KDM3B Suppresses APL Progression by Restricting Chromatin Accessibility

Wang et al. Cancer Cell Int (2019) 19:256 https://doi.org/10.1186/s12935-019-0979-7 Cancer Cell International

PRIMARY RESEARCH Open Access KDM3B suppresses APL progression by restricting chromatin accessibility and facilitating the ATRA‑mediated degradation of PML/RARα Xinrui Wang1, Huiyong Fan1, Congling Xu1, Guojuan Jiang1, Haiwei Wang2* and Ji Zhang1*

Abstract Background: A hallmark of acute promyelocytic leukemia (APL) is the expression of PML/RARα fusion protein. Treat- ment with all-trans retinoic acid (ATRA) results in the terminal diferentiation of neutrophil granulocytes. However, the underlying mechanisms remain largely unknown. Here, we identify and elucidate a novel diferentiation-suppressive model of APL involving the histone demethylase KDM3B, which has been identifed as a suppressor of the tumor genes involved in hematopoietic malignancies. Methods: First, we established a KDM3B knockdown NB4 cell model to determine the functional characteristics of KDM3B by cell proliferation assay and fow cytometry. Then, we performed ChIP-seq and ATAC-seq to search for potential relationships among KDM3B, histone modifcation (H3K9me1/me2) and the chromatin state. Finally, molecular biological techniques and a multi-omics analysis were used to explore the role of KDM3B in diferentiation of the leukemia cells after ATRA treatment. Results: We found that knocking down KDM3B contributed to the growth of NB4 APL cells via the promotion of cell- cycle progression and blocked granulocytic diferentiation. Through global and molecular approaches, we provided futher evidence that knocking down KDM3B altered the global distribution of H3K9me1/me2 and increased the chromatin accessibility. Moreover, knocking down KDM3B inhibited the ATRA-induced degradation of the PML/RARα oncoprotein. Conclusion: Our study suggested that KDM3B was able to inhibit APL progression by maintaining chromatin in a compact state and facilitating the ATRA-mediated degradation of PML/RARα. Taken together, the results show that KDM3B may be an alternative target for the treatment regimens and the targeted therapy for APL by sustaining the function of PML/RARα fusion protein. Keywords: KDM3B, APL, PML/RARα, H3K9me1/me2, Chromatin accessibility, Diferentiation

*Correspondence: [email protected]; [email protected] 1 State Key Laboratory of Medical Genomics, Shanghai Institute of Hematology, Rui‑Jin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China 2 Institute of Health Sciences, Shanghai Institutes for Biological Sciences and Graduate School, Chinese Academy of Sciences, Shanghai 200025, China

© The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Wang et al. Cancer Cell Int (2019) 19:256 Page 2 of 15

Background Tus far, the epigenetic mechanisms of H3K9 methyl- Te epigenetic information encoded in chromatin is transferases or demethylases have been implicated in the crucial for cell fate and diferentiation [1–3]. Post-trans- pathogenesis of hematopoietic malignancies, as deter- lational modifcations in the histone tail are important mined with the information acquired through next-gen- epigenetic marks linked to transcriptional activation or eration sequencing (NGS) [8, 27]. Namely, to thoroughly repression of their specifc target genes [4, 5]. One major analyze, for the frst time, the regulatory mechanisms of type of chromatin-level regulation is histone methyla- KDM3B that are involved in the ATRA-induced difer- tion. In particular, histone lysine residues can be modi- entiation of NB4 cells, we performed a ChIP assay using fed by mono-methylation (me1), dimethylation (me2), or antibodies that recognize H3K9me1/me2 and an ATAC tri-methylation (me3) [6]. Histone lysine methylation is assay. In the present study, we demonstrate that knock- linked to chromatin remodeling and its activity depends ing down KDM3B leads to an increase in H3K9me1 on the specifc lysine targeted [7, 8]. Te methylation of levels and enhanced chromatin accessibility in transcrip- histone H3 lysine 9 (H3K9) is one of the most intensively tion start site (TSS) regions of the global genome, which studied histone modifcations because it is considered an includes the PML/RARα target genes, whereas the levels epigenetic marker of transcriptionally silenced hetero- of modifed H3K9me2 are reduced in these regions. With chromatin [4, 9, 10]. the use of a series of experimental assays and using leu- Histone lysine methylation has been proven to be kemia cell lines (NB4 and PR9), we proved that KDM3B reversible and many histone methylation modifed inhibits the survival of APL NB4 cells and it is required enzymes have been identifed [4, 11]. In the develop- for the diferentiation of leukemia cells after ATRA mental process, the status of H3K9 methylation is tightly treatment. controlled by the dynamic alternation of lysine methyltransferases and demethylases [12–14]. Among these Materials and methods enzymes, KDM3 family members contain a Jumonji C Cell culture (JmjC) domain and possess intrinsic H3K9 demethylat- PR9 and NB4 cells were grown in RPMI-1640 medium ing activity [15]. Te KDM3 family has three important (GibcoBRL, Gaithersburg, MD, USA), supplemented with members, namely, KDM3A (also known as JMJD1A, 10% FBS (GibcoBRL, Gaithersburg, MD, USA). Human JHDM2A or TSGA), KDM3B (also known as JMJD1B, 293T cells were grown in DMEM (GibcoBRL, Gaithers- JHDM2B or 5qNCA) and JMJD1C (also known as burg, MD, USA) supplemented with 10% fetal bovine JHDM2C or TRIP8) [16–18]. serum. All cells were maintained at 37 °C in a humidifed KDM3B is a H3K9me1/me2-specifc demethylase that atmosphere containing 5% CO2. For induction of granu- was initially suspected of inhibiting the tumor activity locytic diferentiation, cells were incubated with 1 mM of hematopoietic malignancies due to its location in the ATRA (Sigma, St. Louis, MO, USA). For induction of 5q31 chromosome region, which is a frequently deleted PML/RARa expression, PR9 cells were incubated with region in myelodysplastic syndromes (MDS) and acute 100 mM ZnSO4 (Sigma, St. Louis, MO, USA) for 4 h. myeloid leukemia (AML) [16, 19–21]. Although, some results have suggested that KDM3B mediates H3K9 shRNA‑induced gene silencing by lentiviral infection methylation, which plays important roles in AML devel- Lentiviral-based vectors for RNA interference-medi- opment, progression and prognosis, due to the lack of ated gene silencing were pLVX-shRNA2-Puro, contain- global information on KDM3B and H3K9 methylation ing shRNAs scrambled (against the sequence 5ʹ-GCC status of chromatin, the mechanisms of H3K9 methyla- ACAACG TCT ATA TCA TGG-3 ʹ), of KDM3B (against tion mediated by KDM3B in hematopoietic malignance the sequence 5ʹ-GCGATCTTTGTAGAATTTGAT-3ʹ development are not known [22–24]. and 5ʹ-GCACTGAGAGAAACAGTTAAT-3ʹ). Lentiviral Previously, our lab focused on the diferentiation pro- particles were produced in HEK293T (human embry- cess in ATRA-mediated acute promyelocytic leukemia onic kidney) cells for 72 h upon transfection of the (APL) and identifed a critical blocked diferentiation shRNA vectors together with psPAX2 and pMD2G plas- mechanism in APL [25, 26]. Our results suggested that mids using polyenthyleneimine (Sigma). Medium was oncogenic fusion protein PML/RARα is bound to and then centrifuged (1000×g, 5 min) and fltered through represses the activation of the PU.1-mediated gene signa- a 0.45 mm membrane (Millipore). Cells were incubated ture which is essential for myeloid diferentiation. Taking with medium containing lentiviral particles for 24 h. advantage of our previous extensive studies on APL, we Ten, medium was replaced to allow the knockdown for chose to study the epigenetic alterations that are induced 2 or more days. pLVX-shRNA2-Puro vectors contain a by KDM3B-mediated H3K9 methylation during APL puromycin resistance site. Puromycin (4 mg/mL; Sigma) pathogenesis. was added to the media to select for resistant cells and Wang et al. Cancer Cell Int (2019) 19:256 Page 3 of 15

