at the mRNA level based on RNA-seq data analysis. We have tested BRD4, which is known to enhance chromatin accessibility (38), but find that it does not mediate HAND2 function at regulating chromatin accessibility. Additional chromatin remodelers will be evaluated to determine how the loss of HAND2 leads to changes in chromatin accessibility in our future studies.

SEs often drive the expression of oncogenes and CRC components in cancers (10, 31) and the Bromo-domain inhibitor JQ1 causes transcriptional repression of SE-associated oncogenes (39). Recent studies showed that the combination of JQ1 and CDK7 inhibitor THZ1 reduced the expression of MYCN and CRC components and synergistically suppressed neuroblastoma growth (10, 11). Here we take a different approach to target these TFs by using clinically-relevant AURKA and HDACis (40, 41). AURKAi treatment increases degradation of MYCN in neuroblastoma cells (27, 28). HDACi treatment of cancer cells spreads hyperacetylated histones and disrupts the three-dimensional organization of SEs, thereby destabilizing CRC TFs and RNA Pol II binding at SEs and selectively suppresses CRC TF transcription (29, 30, 32). As expected, we find that the combination of AURKAi and HDACi treatment of neuroblastoma cells reduces expression of both MYCN and CRC TFs and synergistically suppresses neuroblastoma cell proliferation in MYCN-amplified cells.

The finding that the CRC component HAND2 assists oncogenic protein MYCN to bind to DNA enhancer regions adds another layer to our understanding of how CRC TFs form a combinatorial transcriptional regulatory circuit to determine cell identity in this disease state. Our finding provides a rationale for targeting multiple TFs simultaneously to increase the disruption of this “combinatorial transcription regulatory network” in cancer cells and improve therapeutic efficacy.

No disclosures were reported.

M. Xu: Conceptualization, resources, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. M. Sun: Investigation, methodology, writing–review and editing. X. Zhang: Formal analysis, investigation, methodology, writing–review and editing. R. Nguyen: Investigation, methodology, writing–review and editing. H. Lei: Data curation, investigation, writing–review and editing. J.F. Shern: Investigation, writing–review and editing. C.J. Thiele: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing. Z. Liu: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This work was funded by the Center for Cancer Research, Intramural Research Program at the NCI. The authors thank Bao Tran, Jyoti Shetty, and Yongmei Zhao from NCI Sequencing Facility for DNA and RNA sequencing. This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Lambert
SA
,
Jolma
A
,
Campitelli
LF
,
Das
PK
,
Yin
Y
,
Albu
M
, et al
.
The human transcription factors
.
Cell
2018
;
172
:
650
65
.
2.
Meyer
N
,
Penn
LZ
.
Reflecting on 25 years with MYC
.
Nat Rev Cancer
2008
;
8
:
976
90
.
3.
Sabo
A
,
Kress
TR
,
Pelizzola
M
,
de Pretis
S
,
Gorski
MM
,
Tesi
A
, et al
.
Selective transcriptional regulation by Myc in cellular growth control and lymphomagenesis
.
Nature
2014
;
511
:
488
92
.
4.
Lin
CY
,
Loven
J
,
Rahl
PB
,
Paranal
RM
,
Burge
CB
,
Bradner
JE
, et al
.
Transcriptional amplification in tumor cells with elevated c-Myc
.
Cell
2012
;
151
:
56
67
.
5.
Zeid
R
,
Lawlor
MA
,
Poon
E
,
Reyes
JM
,
Fulciniti
M
,
Lopez
MA
, et al
.
Enhancer invasion shapes MYCN-dependent transcriptional amplification in neuroblastoma
.
Nat Genet
2018
;
50
:
515
23
.
6.
Matthay
KK
,
Maris
JM
,
Schleiermacher
G
,
Nakagawara
A
,
Mackall
CL
,
Diller
L
, et al
.
Neuroblastoma
.
Nat Rev Dis Primers
2016
;
2
:
16078
.
7.
Cheung
NK
,
Dyer
MA
.
Neuroblastoma: developmental biology, cancer genomics and immunotherapy
.
Nat Rev Cancer
2013
;
13
:
397
411
.
8.
Boeva
V
,
Louis-Brennetot
C
,
Peltier
A
,
Durand
S
,
Pierre-Eugene
C
,
Raynal
V
, et al
.
Heterogeneity of neuroblastoma cell identity defined by transcriptional circuitries
.
Nat Genet
2017
;
49
:
1408
13
.
