Leukemia (2021) 35:90–106 https://doi.org/10.1038/s41375-020-0808-y ARTICLE Acute lymphoblastic leukemia H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell marker PROM1/CD133 in MLL-AF4 leukemia cells Laura Godfrey1 Nicholas T. Crump 1 Sorcha O’Byrne2 I-Jun Lau1 Siobhan Rice1 Joe R. Harman1 ● ● ● ● ● ● Thomas Jackson2 Natalina Elliott2 Gemma Buck2 Christopher Connor3 Ross Thorne1 David J. H. F. Knapp4 ● ● ● ● ● ● Olaf Heidenreich 5,6 Paresh Vyas 1,7 Pablo Menendez8,9,10 Sarah Inglott3 Philip Ancliff3 Huimin Geng11 ● ● ● ● ● ● Irene Roberts1,2 Anindita Roy2 Thomas A. Milne 1 ● ● Received: 15 August 2019 / Revised: 12 March 2020 / Accepted: 18 March 2020 / Published online: 2 April 2020 © The Author(s) 2020. This article is published with open access Abstract MLL gene rearrangements (MLLr) are a common cause of aggressive, incurable acute lymphoblastic leukemias (ALL) in 1234567890();,: 1234567890();,: infants and children, most of which originate in utero. The most common MLLr produces an MLL-AF4 fusion protein. MLL-AF4 promotes leukemogenesis by activating key target genes, mainly through recruitment of DOT1L and increased histone H3 lysine-79 methylation (H3K79me2/3). One key MLL-AF4 target gene is PROM1, which encodes CD133 (Prominin-1). CD133 is a pentaspan transmembrane glycoprotein that represents a potential pan-cancer target as it is found on multiple cancer stem cells. Here we demonstrate that aberrant PROM1/CD133 expression is essential for leukemic cell growth, mediated by direct binding of MLL-AF4. Activation is controlled by an intragenic H3K79me2/3 enhancer element (KEE) leading to increased enhancer–promoter interactions between PROM1 and the nearby gene TAPT1. This dual locus regulation is reflected in a strong correlation of expression in leukemia. We find that in PROM1/CD133 non-expressing cells, the PROM1 locus is repressed by polycomb repressive complex 2 (PRC2) binding, associated with reduced expression of TAPT1, partially due to loss of interactions with the PROM1 locus. Together, these results provide the first detailed analysis of PROM1/CD133 regulation that explains CD133 expression in MLLr ALL. Introduction CD133, encoded by the PROM1 gene, is a pentaspan trans- These authors contributed equally: Laura Godfrey, Nicholas T. Crump membrane glycoprotein of great potential value as a pan- Supplementary information The online version of this article (https:// cancer target as it is commonly associated with cancer stem doi.org/10.1038/s41375-020-0808-y) contains supplementary cells in multiple different tumor types, including leukemia material, which is available to authorized users. * Anindita Roy 5 Princess Maxima Centrum for Pediatric Oncology, Utrecht, The anindita.roy@paediatrics.ox.ac.uk Netherlands * Thomas A. Milne 6 Wolfson Childhood Cancer Research Centre, Newcastle thomas.milne@imm.ox.ac.uk University, Newcastle upon Tyne, UK 7 Department of Haematology, Oxford University Hospitals NHS 1 MRC Molecular Haematology Unit, MRC Weatherall Institute of Foundation Trust, Oxford, UK Molecular Medicine, NIHR Oxford Biomedical Research Centre 8 Haematology Theme, Radcliffe Department of Medicine, Josep Carreras Leukemia Research Institute, Barcelona, Spain University of Oxford, Oxford, UK 9 Institucio Catalana of Recerca i Estudis Avançats (ICREA), 2 Department of Paediatrics, University of Oxford, Oxford, UK Barcelona, Spain 10 3 Great Ormond Street Hospital for Children, London, UK Centro de Investigación Biomédica en Red en cancer (CIBERONC)-ISCIII, Barcelona, Spain 4 MRC Weatherall Institute of Molecular Medicine, Radcliffe 11 Department of Medicine, University of Oxford, Oxford, UK Department of Laboratory Medicine, University of California, San Francisco, San Francisco, CA 94143, USA H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 91 [1, 2]. Proof-of-principle studies have shown that targeting the presence of MLL-AF4 through H3K79me2/3-mediated CD133 can be used to deliver nanoparticles to gastric stem enhancer–promoter interactions, and that PROM1 expres- cells [3], or for chimeric antigen receptor T cell therapy in sion contributes to the leukemic growth of CD133+ cells. acute lymphoblastic leukemias (ALL) caused by rearrange- These results for the first time elucidate the mechanism of ments of the Mixed Lineage Leukemia (MLL) gene [4]. PROM1/CD133 upregulation and provide an