to generate stable target-defcient cell lines and Polybrene ChIP‑Seq and bioinformatics analysis (1.75 mg/mL; Sigma) to enhance viral infection. ChIP assay was performed using the Active Motif ChIP- IT kit (Carlsbad, CA, USA) according to manufacturer’s RNA isolation and quantitative PCR (qPCR) instructions. Briefy, cells were lysed and sonicated after Total RNA was isolated from the collected cells using cross-linking by 1% formaldehyde for 10 min at room RNeasy Mini-kit (Qiagen, Hilden, Germany) and reverse temperature. Te crosslinked complex was immuno- transcribed into cDNA that prepared by PrimeScript™ precipitated by target antibody or control rabbit IgG RT reagent Kit (TaKaRa, Tokyo, Japan) as the manufac- (Cell Signaling Technology, Beverly, MA, USA) bound turer’s directions. Relative RNA expression was normal- to protein A/G agarose beads. After overnight incu- ized to β-actin using the 2−ΔΔCt method [28]. QPCR was bation at 4 °C, the complex was eluted and DNA was conducted as follows: 95 °C for 10 min; and 40 cycles of purifed. ChIP products were submitted to Shanghai 95 °C for 15 s and 60 °C for 1 min. QPCR were performed Personal Biotechnology. Libraries were created from on the ABI ViiA™ 7 Real-Time PCR system (Applied Bio- successful large-scaled ChIP experiments using Illumi- systems) using SYBE‑Green qPCR Supermix (TOYOBO, na’s TruSeq Library Prep Kit following instructions in Tokyo, Japan). Gene primer sequences were shown in the user manual (Illumina, San Diego, CA, USA). Te Additional fle 1. library size was estimated using a Bioanalyzer 2100. Pooled libraries were submitted for sequencing on Protein extraction and immunoblots the Illumina HiSeq X ten system at 150 bp pair-ended Nuclear protein was extracted by using the NE-PER sequencing. ChIP-Seq reads were quality controlled Nuclear and Cytoplasmic Extraction Reagents using the and trimmed for adapter sequences using Trim Galore. manufacturer’s instructions. For total protein extrac- Filtered reads were aligned to hg38 using Bowtie2. tion, the cells were collected and lysed in Lysis bufer Sam fles from Bowtie were converted into bam fles by (20 mM Tris/HCl pH 7.5; 120 mM NaCl; 1 mM EDTA; SAMtools. Te profle and correlation plots were made 1 mM EGTA; 1% Triton X-100; 2.5 mM Sodium pyroph- with DeepTools. Visualization of read count data was performed by converting raw bam fles to bigwig fles osphate; 1 mM β Glycerophosphate; 1 mM Na3VO4) in the presence of protease inhibitors (Sigma-Aldrich). Indi- using IGV tools. 3 kb up- and downstream regions of vidual samples were denatured and loaded onto 8–12% the TSS were used for TSS-based profle plots and 3 kb Tris Acrylamide gels and electrotransferred onto PVDF up- and downstream regions PML-RARα peaks were membranes (Millipore). After blocking with 5% skimmed used for PML-RARα-based profle plots. milk in TBST for 2 h at room temperature, membranes were incubated at 4 °C overnight with specifc primary ATAC‑Seq and bioinformatics analysis antibodies with details shown in the supplement table. Subsequent to washing, membranes were incubated with ATAC-seq was performed as described on harvested horseradish peroxidase-conjugated (HRP) secondary cell samples [29, 30]. Libraries were purifed with antibodies for 2 h. Signals were developed by ECL rea- AMPure beads (Agencourt) to remove contaminating gent (Millipore) and captured by FluorChem M (Protein- primer dimers. Te library size was estimated using Simple). Histone 3 and β-actin were used as markers of a Bioanalyzer 2100. Pooled libraries were submitted the nucleus and whole cell lysate, respectively. Details of for sequencing on the Illumina HiSeq X ten system at antibodies were shown in the Additional fle 1. 150 bp pair-ended sequencing. All reads for each sample were combined and aligned to hg38 with bowtie2 Immunofuorescence [18]. Peaks were called for each sample using HOMER Cells were fxed in 4% paraformaldehyde, then washed and individual peaks separated by < 500 bp were joined with phosphate-bufered saline (PBS). Dispersed cells together. Te peak annotation was assessed using the were attached to positively-charged slides using a Cyto- ‘ChIPseeker’ library from R/Bioconductor. Motif anal- Spin and incubated in blocking solution (PBS with 0.3% ysis on peak regions was performed by HOMER func- Triton and 10% fetal bovine serum FBS) [19]. Primary tion. Heatmaps and profle plots were generated using antibodies were incubated overnight at 4 °C, and second- DeepTools. Visualization of genome-wide read cover- ary antibodies were incubated at room temperature for age was performed by converting raw bam fles to big- 3 h. DAPI (4′,6-diamidino-2-phenylindole) was added wig fles using IGV tools. 0.5 kb up- and downstream with the mounting media to counterstain DNA. Pictures regions of the TSS were used for TSS-based heatmaps. of random felds were taken and then analyzed on a Leica 0.5 kb up- and downstream regions PML-RARα peaks TCS SP8 confocal microscope. Details of antibodies were were used for PML-RARα-based heatmaps. shown in the supplement Additional fle 1. Wang et al. Cancer Cell Int (2019) 19:256 Page 4 of 15

(See fgure on next page.) Fig. 1 KDM3B is required for the growth and granulocytic diferentiation of NB4 cells. a, b The mRNA expression levels of KMD3B across various types of human cancers. Data were retrieved from CCLE dataset (a), which contains RNA-seq of 1457 human cancer cell lines, and the TCGA dataset (b). Arrow, leukemia and lymphoma cells. c KDM3B mRNA levels in AML patients and normal counterparts were compared (raw data retrieved from from GSE9476 and GSE7186. d Overall survival of AML patient cohort with regard to KDM3B mRNA levels. e Efects of stable KDM3B-knockdown in APL NB4 cells by Western blot analysis. f Knock-down of KDM3B promotes NB4 cell proliferation. CCK8 assay was performed to determine the cell growth of NB4 cells transfected with scramble control or shKDM3B. g Cell cycle was determined by fow cytometry analysis. The bar chart represented the percentage of cells in G0/G1, S, or G2/M phases respectively. h Flow cytometric analysis of CD11b cell-surface expression in NB4 cells transfected with scramble control or shKDM3B followed by treatment with or without ATRA. Values are derived from three independent experiments, and data were mean SD. *P < 0.05, **P < 0.01, ***P < 0.001 ±

RNA‑Seq and bioinformatics analysis A and 50 g/mL propidium iodide (PI, Sigma-Aldrich, St. Total RNA prepared was extracted using TRIzol Rea- Louis, MO, USA) for 30 min in the dark at 4 °C. Te fow gent (Invitrogen, Carlsbad, CA, USA) and submitted to cytometric data were collected on a BD LSRFortessa fow Shanghai Personal Biotechnology, where RNA intergrity cytometer (BD Biosciences, San Jose, CA, USA) and ana- was confrmed using the Illumina HiSeq X ten system at lyzed using FlowJo software or ModFit LT3.0 software. 150 bp pair-ended. Double-strand cDNA libraries were prepared and constructed using the TruSeq RNA Sam- Statistical data ple Prep Kit (Illumina, San Diego, CA). Two replicates Data are presented as mean SD. A 2-tailed Student of the RNA-Seq experiments were performed. RNA-Seq ± t-test was performed for all compared data by using reads were quality controlled and trimmed for adapter GraphPad Prism version 7. A P < 0.05 was considered to sequences using Trim Galore. Filtered reads were aligned be statistically signifcant (*), a P < 0.01 was regarded as to hg38 using HISAT2. Read counts for each gene were statistically very signifcant (**). carried out using HT-Seq using the hg38 refSeq refFlat GTF fle accessed on July 2015. Diferentially expressed genes (DEGs) were analysed using the DESeq2 package Results KDM3B is highly expressed in hematopoietic malignance (|fold change| ≥ 1.5, P < 0.05). and correlated with the favorable prognosis of AML patient Gene‑set enrichment analysis and gene ontology analysis First, the expression of KDM3B across tumor types GSEA was performed using the Broad Institute web plat- were tested. By analyzing the expression of the Broad form by pre-ranking the RNA-seq list based on log2-fold Institute’s CCLE cell lines, we found that KDM3B was change. Gene ontology analysis and KEGG pathways signifcantly highly expressed in human leukemia and analysis were performed with Gene Ontology Resource lymphoma cell lines compared to other tumor cell lines (http://geneontology.org/) and R package “clusterPro- (Fig. 1a). Tis observation was further confrmed in fler” to assess the biological functions of this subset of tumor patients from Te Cancer Genome Atlas (TCGA) DEGs [31]. dataset. Compared to other tumor type’s patients, AML patients had the highest expression of KDM3B (Fig. 1b). Cell proliferation assay and fow cytometry analysis Tose results emphasized the signifcance of KDM3B in Cell proliferation was measured by CCK-8 (Dojindo Lab- hematopoietic malignance. oratories, Kumamoto, Japan) following the manufactur- Ten, we determined the expression of KDM3B in er’s instructions. Te optical absorbance was measured at normal and malignant hematopoietic cells. Te data 450 nm by a plate reader (BioTek Instruments, Inc., Win- of KDM3B expression in two cohorts of AML patients ooski, VT, USA). were downloaded from GEO datasets with GEO number Antibodies PE-conjugated CD11b were purchased GSE9476 and GSE7186. An abnormal KDM3B expres- from BD PharMingen (CA, USA). Total cells were sion were observed in AML cells. Compared to normal Fcblocked and stained with indicated combinations of cells, KDM3B mRNA level were signifcantly reduced in antibodies for 30 min on ice, then washed three times AML blasts cells (Fig. 1c). Importantly, using the TCGA and resuspended in 1% FBS/PBS. For cell cycle assays, data, we found that low expression of KDM3B was an total cells were harvested and washed three times in PBS unfavorable prognosis in AML patients (P = 0.0058) (pH 7.4), followed by fxation with precooled 70% etha- (Fig. 1d). Tese results were consistent with the observa- nol overnight at 4 °C. Ethanol-fxed cells were centrifuged tion of loss of KDM3B chromosome region in AML and at 1000 rpm for 5 min, washed twice with PBS, and then highlighted the inhibitory roles of KDM3B in hematopoi- incubated with 0.5 mL PBS containing 10 g/mL RNase etic malignancies. Wang et al. Cancer Cell Int (2019) 19:256 Page 5 of 15