9.
van Groningen
T
,
Koster
J
,
Valentijn
LJ
,
Zwijnenburg
DA
,
Akogul
N
,
Hasselt
NE
, et al
.
Neuroblastoma is composed of two super-enhancer-associated differentiation states
.
Nat Genet
2017
;
49
:
1261
6
.
10.
Durbin
AD
,
Zimmerman
MW
,
Dharia
NV
,
Abraham
BJ
,
Iniguez
AB
,
Weichert-Leahey
N
, et al
.
Selective gene dependencies in MYCN-amplified neuroblastoma include the core transcriptional regulatory circuitry
.
Nat Genet
2018
;
50
:
1240
6
.
11.
Decaesteker
B
,
Denecker
G
,
Van Neste
C
,
Dolman
EM
,
Van Loocke
W
,
Gartlgruber
M
, et al
.
TBX2 is a neuroblastoma core regulatory circuitry component enhancing MYCN/FOXM1 reactivation of DREAM targets
.
Nat Commun
2018
;
9
:
4866
.
12.
Wang
L
,
Tan
TK
,
Durbin
AD
,
Zimmerman
MW
,
Abraham
BJ
,
Tan
SH
, et al
.
ASCL1 is a MYCN- and LMO1-dependent member of the adrenergic neuroblastoma core regulatory circuitry
.
Nat Commun
2019
;
10
:
5622
.
13.
Bradner
JE
,
Hnisz
D
,
Young
RA
.
Transcriptional addiction in cancer
.
Cell
2017
;
168
:
629
43
.
14.
Liu
Z
,
Yang
X
,
Tan
F
,
Cullion
K
,
Thiele
CJ
.
Molecular cloning and characterization of human Castor, a novel human gene upregulated during cell differentiation
.
Biochem Biophys Res Commun
2006
;
344
:
834
44
.
15.
Liu
Z
,
Lam
N
,
Wang
E
,
Virden
RA
,
Pawel
B
,
Attiyeh
EF
, et al
.
Identification of CASZ1 NES reveals potential mechanisms for loss of CASZ1 tumor suppressor activity in neuroblastoma
.
Oncogene
2017
;
36
:
97
109
.
16.
Liu
Z
,
Zhang
X
,
Lei
H
,
Lam
N
,
Carter
S
,
Yockey
O
, et al
.
CASZ1 induces skeletal muscle and rhabdomyosarcoma differentiation through a feed-forward loop with MYOD and MYOG
.
Nat Commun
2020
;
11
:
911
.
17.
Ramirez
F
,
Ryan
DP
,
Gruning
B
,
Bhardwaj
V
,
Kilpert
F
,
Richter
AS
, et al
.
deepTools2: a next generation web server for deep-sequencing data analysis
.
Nucleic Acids Res
2016
;
44
:
W160
5
.
18.
Orlando
DA
,
Chen
MW
,
Brown
VE
,
Solanki
S
,
Choi
YJ
,
Olson
ER
, et al
.
Quantitative ChIP-seq normalization reveals global modulation of the epigenome
.
Cell Rep
2014
;
9
:
1163
70
.
19.
Khan
A
,
Mathelier
A
.
Intervene: a tool for intersection and visualization of multiple gene or genomic region sets
.
BMC Bioinformatics
2017
;
18
:
287
.
20.
Heinz
S
,
Benner
C
,
Spann
N
,
Bertolino
E
,
Lin
YC
,
Laslo
P
, et al
.
Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities
.
Mol Cell
2010
;
38
:
576
89
.
21.
McLean
CY
,
Bristor
D
,
Hiller
M
,
Clarke
SL
,
Schaar
BT
,
Lowe
CB
, et al
.
GREAT improves functional interpretation of cis-regulatory regions
.
Nat Biotechnol
2010
;
28
:
495
501
.
22.
Moyle-Heyrman
G
,
Tims
HS
,
Widom
J
.
Structural constraints in collaborative competition of transcription factors against the nucleosome
.
J Mol Biol
2011
;
412
:
634
46
.
23.
Mirny
LA
.
Nucleosome-mediated cooperativity between transcription factors
.
Proc Natl Acad Sci U S A
2010
;
107
:
22534
9
.
24.
Reiter
F
,
Wienerroither
S
,
Stark
A
.
Combinatorial function of transcription factors and cofactors
.
Curr Opin Genet Dev
2017
;
43
:
73
81
.
25.
George
RM
,
Firulli
AB
.
Hand factors in cardiac development
.
Anat Rec
2019
;
302
:
101
7
.
26.