explanation Despite vast improvements in treatment for ALL, MLL for the widespread expression of CD133 in MLLr ALLs. It gene rearrangements (MLLr) still cause very poor prognosis also provides an interesting paradigm illustrating how ALLs especially in infants [5–9]. The most common MLL repression of a single locus (PROM1) can impact regulation rearrangement is the t(4;11) (q21;q23) chromosome translo- of a nearby locus (TAPT1) by suppressing access to key cation that fuses MLL in frame with the AF4 gene producing enhancer sequences. MLL-AF4 and AF4-MLL fusion proteins [7, 10–12]. MLL- AF4 and other MLL fusion proteins (MLL-FPs) bind to gene targets and cause inappropriate gene activation through Materials and methods multiple transcription elongation and epigenetic mechanisms [13], including recruitment of the histone H3 lysine-79 Cell culture and cell lines (H3K79) methyltransferase DOT1L [7, 14–17]. In addition to a role in transcription elongation, recent work from our lab SEM (MLL-AF4 B-ALL line [22]), ML-2 (MLL-AF6 AML has shown that H3K79me2/3 has an important role at a subset line [23]), SHI-1 (MLL-AF6 AML line [24]), and RCH- of enhancers (H3K79me2/3-marked enhancer elements ACV (E2A-PBX1 cALL line [25]) cells were purchased (KEEs)), increasing expression of key gene targets through from DMSZ (www.cell-lines.de) and cultured in Iscove’s the maintenance of enhancer–promoter interactions [18]. Modified Dulbecco’s Medium (IMDM) supplemented with One of the most attractive features of PROM1/CD133 as 10% fetal calf serum (FCS, Gibco) and Glutamax (Ther- a potential therapeutic target derives from the recognition moFisher Scientific). RS4;11 (MLL-AF4 B-ALL line), that the gene is a direct target of MLL-AF4 regulation THP1 (MLL-AF9 AML line), MV4;11 (MLL-AF4 AML [17, 19, 20], suggesting that in MLLr leukemias PROM1/ line), and CCRF-CEM (T-ALL line) cells were purchased CD133 expression is tightly linked to the activity of the from ATCC (www.lgcstandards-atcc.org). RS4;11, THP1, fusion protein itself. However, the exact details of how this and CCRF-CEM were cultured using RPMI with 10% FCS locus is regulated by MLL-AF4 is unclear, and whether and and Glutamax. MV4;11 cells were cultured as per the SEM how PROM1/CD133 contributes to MLLr leukemic growth conditions. All cell lines were confirmed free from myco- is unknown. Understanding these mechanisms is likely to plasma contamination. be key to the future development of PROM1/CD133- directed therapeutic targeting in these leukemias. Cell culture drug treatment To better understand the role of PROM1/CD133 in MLLr leukemias, we have systematically characterized the struc- Cells were treated with EPZ-5676 as previously described ture of the PROM1 gene locus and the mechanism of [17, 18]. Briefly, SEM and RS4;11 cells were treated with 2 PROM1/CD133 regulation for the first time. Importantly, µM EPZ-5676 or dimethyl sulphoxide (control). Cells were we show that two CD133-expressing MLL-AF4 cell lines treated for 7 days with media and EPZ-5676 changed on (SEM and RS4;11) are addicted to its continued expression. days 3 and 6 and diluted to 0.5 × 106 cells/ml and 0.8 × 106 Using a high-resolution chromosome conformation capture cells/ml, respectively. Cells were collected on day 7 and technique (next generation Capture-C [21]) to analyze the processed for downstream applications. Cells were treated enhancer–promoter structure of the PROM1/CD133 locus, for 5 days with UNC1999 at 5 µM and GSK126 at 2 µM, we find that PROM1 expression is activated by a series of with cells split and fresh media and inhibitor added on day intragenic KEEs within PROM1 as well as in the nearby 3. Venetoclax was used at 0.5 µM for 24 h, and camp- TAPT1 gene. As with other well-characterized KEEs [18], tothecin was used at 1 µM for 1 h. we demonstrate that MLL-AF4 aberrantly upregulates PROM1 transcription by controlling enhancer–promoter Primograft interactions via increased H3K79 methylation and H3K27 acetylation. In contrast, we find that in CD133-negative For details of sample preparation see Ref. [17]. leukemia cells the PROM1 promoter is bound by compo- nents of the polycomb repressive complex 2 (PRC2) and Patient samples this is associated with a lack of enhancer features, including enhancer–promoter interactions. Overall, our results show Infant and pediatric ALL samples were obtained from