abKDM3B RNA expression level in various cancer cell lines (RMA, log2) KDM3B RNA expression in various cancers (TPM) giant_cell_tumour LIHC chondrosarcoma mesothelioma DLBC melanoma HNSC osteosarcoma BLCA meningioma thyroid CESC kidney LUSC glioma UCS upper_aerodigestive other OV urinary_tract UCEC Ewings_sarcoma GBM lung_NSC bile_duct SKCM lymphoma_Hodgkin READ NA KIRP breast liver LUAD soft_tissue KICH pancreas COAD ovary multiple_myeloma PRAD prostate ACC esophagus STAD stomach endometrium KIRC T−cell_lymphoma_other THCA colorectal BRCA medulloblastoma B−cell_lymphoma_other LGG lung_small_cell AML AML 050 100 150 lymphoma_DLBCL neuroblastoma lymphoma_Burkitt c 10 GSE9476 1.0 GSE7186 CML *** l leukemia_other Leukemia & lymphoma l 0.5 B−cell_ALL 8 *** T−cell_ALL KDM3 B KDM3 B . 23 45 0 0 6 mRNA leve d 1.00 mRNA leve -0.5 + Low Expression Normalized ++ Normalized + ++ + High Expression 4 -1.0 y Log−rank AMLNormal AMLNormal 0.75 + + + + p-value = 0.0058 + e f 1.5

probabilit + 0.50 Scamble * shKDM3B-1 * + ++++++ ++ m 1.0 shKDM3B-2

Survival ++ +++ + 0.25 + ScrambleshKDM3B-shKDM3B-1 2 ++ ++ + KDM3B OD 450n 0.5 0.00 0255075 100 β-actin Overall Survival Time (Months) 0 g 0 1 2 3 (days) 80 1000 1000 Scramble 1200 * ) shKDM3B-1 80 0 80 0 * 90 0 60 shKDM3B-2 00 00

600 * 40 06 40 06 Cell Count 40 * 300 200 200 0 0 0 20 0 40 80 120 140 200 0 40 80 120140 200 0 40 80 120 140 200 Cell population (% Scamble shKDM3B-1 shKDM3B-2 Propidium Iodide 0 G0/G1 SG2/M h Scamble shKDM3B-1 shKDM3B-2 * * Control 80 Scramble 60 shKDM3B-1

1b shKDM3B-2 40 %CD1

20 ATRA (1μM )

0 66.0% 44.6% 46.9% 0 ControlATRA (1μM ) CD11b Wang et al. Cancer Cell Int (2019) 19:256 Page 6 of 15

(See fgure on next page.) Fig. 2 KDM3B is essential to regulate H3K9me1/me2 histone modifcations and chromatin accessibility. a Correlation heatmap of genome-wide enrichment for H3K9me1, H3K9me2 or H3K9me3 ChIP-seq data in NB4 cells expressing scrambled control versus shKDM3B, organized and ordered using hierarchical clustering, is shown. Each modifcation type was then scaled to 0–1 and the R coefcient was determined with Spearman’s correlation. Colors indicate no correlation (red), intermediate correlation (yellow), and strong correlation (blue). b Plots showing the genome-wide correlation from H3K9me1, H3K9me2 or H3K9me3 ChIP-seq in NB4 cells expressing scrambled control versus shKDM3B, respectively. c Signal intensity plot representing changes in H3K9me1, H3K9me2 and H3K9me3 ChIP-seq signal at promoters regions or gene body following silencing of KDM3B in NB4 cells. The enriched regions were extended 3 kb from their midpoint. d Genome browser tracks of H3K9me1, H3K9me2 and ± H3K9me3 ChIP-seq data at the C11orf95, CBX4 and CDK2AP1 loci following silencing of KDM3B in NB4 cells. e Chromatin accessibility in NB4 cells expressing scrambled control versus shKDM3B at promoters regions. The enriched regions were extended 0.5 kb from their midpoint. f Genome ± browser tracks of ATAC-seq data at the C11orf95, CBX4 and CDK2AP1 loci following silencing of KDM3B in NB4 cells. g Venn diagram representing overlap of accessible sites in NB4 cells expressing scrambled control versus shKDM3B. h TF motif analysis of diferent regions in NB4 cells expressing scrambled control versus shKDM3B in all chromatin accessible sites. The scale of the circles represents motif enrichment

Silencing KDM3B promotes tumor cells growth and inhibits studies have tried to assess KDM3B activity, the conclu- ATRA induced NB4 cell diferentiation sions have been contradictory [22, 24, 32]. Terefore, the To further explore the functions of KDM3B in hemat- H3K9 methylation regulated by KDM3B in genome-wide opoietic malignancies, we knocked down KDM3B were investigated by ChIP-seq in NB4 cells that stably expression in NB4 cells. NB4 cells were derived from expressed either the scramble control or shKDM3B. Cor- granulocytic M3/APL subtype of AML, which expressed relation clustering analyses showed that H3K9me1/me2/ PML/RARα fusion protein. Lentivirus-mediated short me3 binding in NB4 cells with shKDM3B or scramble hairpin RNAs (shKDM3B-1 and shKDM3B-2) target- control was highly correlated (Spearman’s correlation ing KDM3B or a negative control (scramble) were trans- coefcient = 0.77, 0.88 and 0.90, respectively) (Fig. 2a, fected into NB4 cells. Western blot were used to test b). From the metagene plots, we observed that KDM3B the knock down efciency. Compared to the scramble silencing increased the levels of H3K9me1 and reduced control, KDM3B was signifcantly down regulated in the levels of H3K9me2 in both the TSS region and the NB4 cell transfected with shKDM3B-1 and shKDM3B-2 gene-body region, while global levels of H3K9me3 (Fig. 1e). remained largely variable (Fig. 2c). For example, we Te cell proliferation was examined in KDM3B-knock- observed substantial changes in H3K9me1/me2/me3 down NB4 cells. Results showed that shRNA knock- occupancy at the C11orf95, CBX4 and CDK2AP1 loci down of endogenous KDM3B did signifcantly promoted (Fig. 2d). Tese fndings were consistent with publica- the proliferation of NB4 cells compared with the control tions showing that KDM3B is a H3K9me1/me2 demethy- (Fig, 1f). Cell cycle distribution was determined using lase in human tissues. cytometry analysis. NB4 cells with shKDM3B had an To further investigate the genome-wide accessible obvious increased percentage of S phase cells (Fig. 1g), regions regulated by KDM3B, ATAC-seq was used. As suggesting the high DNA synthesis and cell proliferation shown in the heatmap, KDM3B silencing dramatically rate when KDM3B was defcient. increased the global chromatin accessibility in NB4 cells In addition, we investigated the potential roles of (Fig. 2e). Browser track representations showed increased KDM3B in ATRA mediated leukemia cell diferen- chromatin accessibility of C11orf95, CBX4 and CDK2AP1 tiation by fow cytometry analysis using CD11b, a at three loci (Fig. 2f). granulocytic diferentiation marker. Notably, KDM3B- Next, transcription factors associated with KDM3B knockdown NB4 cells clearly had a low expression level were identifed. shKDM3B-specifc open chromatin of the diferentiation marker CD11b despite ATRA treat- regions were selected for further study (Fig. 2g). By using ment (Fig. 1h). Collectively, these results indicated that motif analysis software HOMER, we found that NB4 KDM3B was an important contributor to APL cell sur- cells with shKDM3B-specifc open chromatin regions vival and diferentiation. were enriched with binding sites for transcription factors known as members of bZIP transcription factor and ETS Silencing KDM3B globally increases the level of H3K9me1 transcription factor families (Fig. 2h). bZIP transcription and reduces the level of H3K9me2 factors including AP-1, c-Jun, Jun-B, Atf3, BATF, Fra1 and KDM3B, which belongs to the subfamily of proteins con- Fra2, play important roles in positive regulation of tran- taining a JmjC domain, has been previously shown to scription by RNA polymerase II. Te ETS transcription participate in H3K9me1/me2 demethylation activities factors family, especially the PU.1 factor, is a master regu- at specifc regions within the genome. Although several lator of myeloid diferentiation. Interestingly, the CTCF Wang et al. Cancer Cell Int (2019) 19:256 Page 7 of 15

a c H3K9me1 Chip-seq Scramble shKDM3B Transcription start sitesGene body regions 1.8 1.8 1.4 Scramble-H3K9me2shKDM3B-H3K9me2Scramble-H3K9me3shKDM3B-H3K9me3Scramble-H3K9me1shKDM3B-H3K9me1 1.6 shKDM3B-H3K9me1 1.0