Harmelink
C
,
Peng
Y
,
DeBenedittis
P
,
Chen
H
,
Shou
W
,
Jiao
K
.
Myocardial Mycn is essential for mouse ventricular wall morphogenesis
.
Dev Biol
2013
;
373
:
53
63
.
27.
Brockmann
M
,
Poon
E
,
Berry
T
,
Carstensen
A
,
Deubzer
HE
,
Rycak
L
, et al
.
Small molecule inhibitors of aurora-a induce proteasomal degradation of N-myc in childhood neuroblastoma
.
Cancer Cell
2013
;
24
:
75
89
.
28.
Otto
T
,
Horn
S
,
Brockmann
M
,
Eilers
U
,
Schuttrumpf
L
,
Popov
N
, et al
.
Stabilization of N-Myc is a critical function of Aurora A in human neuroblastoma
.
Cancer Cell
2009
;
15
:
67
78
.
29.
Sanchez
GJ
,
Richmond
PA
,
Bunker
EN
,
Karman
SS
,
Azofeifa
J
,
Garnett
AT
, et al
.
Genome-wide dose-dependent inhibition of histone deacetylases studies reveal their roles in enhancer remodeling and suppression of oncogenic super-enhancers
.
Nucleic Acids Res
2018
;
46
:
1756
76
.
30.
Gryder
BE
,
Wu
L
,
Woldemichael
GM
,
Pomella
S
,
Quinn
TR
,
Park
PMC
, et al
.
Chemical genomics reveals histone deacetylases are required for core regulatory transcription
.
Nat Commun
2019
;
10
:
3004
.
31.
Hnisz
D
,
Abraham
BJ
,
Lee
TI
,
Lau
A
,
Saint-Andre
V
,
Sigova
AA
, et al
.
Super-enhancers in the control of cell identity and disease
.
Cell
2013
;
155
:
934
47
.
32.
Gryder
BE
,
Pomella
S
,
Sayers
C
,
Wu
XS
,
Song
Y
,
Chiarella
AM
, et al
.
Histone hyperacetylation disrupts core gene regulatory architecture in rhabdomyosarcoma
.
Nat Genet
2019
;
51
:
1714
22
.
33.
Bukovsky
J
,
Evans
A
,
Tartaglione
M
,
Kennett
RH
.
Selection of variant neuroblastoma cell line which has lost cell surface expression of antigen detected by monoclonal antibody PI153/3
.
Somat Cell Mol Genet
1985
;
11
:
517
22
.
34.
Bushweller
JH
.
Targeting transcription factors in cancer — from undruggable to reality
.
Nat Rev Cancer
2019
;
19
:
611
24
.
35.
Valentijn
LJ
,
Koster
J
,
Haneveld
F
,
Aissa
RA
,
van Sluis
P
,
Broekmans
ME
, et al
.
Functional MYCN signature predicts outcome of neuroblastoma irrespective of MYCN amplification
.
Proc Natl Acad Sci U S A
2012
;
109
:
19190
5
.
36.
Vashee
S
,
Melcher
K
,
Ding
WV
,
Johnston
SA
,
Kodadek
T
.
Evidence for two modes of cooperative DNA binding in vivo that do not involve direct protein-protein interactions
.
Curr Biol
1998
;
8
:
452
8
.
37.
Miller
JA
,
Widom
J
.
Collaborative competition mechanism for gene activation in vivo
.
Mol Cell Biol
2003
;
23
:
1623
32
.
38.
Devaiah
BN
,
Case-Borden
C
,
Gegonne
A
,
Hsu
CH
,
Chen
Q
,
Meerzaman
D
, et al
.
BRD4 is a histone acetyltransferase that evicts nucleosomes from chromatin
.
Nat Struct Mol Biol
2016
;
23
:
540
8
.
39.
Puissant
A
,
Frumm
SM
,
Alexe
G
,
Bassil
CF
,
Qi
J
,
Chanthery
YH
, et al
.
Targeting MYCN in neuroblastoma by BET bromodomain inhibition
.
Cancer Discov
2013
;
3
:
308
23
.
40.
Bavetsias
V
,
Linardopoulos
S
.
Aurora kinase inhibitors: current status and outlook
.
Front Oncol
2015
;
5
:
278
.
41.
Suraweera
A
,
O'Byrne
KJ
,
Richard
DJ
.
Combination therapy with histone deacetylase inhibitors (HDACi) for the treatment of cancer: achieving the full therapeutic potential of HDACi
.
Front Oncol
2018
;
8
:
92
.