that PROM1/CD133 expression is directly upregulated by patients being treated at Great Ormond Street Hospital for 92 L. Godfrey et al. Children, London after informed consent, and analyzed as AAGGGGCTAG. SEM cells were lentivirally transduced part of their diagnostic workup. Additional MLLr ALL with either shRNA plasmid and clonal lines were estab- samples were obtained from Bloodwise Childhood Leu- lished using 0.5 µg/ml puromycin. kaemia Cell Bank, UK (REC: 16/SW/0219). All ALL samples were anonymised at source, assigned a unique Colony-forming assay study number and linked. Donated fetal tissue was provided by the Human Developmental Biology Resource (www. Wild-type or PROM1 shRNA SEM cell lines were treated hdbr.org, covered under REC: 08/H0906/21+5) regulated with 0.5 µg/ml doxycycline 24 h prior to colony assay by the UK Human Tissue Authority (www.hta.gov.uk). plating. Five hundred doxycycline-treated or -untreated Second-trimester fetal bone marrow (BM) and liver was cells were plated in IMDM MethoCult media with 20% processed as previously described [26]. CD34+ and FCS (H4100; STEMCELL Technologies) per dish, in tri- CD34− populations were isolated using magnetic beads plicate, with 0.5 µg/ml doxycycline added to the media (Miltenyi). where appropriate. Colonies were incubated for 14–16 days (37 °C, 5% CO2) and then counted. For replating, after Flow cytometry analysis counting cells were washed in IMDM, and 500 cells were added to fresh IMDM MethoCult media, with doxycycline Cells were stained with fluorophore-conjugated monoclonal added where appropriate, and plated for 14 days. antibodies in PBS with 2% FBS and 1 mM EDTA for 30 min and analyzed using BD LSR II or Fortessa X50 Cell cycle analysis instruments. For viability eBioscience Fixable Viability Dye eFluor 506 (ThermoFisher Scientific) or Hoechst 33342 1 × 106 cells were fixed in 70% ethanol and incubated at 4 ° (ThermoFisher Scientific) was used. Annexin V binding C overnight, then stained with FxCycle PI/RNase staining was assayed using the eBioscience Annexin V Apoptosis solution (ThermoFisher Scientific). Cells were analyzed Detection Kit APC (ThermoFisher Scientific). Antibodies using 532 nm excitation and collected using 585/42 band- used are detailed in Supplementary Table 1. Analysis was pass. Analysis was performed using FlowJo software. performed using FlowJo software with gates set using unstained and fluorescence-minus-one controls. Immunofluorescence siRNA knockdown Cells were prepared for imaging by attachment to coverslips coated with 0.01% Poly-L-Lysine and fixed for 15 min with siRNA knockdowns were performed in SEM and 4% PFA. Cells were incubated with anti-γH2AX primary RS4;11 cells as described [17]. For PROM1 KD, Ambion antibody (Cell Signaling Technology 9178) for 1 h at room Silencer select siRNAs (ThermoFisher Scientific) were temperature. Cells were then washed with PBS and incu- used: siRNA 1: s16879; siRNA 2: s16879; NT: Silencer bated with secondary antibody for 30 min at room tem- Negative Control No. 2. For MLL-AF4 KD, NT siRNA: perature. Cells were washed with PBS and mounted onto sense AAAAGCUGACCUUCUCCAAUG; antisense slides using Vectashield with DAPI. Images were acquired CAUUGGAGAAGGUCAGCUUUUCU. MLL-AF4 KD at room temperature using a Zeiss AXIO Observer.Z1 siRNA: sense AAGAAAAGCAGACCUACUCCA; anti- inverted microscope equipped with a Zeiss LSM‐880 con- sense UGGAGUAGGUCUGCUUUUCUUUU. focal system using a Plan-Apochromat 63x/1.4 Oil DIC M27 objective. Images were then processed using OMERO SEM PROM1/CD133 shRNA lines (v5.4.9). shRNA sequences, which were designed using the siRNA Western blotting targeting sequences, were cloned into a doxycycline- inducible shRNA vector (miRE_18_LT3GEPIR) using the Salt-soluble protein and histone extraction from 1 × 106 following sequences. PROM1 shRNA 1: GAAGGTATA SEM cells was conducted as previously described [18], TTGCTGTTGACAGTGAGCGGTTTACCGCAAAAAT followed by western blotting [27]. Antibodies used are TAGAGCCTAGTGAAGCCACAGATGTAGGCTCTAA detailed in Supplementary Table 1. TTTTTGCGGTAAAATGCCTACTGCCTCGGACTTCA AGGGGCTAG; PROM1 shRNA 2: TCGAGAAGGTA qRT-PCR TATTGCTGTTGACAGTGAGCGGCATTTGTTACAGC AACGACACTAGTGAAGCCACAGATGTAGTGTCGT For qPCR applications, RNA was extracted from 1 × 106 TGCTGTAACAAATGTTGCCTACTGCCTCGGACTTC cells using the RNeasy Mini Kit (Qiagen). RNA was reverse