Scramble-H3K9me1 0.6 0.6 -3.0kb TSS 3.0kb -3.0kb TSSTES 3.0kb shKDM3B-H3K9me3 H3K9me2 Chip-seq Scramble-H3K9me3 Transcription start sitesGene body regions 3.5 3.0 shKDM3B-H3K9me2 2.5 2.0 Scramble-H3K9me2 1.5 1.0 0.00.2 0.40.6 0.81.0 0.5 0 b Normalized average signal -3.0kb TSS3.0kb -3.0kb TSSTES 3.0kb coefficient=0.77 coefficient=0.88 coefficient=0.90 1 2 3 H3K9me3 Chip-seq 10 10 10 Transcription start sitesGene body regions 8 8 8 6 6 6 5.0 4.0 4 4 4 3.0 2 2 2 3.0 0 0 0 2.0 Scramble H3K9me Scramble H3K9me 024 6810 0 2 46810 Scramble H3K9me 0246810 shKDM3B H3K9me1 shKDM3B H3K9me2 shKDM3B H3K9me3 1.5 1.0 -3.0kb TSS3.0kb -3.0kb TSSTES 3.0kb

10 10 15 d Scramble H3K9me1 10 10 15 shKDM3B

30 20 20 Scramble H3K9me2 30 20 20 shKDM3B 30 50 30 Scramble H3K9me3 30 50 30 shKDM3B

C11orf95 CBX4 CDK2AP1 e Scramble shKDM3B f h Motif CTCF 25 90 Scramble BORIS SpiB(ETS) 25 shKDM3B PU.1 80 GABPA Fli1 Etv2 %of Targets C11orf95 40 70 ETV1 40 Scramble ETS1 30 40 Ets1−distal shKDM3B 60 ETS

factor famil y 20 ERG ETS transcription ELF5 -log10(p-value) 50 CBX4 ELF3 2000 25 ELF1 Scramble EHF 4000 40 25 6000 shKDM3B JunB Jun Fra2 30 Fra1 CDK2AP1 Fosl2 g BATF Shared peaks factor famil y 20 Atf3 bZIP transcription AP−1

s s s 10 16166 37352 10331

hared peak 0 S -0.5kb TSS 0.5kb -0.5kb TSS 0.5kb Scramble unique peaks shKDM3B unique peaks shKDM3BScramble unique unique peak peak Wang et al. Cancer Cell Int (2019) 19:256 Page 8 of 15

factor was most enriched in ATAC-seq peaks (Fig. 2h). between the KDM3B modulatory efect and gene expres- Previous results have suggested that transcription factor sion changes related to the APL diferentiation process, CTCF is associated with open chromatin regions. Taken the diferential gene expression patterns of KDM3B- together, our results showed that KDM3B specifcally knockdown NB4 cells were compared against those of regulated chromatin accessibility and H3K9me1/me2 the scrambled control cells in the presence of ATRA, and modifcation and taht these regulations are crucial for the the diferentiated control NB4 cells were compared with NB4 leukemia cell survival and diferentiation. the undiferentiated NB4 cells, as schematically depicted in Fig. 4a (Additional fles 3, 4). As shown by the venn The genes activated after KDM3B silencing are located diagram, 291 ATRA-induced genes showed reduced in chromatin accessible regions expression in KDM3B-knockdwon NB4 cells (cluster 1), We used RNA-seq to further study the regulatory mecha- while only 80 ATRA-inhibited genes were upregulation nisms of KDM3B. A P-value < 0.05 and |fold change| ≥ 1.5 in cells lacking KDM3B (cluster2). Tus, KDM3B has were set to be the standards to identify KDM3B regula- an important facilitating role in the induction of certain tory genes to be compared against the scramble con- genes characteristic of the diferentiation process. trol. According to these parameters, 774 diferentially Moreover, we performed gene set enrichment analy- expressed genes were acquired, of which 399 genes were sis (GSEA) to defne the functions of the 1198 diferen- upregulated and 357 genes were down regulated in the tial expression genes of the KDM3B-knockdown NB4 shKDM3B NB4 cells (Additional fle 2). All the genes cells versus those of the scrambled control cells induced are depicted in heatmap (Fig. 3a). A panel of the top hits by ATRA (P-value < 0.05 and |fold change| ≥ 1.5). In line from our RNA-seq data was validated by conventional with the fow cytometric results, knocking down KDM3B qPCR (Fig. 3b, c). We used Gene Ontology (GO) clas- led to suppressed expression of a leukemia stem cell sifcation and KEGG to explore the biological functions (LSC) gene signature and a hematopoietic stem cell gene of the KDM3B regulatory genes. Te GO classifcation signature (Fig. 4b–d). We also found a negative associa- included leukocyte diferentiation and neutrophil activation between the NPM1-mutated AML gene signature tion (Fig. 3d). Te KEGG pathway analysis showed sig- and KDM3B-knockdown gene signature in NB4 cells nifcant enrichment of the pathways typically found in (Fig. 4e). cancer (Fig. 3e). To further characterize the putative transcription fac- Furthermore, to determine whether chromatin open tors that control the aberrant expression of these genes, regions from the ATAC-seq analysis correlated with the we performed an enrichment analysis of the ChIP-X KDM3B target gene expression, we integrated our ATAC- (ChEA) database with Enrichr. Among the transcription seq data with and RNA-seq data. Venn diagrams showed factors in the database, PU.1 was ranked as one of the that 83% of upregulated genes in KDM3B-knockdown most potent transcription factors involved in regulating NB4 cells had open chromatin regions (Fig. 3f). Tese the diferentially expressed genes (Fig. 4f). results suggest that open chromatin may be a predictor We previously demonstrated that PML/RARα is of gene activation in NB4 cells. In contrast, only 2% of recruited to PU.1-bound chromatin containing RARE ATAC-seq peaks were mapped to diferentially expressed half and PU.1 sites [25, 26]. In the present study, we genes (Fig. 3f). As shown by the genomic browser tracks found that over 53% (1885/3549) of the PML/RARα bind- (Fig. 3g), chromatin openness and MYC expression were ing regions are targeted by PU.1 in NB4 cells (Fig. 4g). markedly increased, as shown in the left panel. However, De novo motif analysis revealed that the PU.1 and Jun- an increased ATAC-seq signal did not signifcantly alter AP1 binding motifs are the most signifcant motifs found ZNF503 transcript levels (middle panel), and increased in PML/RARα bound regions (Fig. 4h). It has been sug- KDM5B expression did not accompany by increased gested that APL-specifc PML/RARα fusion protein is chromatin accessibility (right panel). Tese results con- bound to chromatin open regions [35]. Integrating high- cur with the growing understanding that gene activation throughput sequencing data from a variety of ChIP-seq depends on multiple regulatory regions, many of which experiments and results from analysis of the accessible are located far from the gene locus itself. chromatin landscapes, including those based on DNase- seq and ATAC-seq, we found that the vast majority of KDM3B facilitates the transcriptional alterations induced PML/RARα binding sites overlap with epigenetic marks by ATRA of active chromatin (H3K4me3, H3K4me1, H3K9ac and ATRA is widely used as a diferentiating agent in the RNAPII), while a minority overlap with inhibitory chro- clinic [33, 34]. Notably, KDM3B-knockdown NB4 matin marks (H3K27me3, H4K20me3 and H3K9me3) cells clearly inhibited ATRA-mediated APL diferen- (Fig. 4i). Combined with the results previously obtained tiation (Fig. 1h). To further characterize the relationship (Fig. 2), we speculate that KDM3B might contribute to Wang et al. Cancer Cell Int (2019) 19:256 Page 9 of 15

ab d Count Scramble 6 Leukocyte differentiation shKDM3B 10 4 Defense response to other organism 20 Granulocyte activation 2 30 1 Neutrophil activation 0 40 Positive regulation of cytokine production -2 p.adjust (fold change) with