H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 93 transcribed using Superscript III (ThermoFisher Scientific) ChIP-qPCR and ChIP-seq with random hexamer primers (ThermoFisher Scientific). Samples were analyzed by Taqman qPCR, using the The full protocol is described in [17, 18]. Briefly, 107–108 housekeeping gene YWHAZ for gene expression normal- cells were sonicated (Covaris) following the manufacturer’s ization. For list of qPCR primers used see Supplementary protocol. Magnetic protein A and G beads (ThermoFisher Table 2. Scientific) were used to isolate antibody-chromatin com- plexes. Antibodies used are detailed in Supplementary Poly(A) RNA purification and sequencing Table 1. Beads were washed three times using a solution of 50 mM HEPES-KOH (pH 7.6), 500 mM LiCl, 1 mM RNA was extracted from 1 × 107 PROM1 shRNA 1 SEM EDTA, 1% NP40 and 0.7% sodium deoxycholate, and once cells lines using the Directzol RNA MiniPrep Kit (Zymo with Tris-EDTA. Samples were eluted and Proteinase R2050). Poly(A) purification was conducted using the NEB K/RNase A-treated. Samples were purified using the Qiagen Poly(A) mRNA magnetic isolation module following the PCR purification kit. DNA content was analyzed by qPCR manufacturer’s protocol. Library preparation was carried or ChIP-sequencing. For list of qPCR primers used see out using the Ultra II Directional RNA Library Prep Kit Supplementary Table 2. For ChIP-seq, DNA libraries were (NEB, E7765). RNA libraries were sequenced by paired- generated using the NEBnext Ultra DNA library preparation end sequencing using a 150 cycle high output kit on a kit for Illumina (NEB). Libraries were sequenced by paired- Nextseq 500 (Illumina). end sequencing using a 75 cycle high output kit on a Nextseq 500 (Illumina). Gene expression analysis ATAC-seq Following sequencing, QC analysis was conducted using the fastQC package (http://www.bioinformatics.babraham. ATAC-seq was performed using the Nextera Tn5 transpo- ac.uk/projects/fastqc). Reads were mapped to the human sase (Illumina) on 6 × 104 THP1 cells and whole BM from genome assembly hg19 using STAR [28]. PCR duplicates an MLL-AF4 ALL pediatric patient, as previously descri- were removed from the mapped reads using Samtools [29]. bed [18, 39]. Libraries were sequenced by paired-end The featureCounts function from the Subread package was sequencing using a 75 cycle high output kit on a Nextseq used to quantify gene expression levels using standard 500 (Illumina). parameters. This was used to identify differential gene expression globally, generating RPKM values, using the Sequence analysis edgeR package [30]. Differential gene expression was defined by an adjusted p value (FDR) of <0.05. Infant ALL For ChIP-seq and ATAC-seq, quality control of FASTQ [31–33] and normal BM [34–36] RNA-seq datasets were reads, alignment, PCR duplicate filtering, blacklisted region analyzed as described previously [36]. filtering, and UCSC data hub generation was performed using an in-house pipeline (https://github.com/Hughes- Single-cell gene expression analysis Genome-Group/NGseqBasic/releases) as described [18]. The Homer tool makeBigWig.pl command was used to Single-cell data was analyzed using the Seurat package [37] generate bigwig files for visualization in UCSC, normal- following standard methods. Briefly, a random subset of 105 izing tag counts to tags per ten million. Gene profiles were cells were chosen for analysis. Cells were initially filtered generated using annotatePeaks.pl. removing those that contained less than 200 and more than 2500 genes. Cells with >5% mitochondrial genes were Capture-C removed. Genes detected in less than four cells were removed. Data were normalized using the log normalize For details on Capture-C methods used see Refs. [18, 21]. method with a scale factor of 10,000. FindVariableFeatures, Biotinylated oligo probes, with the following sequences, RunPCA, and RunTSNE were used for dimension reduc- were designed to the PROM1 and TAPT1 promoters. tion. FindNeighbours and FindCluster functions were used PROM1_F: GCGGCTTCCTCTGTCCCGGACGGGGAC to find cell clusters. FindAllMarkers was used to marker CTAGGTATGGGGCCGCCGTAATGGAAAGGATTCC genes of each cluster and cell types were identified using TTAAACATACTCACCGCGGCGGGAGAGCTGAGAG published marker signatures [38] and a custom gene voting CATGGCCAGGTGCCGCGTGGGGATC; PROM1_R: system. Cell clusters and gene expression were visualized GATCCCGAGCCTCTGACTTTCTACAGCCGTGAAG using DimPlot and FeatureScatter functions. CTCCTGGTCTGTCCACGCTCCTCTTTGTTGTCCGG 94 L. Godfrey et al. TCGGCGTGTTCTGGCAGGGTGCGCCTTGAGCACC that it is the active transcription of the PROM1 locus itself CCAGTTCTCCATCTAGC; TAPT1_F: GATCAACTTT that may contribute to the upregulation of TAPT1. ATGCCCCTTAAGTAGTTTTAAAATTTCAGTACTCA To determine if heterogeneity in CD133 expression was AGTACAATCAATGATGTAATCAGCCTCACAACTT observed in ALL patient samples, we analyzed 44 primary AGTTATTTGCATGAACTTCAGGAAATAACGGCTA precursor B-ALL samples: 12 MLLr and 32 non-MLLr ACACAGC;TAPT1_R: GTGTCCGGCGGTCCGCTCA (sorting strategy shown in Supplementary Fig. 1e). GGGCCTCTTTTGCAGACTCGGTGCCCGGAGTGCG Although there was a wide range of CD133 expression CCGGCGCCGCCCGCCAGGTCTTGGCACTGCTGGC (measured by the proportion of blasts with detectable GCGGCCGCGGCGGCGGGGGCCCGGCTCCAGATC. CD133) in both MLLr and non-MLLr ALL, mean and Data analysis was performed using an in-house pipeline median levels of CD133 expression were markedly higher in (https://github.com/Hughes-Genome-Group/CCseqBasicF/ MLLr cases (Fig. 1b, p < 0.0001, mean 64.9% ± 10.2 s.e.m. releases; Davies 2016) and statistical analysis was per- vs. 15.7% ± 4.2 s.e.m., median 74.9% vs. 2.5% for MLLr formed as described [18]. and non-MLLr blasts, respectively; Supplementary Fig. 1f), and a significantly greater number of MLLr ALL samples Statistical analysis (9/12) expressed CD133 on the majority of cells (>50%) compared with 4/32 non-MLLr ALL. To correlate this with Statistical analyses used and sample sizes are indicated in gene expression patterns, we analyzed RNA-seq data from figure legends; n numbers refer to independent experiments published datasets of large cohorts of infant and childhood (for cell lines) or biological samples (for patient data). All ALL patient blasts [32, 33]. In keeping with previous tests were conducted two-tailed. Samples were only exclu- findings [40], MLLr leukemia cells were found to ded from analysis if positive and negative controls did not express PROM1 at much higher levels than the non-MLLr give the expected results. cohort (Fig. 1c). To test whether the specific MLL fusion partner con- tributes to PROM1 expression status we analyzed two ALL Results datasets for which fusion partner status was available [31, 32]. Strikingly, whilst MLL-AF4 and MLL-ENL ALL MLLr leukemia cell lines and primary ALL blasts samples displayed high levels of PROM1 expression, other show heterogeneous expression of PROM1/CD133 fusion proteins, including MLL-AF9, were associated with lower levels of expression, more comparable to non-MLLr In order to model PROM1/CD133 behavior and regulation, leukemias (Supplementary Fig. 1g). Taken together, these we first evaluated CD133 expression by flow cytometry to data indicate that although a CD133+ immunophenotype is identify CD133-positive and -negative leukemia cell lines not restricted to MLLr ALL, PROM1/CD133 is aberrantly (Fig. 1a). As also observed by others [4], while two MLL- expressed in a large subset of cell lines and MLLr leukemia AF4 ALL cell lines (SEM and RS4;11) displayed high patients, particularly those associated with MLL-AF4 and levels of CD133 expression, CD133 was undetectable on MLL-ENL fusion proteins. the surface of THP1 (MLL-AF9 AML), MV4;11 (MLL- AF4 AML), and RCH-ACV (non-MLLr ALL) cells CD133 is important for MLL-AF4 ALL (Fig. 1a). This was confirmed at the RNA level by qRT- leukemogenesis PCR (Supplementary Fig. 1a). Interestingly, PROM1 demonstrated a similar expression pattern to the neighbor- To test whether CD133 is important for MLL-AF4 leuke- ing gene TAPT1 (Supplementary Fig. 1a, b). Although mia growth we generated two SEM cell lines containing TAPT1 is expressed even when PROM1 is not (albeit at different doxycycline-inducible shRNA sequences targeting lower levels), higher PROM1 expression correlated with PROM1. Levels of PROM1 RNA were significantly higher TAPT1 expression in the cell lines analyzed (Sup- reduced 48 h after induction (Fig. 1d). To test whether plementary Fig. 1a). We observed a similar correlation of PROM1 is important for leukemogenesis we performed expression in a published dataset [32] of leukemia patient colony-forming assays on cells after 24 h doxycycline samples (Supplementary Fig. 1c, R2 = 0.4251), confirming treatment. A reduction