2 Neutrophil activation in immune response

KDM3B knock down -4

lo g Lymphocyte differentiation 0.5 1 0.002 LMO2 ATA Regulation of vasculature development CREB5 GRID1 G MAGEB2 ADGRE1 TMPRSS4 0.001

0.010.02 0.03 0.04 0.050.06 0 c e GeneRatio

8 * Count Pathways in cancer 15 6 Scramble 20 −0.5 A levels shKDM3B PI3K−Akt signaling pathway * 25 4 Transcriptional misregulation in cancer 30 * 35 2 * AGE−RAGE signaling pathway * * * * p.adjust ECM−receptor interaction

Relative RN 0 −1 B Acute myeloid leukemia 0.004 TA1 LMO2 ACTB CREB5 GRID1 GA KDM3 MAGEB2 ADGRE1 0.05 0.07 0.09 0.11 0.002 TMPRSS4 Scramble shKDM3B GeneRatio

fg 2000 60 80 Scramble RNA-seq 2000 60 80 shKDM3B 150 Scramble 150 50 shKDM3B ATAC-seq 150 150 50 shKDM3B 20 15 10 Scramble H3K9me1 15811 330 69 20 15 10 shKDM3B 15 40 10 RNA-seq Scramble H3K9me2 15 40 10 shKDM3B ATAC-seq 30 100 20 Scramble

H3K9me3 30 100 20 shKDM3B

MYC ZNF503 KDM5B Fig. 3 Integration of ATAC-seq data with RNA-seq for the activated genes after KDM3B silencing in NB4 cells. a Heatmap of diferentially expressed genes (defned as |fold change| 1.5, with a P-value < 0.05) in KDM3B-knockdown NB4 cells relative to scrambled control cells. b Histogram ≥ showing a subset of the genes identifed diferentially expressed in NB4 cells expressing scrambled control versus shKDM3B from RNA-seq data. c RT-qPCR validation of diferentially expressed genes in “B” following silencing of KDM3B in NB4 cells. Relative mRNA levels were calculated by ΔΔCT 2− method and normalized to ACTB (β-actin). d GO analysis performed on the diferentially expressed genes. e Pathway analysis using KEGG was performed on the genes diferentially expressed. f Venn diagrams indicating the proportion of genes that diferentially expressed in NB4 cells expressing scrambled control versus shKDM3B based on RNA-seq data that have ATAC-seq peaks (identifed only in KDM3B knock-down NB4 cells). g Genome browser tracks of RNA-seq data; ATAC-seq data; H3K9me1, H3K9me2 and H3K9me3 ChIP-seq data at the MYC, ZNF503 and KDM5B loci following silencing of KDM3B in NB4 cells. Values are derived from three independent experiments, and data were mean SD. *P < 0.05 ± genome-wide binding of PML/RARα. Notably, when in NB4 cells led to increased chromatin accessibility in KDM3B was silenced, the PML/RARα binding sites PML/RARα binding regions (Fig. 4k). Together, these were occupied by a substantially increased number of data strongly implied that the KDM3B expression is H3K9me1 marks, and a modest loss of H3K9me2 signal closely linked to the state of normal granulocytic difer- was observed (Fig. 4j). In addition, KDM3B silencing entiation by altering chromatin accessibility. Wang et al. Cancer Cell Int (2019) 19:256 Page 10 of 15

(See fgure on next page.) Fig. 4 KDM3B infuences the expression of ATRA-regulated genes in diferentiation of NB4 cell. a Venn diagram illustrates the number of genes regulated by ATRA, the genes infuenced by KDM3B in ATRA-induced diferentiating, and the overlaps of these 2 gene sets. 291 genes are less induced by ATRA in KDM3B-knockdown NB4 cells than normally expected (cluster1), 80 genes are more up-regulated in KDM3B-knockdown NB4 cells as compared with the control (cluster 2). b–e GSEA of the expressing profle of NB4 cells transfected with scramble control or shKDM3B followed by treatment with ATRA using a Leukemic stem cell down-regulated signature (b), a Hematopoietic stem cell down-regulated signature (c), a positive regulation of hemopoiesis-associated signature (d), and a AML with NPM1 mutated-associated signature (e). f ChIP enrichment analysis of diferentially expressed genes in KDM3B-knockdown NB4 cells relative to scrambled control cells followed by treatment with ATRA. The top 9 ranked transcription factors are shown (ranked by adjusted P-value). g Overlap of PU.1 and PML/RARα binding revealed by ChIP-seq. The peak numbers of each category are shown on the venn diagram. h De novo motif analysis of PML/RARα binding sites. Signifcantly enriched motifs are shown. i PML/RARα binding sites overlap with the active and repress histone modifcation. j Signal intensity plot representing changes in H3K9me1 and H3K9me2 ChIP-seq signal at PML/RARα binding regions followed by silencing of KDM3B in NB4 cells. The enriched regions were extended 3 kb from their midpoint. k Chromatin accessibility in NB4 cells expressing scrambled control versus shKDM3B at PML/RARα binding ± regions. The enriched regions were extended 0.5 kb from their midpoint ±

KDM3B regulates ATRA‑induced degradation of PML/RARα RARα ChIP-seq data were derived from the public data and chromatin state in NB4 cells on NB4 cells and PR9 cells and analyzed. Altogether, we In our study, we confrmed that KDM3B is required identifed a set of 436 PML/RARα-binding genes shared for ATRA-induced granulocytic diferentiation in NB4 by these two cell types. Among the PML/RARα binding cells. Ten, we sought to determine whether KDM3B is targets that we identifed, 29 genes showed a signifcant involved in the regulation of expression of APL-specifc decrease in mRNA when KDM3B was knocked down PML/RARα fusion protein expression. To test this pos- (Additional fle 5). Tese 29 genes were considered as the sibility, we frst used immunostaining to detemine that PML/RARα-KDM3B coregulated gene signature, which KDM3B is primarily distributed in the nucleus of NB4 included key regulators of cell fusion and substrate bind- cells (Fig. 5a). Second, we examined KDM3B expression ing (Fig. 5j) and contributed to prediction of the overall in an ATRA-induced diferentiation time series experi- survival for patients with AML (Fig. 5k). ment using NB4 cells. Upregulation of KDM3B is clearly observed at total protein level, whereas the downregu- Discussion lation of KDM3B is observed at nuclear protein level In this study, we provide several lines of evidence that (Fig. 5b, c). Finally, we investigated whether the inhibition KDM3B, H3K9me1/me2 demethylase, exerted a key role of KDM3B can attenuate the ATRA-induced clearance of in the epigenetic regulation and cellular responses to PML/RARα. It was shown that knocking down KDM3B ATRA treatment in APL by using a genetically defned impaired the degradation of PML/RARα during cell dif- human APL model of PML/RARα positive NB4 cells. ferentiation in both PR9 cells and NB4 cells (Fig. 5d–f). We propose a mechanism that KDM3B was required Few studies have verifed that ATRA treatment leads to to maintain a compact chromatin state in proper devel- acetylation of histones H3 and H4, which induces wide- opment and function of APL (Fig. 6a). Te absence of spread chromatin opening with low modifcation levels of KDM3B could alter H3K9me1/me2 modifcation level epigenetic marks of suppressed chromatin [35, 36], such companied with increasing of chromatin accessibility as H3K27me3 (Fig. 5g). In addition, we also found that and inhibit ATRA-induced degradation of PML/RARα ATRA induces an increase in H3K9me1 and a decrase in and, consequently, allow “easier” chromatin access and H3K9me2 near the TSS regions in the NB4 cells (Fig. 5h), binding of PML/RARα protein in APL, which eventually a fnding that is consistent with epigenetic changes leads to attenuate neutrophil diferentiation of APL cells induced by knocking down KDM3B. Together, these data (Fig. 6b). indicate that KDM3B is not only an ATRA-responsive Previous studies have demonstrated that KDM3B is gene but is also a contributor to sustained PML/RARα located at a locus (5q31 chromosome region) commonly levels and chromatin state maintenance during cell difer- deleted in approximately 15% of primary MDS cases, 10% entiation, showing a close correlation between KDM3B of AML cases, and 40% of the therapy-induced cases of levels and PML/RARα regulation. MDS/AML [19, 20, 37]. In addition, KDM3B−/− mice To improve the sensitivity and specifcity of potential displayed abnormal phenotypes in the hematopoietic sys- target gene identifcation to fnd those that are regu- tem [32]. Terefore, aberrantly low expression of KDM3B lated by KDM3B and PML/RARα, we conducted an inte- is thought to have latent functions in AML and MDS. grated analysis of the ChIP-seq DNA–protein binding Unexpectedly, it was recently shown that KDM3B acts and RNA-seq diferential expression data (Fig. 5i). PML/ as an H3K9me1/me2 demethylase and induces leukemic Wang et al. Cancer Cell Int (2019) 19:256 Page 11 of 15

a b Gal: Leukemic stem cell down c Jaatinen: Hematopoietic stem cell down 0.6 Down-regulated genes in ATRA Up-regulated genes 0.6 NES: 2.02 NES: 1.88 treated KDM3B knock-down NB4 in ATRA treated NB4 0.4 FDR-q: 0.038 0.4 FDR-q: 0.04 0.2 0.2 0.0 cluster1 0.0 Enrichment score Enrichment score Scramble shKDM3B Scramble shKDM3B +ATRA +ATRA +ATRA +ATRA

d GO: Positive regulation of hemopoiesis e Verhaak: AML with NPM1 mutated up differentiation 0.6 0.6 NES: 1.95 NES: 1.87 Effect of ATRA-induced 0.4 FDR-q: 0.019 0.4 FDR-q: 0.04 Effect of KDM3B knock-down in ATRA-induced differentiation 0.2 0.2 cluster2 0.0 0.0 Up-regulated genes in ATRA Down-regulated genes Enrichment score treated KDM3B knock-down NB4 in ATRA treated NB4 Enrichment score Scramble shKDM3B Scramble shKDM3B +ATRA +ATRA +ATRA +ATRA

f g NB4 ChIP-seq peak overlap TF PMID of the respective studiesAdjusted P-valueCombined Score

PU.1 23547873_ChIP-Seq_NB4_Human 1.28E-1032.52612 PU.1 25018 1885 1664 PML-RARα RELA 24523406_ChIP-Seq_Fibrosarcoma_Human 2.65E-06 31.42454 MYB 26560356_ChIP-seq_TH2_Human 1.21E-0630.87648 MAF 26560356_ChIP-Seq_TH1_Human 0.00025219.84758 h KDM2B 26808549_ChIP-Seq_DND41_Human 0.000644 17.64273 Jun-AP1 motif (p-value: 1E-261) KDM2B 26808549_ChIP-Seq_SIL-ALL_Human0.00049517.63475 A GCC GCG C CTG A G A A C CG G T C C T UTX 26944678_ChIP-Seq_JUKART_Human0.00101616.93838 TCTG TA GTA TAGTACTGCA GATAAT PU.1 motif (p-value: 1E-222) MAF 26560356_ChIP-Seq_TH2_Human 0.001016 16.55802 TC KDM2B 26808549_ChIP-Seq_JURKAT_Human0.00137215.62531 CA CT T GG CA CTG A CG CG A G T A G T A G A GA GAGA TACTTATCCCTTC G

ij k Scramble shKDM3B % PML-RARα peaks PML-RARα binding sites Scramble 05205 75 100 H3K9me1 ChIP-seq shKDM3B 3.5 80 ATAC-seq 70 2.5 DNAse-seq s 60 RNAPII 1.5

H3K4me1 50 0.5 Normalized average signal -3.0kb Center 3.0kb H3K4me3 40 H3K9me2 ChIP-seq 1.6 H3K9ac 30 1.2 PML-RARα binding site H3K27me3 20 0.8 H4K20me3 0.4 10 H3K9me3 Proportion of overlap peaks 0.0 0 Proportion of unique peaks Normalized average signal -3.0kb Center 3.0kb b b b 0.5k 0.5k -0.5k Center -0.5kb Center

transformation in AML subtypes, such as HL-60 (AML changes in H3K9me1 after KDM3B was knocked down, FAB M2) and MUTZ-8 (AML-derived) [22, 24]. How- while some others report fnding a negative correlation ever, there is controversy surrounding the mode by between H3K9me1 and KDM3B. Considerating that all which the marks are distributed on the histones, as these studies were focused on specifc chromatin loci, the some authors suggested that there were no remarkable underlying mechanism of KDM3B combined with the Wang et al. Cancer Cell Int (2019) 19:256 Page 12 of 15

(See fgure on next page.) Fig. 5 KDM3B is associated with PML/RARα oncoprotein during NB4 cell diferentiation. a Confocal staining of KDM3B in NB4 cells. Nuclei were counterstained with DAPI. Representative images are shown. Scale bar: 10 μm. b, c Western blot analyses of KDM3B and PML/RARα in NB4 cells at total/nuclear protein level, followed by treatment with ATRA. β-actin was used as a loading control for total protein (b), histone H3 was used for nuclear proteins (c). d Efects of stable KDM3B-knockdown in PR9 cells by Western blot analysis. e Western blot analyses of PML/RARα in PR9 cells 2 transfected with scramble control or shKDM3B, with or without Zn + pretreatment for 4 h, followed (or not) by treatment with ATRA. f Western blot analyses of PML/RARα in NB4 cells transfected with scramble control or shKDM3B followed (or not) by treatment with ATRA. g, h Signal intensity plot representing changes in H3K9ac, H3K27me3, H3K9me1 and H3K9me2 ChIP-seq signal at promoters regions in NB4 cells followed by treatment with ATRA. The enriched regions were extended 0.5 kb from their midpoint. i Strategy and workfow of integrative analysis to identify 29 target ± genes directly regulated by KDM3B and PML/RARα. j GO analysis performed on KDM3B and PML/RARα co-regulated 29-gene signature. k Overall survival of AML patient cohort with regard to the concomitant KDM3B and PML/RARα co-regulated 29-gene signature. Values are derived from three independent experiments, and data were mean SD. *P < 0.05 ± genomic landscape of H3K9me1/me2 modifcations and of accessible chromatin with some hallmarks of active chromatin remodeling, deserves further exploration. chromatin [25, 35, 36]. Moreover, ATRA exerts its thera- Te levels of H3K9me2/me3 have been found to be peutic efect by promoting the degradation of the PML/ enriched in nongenic regions, which resulted in hetero- RARα oncogenic protein, which subsequently promotes chromatin formation and epigenetic silencing, while cell diferentiation [33, 50]. In particular, the observa- H3K9me1 levels were enriched at active genes and open tions in this study indicated that ATRA-induced degrada- chromatin [38, 39]. Moreover, histone demethylation tion of PML/RARα increases KDM3B expression at the and chromatin decondensation can lead to dysregu- total protein level while decreasing of KDM3B expression lated gene expression and transcriptional activation of at the nuclear protein level. In addition, knocking down gene targets [8, 27, 40]. Our results suggest that knock- KDM3B impairs the degradation of PML/RARα during down KDM3B could increase the level of H3K9me1 ing cell diferentiation, suggesting the positive feedback and reduce the level of H3K9me2 in both the TSS between PML/RARα and KDM3B. All of these results, region and gene-body region. Tis fnding supports the as discussed above, strongly suggest that KDM3B plays notion that KDM3B induces the specifc demethyla- a critical role in regulating ATRA treatment in APL cells tion of H3K9me1 and H3K9me2. In addition, knocking via KDM3B mediated chromatin opening. down KDM3B globally enhanced chromatin accessibil- In summary, our studies suggest that histone demethy- ity, which facilitated transcription release. Furthermore, lase KDM3B correlates with APL disease and drug resist- by integrating ATAC-seq data with RNA-seq data using ance, in addition, our data provide a novel insight into KDM3B-knockdown NB4 cells, our results could also PML/RARα-driven APL pathogenesis and treatment. refect the fact that opening chromatin only in certain In this study, APL was used as an AML model to inves- regions is not sufcient for gene activation. tigate the functions of KDM3B. Tese results may be In the case of KDM3B-knockdown cells, we found that achieved in other AML types. We show that, compared the CTCF consensus motif was signifcantly enriched in with those in normal cells, KDM3B mRNA levels were open chromatin regions. Importantly, CTCF is a widely signifcantly reduced in AML blast cells (Fig. 1c), suggest- expressed transcription factor and a “gatekeeper factor” ing that KDM3B is critical for overall AML cell survival. that regulates transition from dormancy to activation in Future studies using multiomics resource are warranted hematopoietic stem cell (HSC) development and con- to investigate the role of KDM3B in other types of leuke- trols stem cell pool throughout life [41–46]. Moreover, mia from. after KDM3B silencing, selective open chromatin regions were enriched in bZIP transcription factor, which plays Conclusion an important roles in AML development [47–49]. Tus, Histone demethylase KDM3B is considered to play further study of these transcription factors in APL is a critical role in leukemogenesis. To the best of our expected to reveal insights into the temporally and spa- knowledge, this study is the frst attempt to probe the tially regulated chromatin structure that programs in detailed genetic and epigenetic mechanisms underlying myeloid diferentiation. the regulation of KDM3B in the development of APL PML/RARα is the hallmark fusion oncoprotein from the perspectives of multi-layer omics. Results of implicated in APL pathogenesis and results in the the present study demonstrate that KDM3B exerts anti- attenuation of gene expression needed for myeloid dif- APL efect by directly modulating H3K9me1/me2 levels ferentiation [34]. Previous studies have shown that the to maintain compact chromatin status, but also indi- majority of PML/RARα binding sites tend to be in regions cate the interaction between KDM3B and PML/RARα Wang et al. Cancer Cell Int (2019) 19:256 Page 13 of 15

a NB4 b c

12h 24h 12h 24h DAPI RA RA ATRA

NB4 NB4+ATNB4+RA AT NB4 NB4+ NB4+AT PML-RARα PML-RARα KDM3B KDM3B KDM3B β-actin H3 Merged