in CD133 expression prior to colony this phenomenon is not restricted to cell lines and sug- assay (day 1) was verified by flow cytometry (Fig. 1e and gesting the genes may be co-regulated in leukemia. Cell Supplementary Fig. 1h), and levels were further decreased surface expression of CD133 does not cause upregulation of in the cells post-colony assay (day 17). We observed a TAPT1, as transfection of CD133-negative THP1 cells with striking reduction in colony-forming ability following a PROM1 expression plasmid did not result in an increase in PROM1 knockdown using either shRNA, and not with TAPT1 expression (Supplementary Fig. 1d). This suggests doxycycline-treated control SEM cells, with the number of H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 95 a b **** c Andersson et al. Agraz-Doblas et al. 100 200 400 SEM *** * PROM1 expression (TPM) PROM1 expression (TPM) normalized to mode RS4;11 150 300 % CD133+ THP1 50 100 200 MV4;11 RCH-ACV 50 100 -3 3 10 0 10 104 105 CD133 0 0 0 MLLr ALL non-MLLr ALL MLLr non-MLLr MLLr non-MLLr d ns *** ** e 100 Con 100 Con Day 1 Day 17 gene expression relative 1.5 +Dox normalized to mode +Dox to YWHAZ/Control 80 80 60 60 1.0 40 40 0.5 20 20 0 0 -10-3 0 103 104 105 -10-3 0 103 104 105 0.0 CD133 CD133 24h 48h 72h f ns ** ** g 300 j 25 WT Con downregulated genes Con 100 +Dox WT +Dox 20 cumulative cell % -Dox colonies number x106 200 shRNA 1 Con -log10(Padj) 15 shRNA 1 +Dox 50 shRNA 2 Con 10 100 shRNA 2 +Dox 5 0 0 0 RNA transport Cellular senescence Human T-cell leukemia virus 1 infection Fanconi anemia pathway Spliceosome Cell cycle Herpes simplex virus 1 infection DNA replication Colorectal cancer Mismatch repair Control shRNA 1 shRNA 2 0 3 6 9 12 time (days) h 80 ** *** ns i 2 865 upregulated Con +Dox 60 1 % cells logFC 40 0 20 -1 -2 1147 downregulated 0 G1 S G2 0 4 8 12 logCPM colonies reduced to <50% of the uninduced control AF4 CD133+ cell line, RS4;11 (Fig. 1a), where siRNA- (Fig. 1f). This effect was reproducible following two rounds mediated knockdown of PROM1 (Supplementary Fig. 2a) of serial replating of colonies (Supplementary Fig. 1i). had a similarly dramatic effect on cell growth (Supple- Thus, CD133 is clearly an important protein for the leu- mentary Fig. 2b). The loss of cell growth following PROM1 kemic potential of SEM cells. KD may be explained by a reduced growth rate and/or The inability of PROM1 KD SEM cells to form colonies increased apoptosis. We tested the former possibility by was matched by a dramatically reduced growth rate in performing cell cycle analysis of SEM cells after shRNA liquid culture (Fig. 1g). We confirmed this in another MLL- induction (Fig. 1h and Supplementary Fig. 2c). This 96 L. Godfrey et al. Fig. 1 CD133 expression is essential for MLL-AF4 ALL cell downregulated genes were enriched for genes involved in growth. a Flow cytometry data showing CD133+ expression in dif- the cell cycle and other processes associated with dereg- ferent leukemia cell lines. Representative of two biological replicates. ulation of the cell cycle, including DNA replication b Flow cytometry data showing the percentage of CD133+ cells in blast populations from 12 MLLr and 32 non-MLLr infant and child- (Fig. 1j). We also observed downregulation of DNA hood ALL patients. Box represents median and interquartile range; repair genes, including genes involved in the Fanconi whiskers show minimum and maximum values, ****p < 0.0001. Mean anemia pathway and mismatch repair (Fig. 1j), which values: 64.9% ± 10.2 s.e.m. and 15.7% ± 4.2 s.e.m.; median: 74.9% may be a consequence of the reduced S phase (Fig. 1h) as and 2.5% for MLLr and non-MLLr blasts, respectively. c Expression of PROM1 by RNA-seq from two independent published datasets Left: these processes are associated with DNA replication. expression of PROM1 in MLL-AF4 ALL patients compared with non- This was not associated with an increase in double-strand MLLr ALL patient blasts from Ref. [32], ***p < 0.001; right: breaks, as measured by γH2AX levels (Supplementary expression of PROM1 in MLLr ALL patients compared with non- Fig. 2f, g). MLLr ALL patient blasts from Ref. [33], *p < 0.05. d qRT-PCR of PROM1 in untreated control (black) and doxycycline-treated (orange) PROM1 shRNA 1 SEM cells after 24, 48, and 72 h. Error bars MLL-AF4 directly binds to and regulates PROM1 represent s.e.m. of three biological