(63X10)10μm h 24 24h 48h 48h RA RA 24h 24h TRAAT TRAAT n n TRAAT RA e B B de f

PR9-ScramblPR9-shKDM3 PR9-ScramblePR9-shKDM3PR9-Scramble+ZPR9-shKDM3B+ZPR9-Scramble+Zn+APR9-shKDM3B+Zn+PR9-Scramble+Zn+APR9-shKDM3B+Zn+ NB4-ScrambleNB4-shKDM3BNB4-Scramble+ANB4-shKDM3B+ KDM3B PML-RARα PML-RARα β-actin β-actin β-actin

g h

H3K9ac H3K27me3 H3K9me1 3.2 H3K9me2 14 0.7 1.4 2.8 10 0.6 Normalized Normalized 1.0 6 0.5 2.4 0.4 average signal

2 average signal 0.6 2.0 -0.5kb TSS0.5kb -0.5kb TSS0.5kb -0.5kb TSS0.5kb -0.5kb TSS0.5kb ChIP-seq ChIP-seq NB4NB4+ATRA i k NB4 PML-RARα bound genes PR9 PML-RARα bound genes Overall Survival from ChIP-seq from ChIP-seq 1.0 cluster1 cluster2 l 0.8 Logrank p=0.0029 vi va 1389 407 1270 0.6

29 29 PML-RARα & KDM3B 0.4 39 36 targeted genes ercent sur P 0.2 271 0.0 Do wn-regulated genes 020406080 in KDM3B knock-down Months

j GO terms Term Name Binom P-Value Regulation of binding 5.18E-04 Leukocyte differentiation 3.81E-03 Negative regulation of cell differentiation 4.52E-03 Cell communication 9.92E-03 Wang et al. Cancer Cell Int (2019) 19:256 Page 14 of 15

KDM3B -depleted a b

H3K9me1 H3K9me2H3K9me1 KDM3B H3K9me2 Co-factors PML-RARα ATRA TFs TFs Differentiation

H3K9me1 H3K9me1 Closed Chromatin Open Chromatin Fig. 6 A proposed model for KDM3B and PML/RARα in APL cell diferentiation with KDM3B existence (a) or depletion (b), respectively. See “Discussion” for details regulates degradation of PML/RARα. Tese fndings Ethics approval and consent to participate Not applicable. will provide insight into PML/RARα-driven APL pathogenesis, and theoretical and experimental data for the Consent for publication treatment regimens and target therapy of APL. Not applicable. Competing interests Supplementary information The authors declare that they have no competing interests. Supplementary information accompanies this paper at https://doi. org/10.1186/s12935-019-0979-7. Received: 2 May 2019 Accepted: 26 September 2019

Additional fle 1. Quantitative PCR primers and details of antibodies. Additional fle 2. Diferentially expressed genes between KDM3B-knockdown NB4 versus scrambled control NB4 cells. References 1. Maury E, Hashizume R. Epigenetic modifcation in chromatin machinery Additional fle 3. Diferentially expressed genes between KDM3B- and its deregulation in pediatric brain tumors: insight into epigenetic knockdown NB4 versus scrambled control NB4 cells followed with ATRA therapies. Epigenetics. 2017;12(5):1–17. treatment. 2. Flavahan WA, Gaskell E, Bernstein BE. Epigenetic plasticity and the hall- Additional fle 4. Diferentially expressed genes between ATRA-induced marks of cancer. Science. 2017;357(6348):eaal2380. diferentiated control NB4 versus undiferentiated NB4 cells. 3. Atlasi Y, Stunnenberg HG. The interplay of epigenetic marks during stem cell diferentiation and development. Nat Rev Genet. 2017;18(11):643–58. Additional fle 5. PML/RARα-KDM3B coregulated 29-genes signature. 4. Barski A, Cuddapah S, Cui K, Roh T, Schones D, Wang Z, Wei G, et al. High- resolution profling of histone methylations in the human genome. Cell. 2007;129(4):823–37. Abbreviations 5. Zhang T, Cooper S, Brockdorf N. The interplay of histone modifcations— KDM3B: histone lysine-specifc demethylase 3B; APL: acute promyelocytic leu- writers that read. EMBO Rep. 2015;16(11):1467–81. kemia; ATRA: all-trans retinoic acid; PML: promyelocytic leukemia protein; MDS: 6. Declerck K, Vel Szic KS, Palagani A, Heyninck K, Haegeman G, Morand C, myelodysplastic syndromes; AML: acute myeloid leukemia; TSS: transcription et al. Epigenetic control of cardiovascular health by nutritional polyphe- start site; TCGA: The Cancer Genome Atlas; GSEA: gene set enrichment analy- nols involves multiple chromatin-modifying writer-reader-eraser proteins. sis; ChIP: chromatin immunoprecipitation; NGS: next-generation sequencing. Curr Top Med Chem. 2016;16(7):788–806. 7. Greer EL, Yang S. Histone methylation: a dynamic mark in health, disease Acknowledgements and inheritance. Nat Rev Genet. 2012;13(5):343–57. We acknowledge and appreciate our colleagues for their valuable eforts and 8. Faundes V, Newman WG, Bernardini L, Canham N, Claytonsmith J, comments on this paper. Dallapiccola B, et al. Histone lysine methylases and demethylases in the landscape of human developmental disorders. Am J Hum Genet. Authors’ contributions 2018;102(1):175–87. XW designed experiments, generated data, analyzed/interpreted data, and 9. Bo W, Hao W, Yoichi S, Irizarry RA, Feinberg AP. Large histone H3 lysine 9 prepared the manuscript; HF, GJ, CX and HW participated in study design dimethylated chromatin blocks distinguish diferentiated from embry- and edited the article. JZ proposed the study, supervised the experiment and onic stem cells. Nat Genet. 2009;41(2):246. revised the drafted article. All authors critically revised the manuscript. All 10. Chen J, Liu H, Liu J, Qi J, Wei B, Yang J, et al. H3K9 methylation is a authors read and approved the fnal manuscript. barrier during somatic cell reprogramming into iPSCs. Nat Genet. 2013;45(1):34–U62. Funding 11. De S, Kassis JA. Passing epigenetic silence to the next generation. Sci- This work was supported by National Science Foundation of China (81370655). ence. 2017;356(6333):28–9. 12. Yoshimi A, Kurokawa M. Key roles of histone methyltransferase and dem- Availability of data and materials ethylase in leukemogenesis. J Cell Biochem. 2011;112(2):415–24. H3K9ac, H3K27me3, PML and RARα ChIP-seq data for NB4 and PR9 cells were 13. Morera L, Lübbert M, Jung M. Targeting histone methyltransferases obtained from GSE18886 [36]. Microarray gene expression data for AML and demethylases in clinical trials for cancer therapy. Clin Epigenet. patients were obtained from GSE9476 and GSE7186 [51, 52]. The data for 2016;8(1):57. ChIP-seq and RNA-seq analyses can be found at Gene Expression Omnibus 14. Majello B, Gorini F, Saccà CD, Amente S. Expanding the role of the histone GSE137105. lysine-specifc demethylase LSD1 in cancer. Cancers. 2019;11(3):324. Wang et al. Cancer Cell Int (2019) 19:256 Page 15 of 15

15. Tsukada Y, Fang J, Erdjumentbromage H, Warren ME, Borchers CH, Tempst and AML1-ETO binding sites in acute myeloid leukemia. Blood. P, et al. Histone demethylation by a family of JmjC domain-containing 2012;120(15):3058–68. proteins. Nature. 2006;439(7078):811–6. 36. Martens JHA, Brinkman AB, Femke S, Kees-Jan F, Angela N, Felicetto F, 16. Michael B, Yao Z, Rishi A, Sachin T, Ieuan C, Bruno I, et al. Protein complex et al. PML-RARalpha/RXR alters the epigenetic landscape in acute pro- interactor analysis and diferential activity of KDM3 subfamily members myelocytic leukemia. Cancer Cell. 2010;17(2):173–85. towards H3K9 methylation. PLoS ONE. 2013;8(4):e60549. 37. Le Beau MM, Espinosa R, Neuman WL, Stock W, Roulston D, Larson RA, 17. Ohguchi H, Hideshima T, Bhasin MK, Gorgun GT, Santo L, Cea M, et al. The et al. Cytogenetic and molecular delineation of the smallest commonly KDM3A–KLF2–IRF4 axis maintains myeloma cell survival. Nat Commun. deleted region of chromosome 5 in malignant myeloid diseases. Proc 2016;7:10258. Natl Acad Sci USA. 1993;90(12):5484–8. 18. Chen M, Zhu N, Liu X, Laurent B, Tang Z, Eng R, et al. JMJD1C is required 38. Kang X, Feng Y, Gan Z, Zeng S, Guo X, Chen X, et al. NASP antagonize for the survival of acute myeloid leukemia by functioning as a coactivator chromatin accessibility through maintaining histone H3K9me1 in hepa- for key transcription factors. Genes Dev. 2015;29(20):2123. tocellular carcinoma. Biochim Biophys Acta. 2018;1864(10):3438–48. 19. Hu Z, Gomes I, Horrigan SK, Kravarusic J, Mar B, Arbieva Z, et al. A novel 39. Srimongkolpithak N, Sundriyal S, Li F, Vedadi M, Fuchter MJ. Identifcation nuclear protein, 5qNCA (LOC51780) is a candidate for the myeloid leu- of 2,4-diamino-6,7-dimethoxyquinoline derivatives as G9a inhibitors. kemia tumor suppressor gene on chromosome 5 band q31. Oncogene. Medchemcomm. 2014;5(12):1821–8. 2001;20(47):6946. 40. Kooistra SM, Helin K. Molecular mechanisms and potential functions of 20. Mackinnon RN, Kannourakis G. A cryptic deletion in 5q31.2 provides histone demethylases. Nat Rev Mol Cell Biol. 2012;13(5):297. further evidence for a minimally deleted region in myelodysplastic 41. Kim TH, Abdullaev ZK, Smith AD, Ching KA, Loukinov DI, Green R, et al. syndromes. Cancer Genet. 2011;204(4):187–94. Analysis of the vertebrate insulator protein CTCF-binding sites in the 21. Vajen B, Thomay K, Schlegelberger B. Induction of chromosomal insta- human genome. Cell. 2007;128(6):1231–45. bility via telomere dysfunction and epigenetic alterations in myeloid 42. Igor C, Shaharum S, Sung Yun K, Rosita BM, Yoo-Wook K, Wenqiang Y, et al. neoplasia. Cancers. 2013;5(3):857–74. CTCF interacts with and recruits the largest subunit of RNA polymerase II 22. Kim JY, Kim KB, Eom GH, Choe N, Kee HJ, Son HJ, et al. KDM3B is the H3K9 to CTCF target sites genome-wide. Mol Cell Biol. 2007;27(5):1631. demethylase involved in transcriptional activation of lmo2 in leukemia. 43. Ohlsson R, Bartkuhn MR. CTCF shapes chromatin by multiple mecha- Mol Cell Biol. 2012;32(14):2917. nisms: the impact of 20 years of CTCF research on understanding the 23. Kasioulis I. Dissecting the biological roles of Kdm3b and Kdm3a lysine workings of chromatin. Chromosoma. 2010;119(4):351–60. demethylases. Edinburgh: The University of Edinburgh; 2015. 44. Sanjeev S, Ersen K, Melissa G, Masahiko I, Bojan S, Mikhail K, et al. CTCF- 24. Xu X, Nagel S, Quentmeier H, Wang Z, Pommerenke C, Dirks WG, et al. promoted RNA polymerase II pausing links DNA methylation to splicing. KDM3B shows tumor-suppressive activity and transcriptionally regulates Nature. 2011;479(7371):74–9. HOXA1 through retinoic acid response elements in acute myeloid leuke- 45. Merkenschlager M, Nora EP. CTCF and cohesin in genome folding mia. Leuk Lymphoma. 2018;59(1):1. and transcriptional gene regulation. Annu Rev Genomics Hum Genet. 25. Kankan W, Ping W, Jiantao S, Xuehua Z, Miaomiao H, Xiaohong J, et al. 2016;17(1):17. PML/RARalpha targets promoter regions containing PU.1 consensus 46. Luo H, Wang F, Zha J, Li H, Yan B, Du Q, et al. CTCF boundary remodels and RARE half sites in acute promyelocytic leukemia. Cancer Cell. chromatin domain and drives aberrant HOX gene transcription in acute 2010;17(2):186. myeloid leukemia. Blood. 2018;132(8):837–48. 26. Qian M, Jin W, Zhu X, Jia X, Yang X, Du Y, et al. Structurally diferentiated 47. Shen Q, Uray I, Li Y, Krisko T, Strecker T, Kim H, Brown P. The AP-1 transcrip- cis-elements that interact with PU1 are functionally distinguishable in tion factor regulates breast cancer cell growth via cyclins and E2F factors. acute promyelocytic leukemia. J Hematol Oncol. 2013;6(1):25. Oncogene. 2008;27(3):366. 27. Andricovich J, Kai Y, Tzatsos A. Lysine-specifc histone demethylases in 48. Fan F, Bashari MH, Morelli E, Tonon G, Malvestiti S, Vallet S, et al. The normal and malignant hematopoiesis. Exp Hematol. 2016;44(9):778–82. AP-1 transcription factor JunB is essential for multiple myeloma cell 28. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using proliferation and drug resistance in the bone marrow microenvironment. real-time quantitative PCR and the 2( Delta Delta C(T)) method. Meth- Leukemia. 2017;31(7):1570–81. ods. 2001;25(4):402–8. − 49. Wahlestedt M, Säwén P, Ladopoulos V, Norddahl G, Gottgens B, Bryder D, 29. Wu J, Huang B, Chen H, Yin Q, Liu Y, Xiang Y, et al. The landscape of et al. Critical modulation of hematopoietic lineage fate by the PAR/bZIP accessible chromatin in mammalian preimplantation embryos. Nature. transcription factor hepatic leukemia factor. Exp Hematol. 2014;42(8):S64. 2016;534(7609):652. 50. De TH, Pandolf PP, Chen Z. Acute promyelocytic leukemia: a paradigm for 30. Buenrostro JD, Wu B, Chang HY, Greenleaf WJ. ATAC-seq: a method for oncoprotein-targeted cure. Cancer Cell. 2017;32(5):552–60. assaying chromatin accessibility genome-wide. Curr Protoc Mol Biol. 51. Lilljebjorn H, Heidenblad M, Nilsson B, Lassen C, Horvat A, Heldrup J, et al. 2015;109:21–9. Combined high-resolution array-based comparative genomic hybridiza- 31. Yu G, Wang LG, Han Y, He QY. clusterProfler: an R package for com- tion and expression profling of ETV6/RUNX1-positive acute lympho- paring biological themes among gene clusters. Omics J Integr Biol. blastic leukemias reveal a high incidence of cryptic Xq duplications 2012;16(5):284–7. and identify several putative target genes within the commonly gained 32. Li S, Ali S, Duan X, Liu S, Du J, Liu C, et al. JMJD1B demethylates region. Leukemia. 2007;21(10):2137–44. H4R3me2 s and H3K9me2 to facilitate gene expression for development 52. Stirewalt DL, Meshinchi S, Kopecky KJ, Fan W, Pogosova-Agadjanyan EL, of hematopoietic stem and progenitor cells. Cell Rep. 2018;23(2):389. Engel JH, et al. Identifcation of genes with abnormal expression changes 33. Akihiro T, Hitoshi K, Tomoki N. Mechanisms of action and resistance to all- in acute myeloid leukemia. Genes Chromosom Cancer. 2008;47(1):8–20. trans retinoic acid (ATRA) and arsenic trioxide (As2O 3) in acute promyelocytic leukemia. Int J Hematol. 2013;97(6):717–25. 34. Doucas V, Brockes JP, Yaniv M, Thé H, Dejean A. The PML-retinoic acid Publisher’s Note receptor alpha translocation converts the receptor from an inhibitor to a Springer Nature remains neutral with regard to jurisdictional claims in pub- retinoic acid-dependent activator of transcription factor AP-1. Proc Natl lished maps and institutional afliations. Acad Sci USA. 1993;90(20):9345–9. 35. Sadia S, Colin L, Kees-Jan F, Gianmaria F, Mauro R, Nielsen FG, et al. Chromatin accessibility, p300, and histone acetylation defne PML-RARα