replicates, ***p < 0.001, **p < expression 0.01, ns = no significant difference. e Flow cytometry analysis showing CD133 level in PROM1 shRNA 1 SEM cell line in control (black) and induced (orange) conditions at day 1 (24 h doxycycline To understand how PROM1 expression is regulated in treatment) and day 17 (following colony assay). Histograms are CD133+ MLLr leukemias, we analyzed ChIP-seq datasets representative of three replicates. f Colony assay showing percentage from multiple cell types, using MLL-N and AF4-C binding of colonies formed in control and PROM1 shRNA 1 and 2 SEM cell as a proxy for MLL-AF4 fusion protein binding. In line lines in the presence of 0.5 µg/ml doxycycline (+Dox) compared with uninduced control (Con = −Dox). Error bars represent s.d. of three with previous findings [17, 20, 27, 41], we identified MLL- (control) or four (shRNA) biological replicates, **p < 0.01, ns = no AF4 binding at the promoter and spreading into the gene significant difference. g Growth curve showing cumulative cell num- body of PROM1 in CD133+ ALL cell lines (SEM and ber for control (circles) and PROM1 shRNA 1 (triangles) and 2 RS4;11; see Fig. 1a) and in MLL-AF4 blasts from a CD133 (squares) SEM cell lines grown in the presence (+Dox, orange) or absence (Con, black) of 0.5 µg/ml doxycycline. Cells were split into + ALL patient primograft [17] (Fig. 2a and Supplementary fresh medium every 3 days. Error bars represent s.e.m. of three bio- Fig. 3a). Published FLAG-MLL-Af4 ChIP-seq from MLL- logical replicates. h Cell cycle analysis using flow cytometry in control Af4-transformed CD34+ cord blood cells [20] revealed a (black) and 72 h post induction (+Dox, orange) of PROM1 shRNA 1 similar binding distribution at the PROM1 locus (Fig. 2a, cell line. Error bars represent s.e.m. from three biological replicates, ***p < 0.001, **p < 0.01, ns = no significant difference. i MA plot bottom panel). Notably, in all four cell types the neigh- showing differentially expressed genes from RNA-seq 72 h post boring TAPT1 gene was also marked by a broad domain of induction of PROM1 shRNA 1. Statistically significant differences MLL-AF4 binding (Fig. 2a). H3K79me2/3, a histone (865 upregulated: red; 1147 downregulated: blue; and 10,385 modification found at high levels at MLL-FP gene targets unchanged: black) from three biological replicates, FDR < 0.05. j KEGG analysis of significantly downregulated genes from RNA-seq [16, 17], was observed to colocalize with MLL-AF4 at analysis, displaying ten most enriched processes. PROM1 in SEM, RS4;11, and the MLL-AF4 ALL primo- graft cells, further validating the gene as a bona fide MLL- AF4 target (Fig. 2b). revealed a shift in cell cycle status following PROM1 Binding of MLL-AF4 to PROM1 suggests that it may depletion, with an increase in the proportion of cells in G1 directly regulate PROM1 transcription. To test this, we phase and a decrease in S phase, suggesting a delay in interrogated nascent RNA-seq data from SEM cells fol- transition between the stages (Fig. 1h and Supplementary lowing siRNA-mediated knockdown of MLL-AF4 [17, 18], Fig. 2c). We tested for apoptosis by Annexin V staining which produced clear reductions in MLL-N and AF4-C (Supplementary Fig. 2d) and western blot for cleaved PARP binding across PROM1 and TAPT1, as well as a known (Supplementary Fig. 2e), a known caspase 3 substrate. MLL-AF4 gene target, BCL11A (Fig. 2c) [17, 18]. We These demonstrated an increase in apoptosis following observed a dramatic reduction in PROM1 transcription PROM1 shRNA induction, but not doxycycline treatment of following MLL-AF4 knockdown (p < 0.001, Fig. 2d and control SEM cells (Supplementary Fig. 2d, e). Taken Supplementary Fig. 3b), confirmed by qRT-PCR (Supple- together, these data suggest that PROM1 is likely involved mentary Fig. 3c), demonstrating that MLL-AF4 is required in promoting both cell growth and survival. for expression of PROM1. MLL-AF4 knockdown also To further investigate the dependence of SEM cells on reduced the expression of TAPT1, consistent with the idea CD133 expression, we performed RNA sequencing in the that the two genes are co-regulated by MLL-AF4 (p < PROM1 shRNA cell line. Differential gene expression 0.001, Fig. 2d and Supplementary Fig. 3b, c). Taken toge- analysis revealed 865 up- and 1147 down-regulated genes ther, these data demonstrate that PROM1 is directly bound 72 h after PROM1 shRNA induction (Fig. 1i). Consistent and transcriptionally regulated by MLL-AF4 in CD133+ with the disrupted cell cycle of these cells, the MLL-AF4 ALL. H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 97 a 100 kb c 2.0 NT siRNA FGFBP2 PROM 1 BC010030 TAPT1 MLL-AF4 KD 110 SEM MLL-N 1.5 % Input MLL-N 1.0 81 SEM AF4-C MLL-N and AF4-C ChIP-seq 0.5 78 RS4;11 MLL-N 0.0 CON BCL11A P1 P2 P3 T1 T2 T3 52 RS4;11 AF4-C PROM1 TAPT1 2.5 25 Primograft MLL-N 2.0 1.5 % Input AF4-C 38 Primograft AF4-C 1.0 74 FLAG-MLL-Af4 0.5 0.0 CON BCL11A P1 P2 P3 T1 T2 T3 P3 P2 P1 T3 T2 T1 PROM1 TAPT1 b e 3.0 52 SEM H3K79me2 H3K79me2/3 ChIP-seq H3K79me2 2.0 % Input 545 SEM H3K79me3 28 RS4;11 H3K79me2 1.0 96 Primograft H3K79me3 0.0 CON BCL11A P1 P2 P3 T1 T2 T3 PROM1 TAPT1 d 1.5 50 Control H3K79me3 1.0 Nascent RNA-seq % Input -370 0.5 50 MLL-AF4 knockdown 0.0 CON BCL11A P1 P2 P3 T1 T2 T3 -370 PROM1 TAPT1 f g DMSO 6 DOT1Li 545 DMSO H3K79me3 cell count x103 545 DOT1Li 4 Nascent RNA-seq 10 2 -125 DMSO 10 0 DOT1Li -10-3 0 103 104 105 -125 CD133 Fig. 2 MLL-AF4 regulates PROM1 via H3K79me2/3. a MLL-N and PROM1 and TAPT1. e H3K79me2 and H3K79me3 ChIP-qPCR in AF4-C ChIP-seq in SEM, RS4;11, MLL-AF4 ALL primograft, and control (dark shade) and MLL-AF4 siRNA knockdown (light shade) FLAG-tagged MLL-Af4 cell line at PROM1 and TAPT1. Location of SEM cells at a negative control locus (CON), BCL11A and several primers used for ChIP-qPCR is indicated below the traces. b regions of PROM1 and TAPT1 in SEM cells (primer locations indi- H3K79me2/3 ChIP-seq at PROM1 and TAPT1 in SEM, RS4;11, and cated in a). Error bars represent s.e.m. from five biological replicates. f primograft cells. c MLL-N and AF4-C ChIP-qPCR in control (red) and Reference-normalized H3K79me3 [18, 58] ChIP-seq and nascent MLL-AF4 siRNA knockdown (pink) SEM cells at a negative control RNA-seq in control (DMSO; dark purple/black) and EPZ-5676-treated locus (CON), BCL11A and several regions of PROM1 and TAPT1 in (DOT1Li; light purple/orange) SEM cells at PROM1 and TAPT1. g SEM cells (primer locations indicated in a). Error bars represent s.e.m. Flow cytometry analysis of CD133 level in SEM cells following from five biological replicates. d Nascent RNA-seq from control control (DMSO; black) or EPZ-5676 treatment (DOT1Li; orange). (black) and MLL-AF4 siRNA knockdown (orange) SEM cells at Representative of two biological replicates. 98 L. Godfrey et al. MLL-AF4 controls PROM1 expression via This strongly argues that these genes exist within the same H3K79me2/3 regulatory domain in SEM and RS4;11 cells, with a potential clover leaf structure whereby both promoters and We and others have shown that a common mechanism of enhancers may be in close proximity at the same time MLL-AF4 gene activation is via aberrant recruitment of (Fig. 3c), consistent with their correlated expression in DOT1L leading to high levels of H3K79me2/3 [16, 17]. To leukemia. test whether MLL-AF4 directs DOT1L activity at PROM1 As H3K27ac is commonly used as a marker of active we performed DOT1L and H3K79me2/3 ChIP-qPCR in enhancers, a reduction in H3K27ac may be indicative of SEM cells following MLL-AF4 knockdown. We observed a reduced enhancer activity. We have previously demon- reduction in both DOT1L and H3K79me2/3 at PROM1 and strated that DOT1Li in SEM cells results in a loss of TAPT1, as well as BCL11A (Fig. 2e and Supplementary H3K27ac at KEEs [18]. Indeed, there was a striking Fig. 3d). This indicates that MLL-AF4 may control PROM1 reduction in H3K27ac at the KEEs of PROM1 and TAPT1 and TAPT1 transcription via H3K79me2/3. following loss of H3K79me2/3 (Fig. 3d). We observed a To test this we used the DOT1L-specific small molecule similar effect by H3K27ac ChIP-qPCR in both SEM and inhibitor EPZ-5676 (DOT1Li) to deplete levels of H3K79 RS4;11 cells (Supplementary Fig. 3e, f). Notably, MLL- methylation, with a near-complete loss of H3K79me3 at AF4 siRNA knockdown produced a similar reduction in PROM1 and TAPT1 (Fig. 2f). Expression of both PROM1 H3K27ac (Supplementary Fig. 3g), further suggesting that and TAPT1 was sensitive to DOT1Li, with nascent RNA- upregulation of TAPT1 and PROM1 by MLL-AF4 is seq [17, 18] and qRT-PCR demonstrating a significant achieved via KEE enhancer function. DOT1Li also led to a decrease in transcription, comparable to the downregulation reduction in chromatin accessibility at the KEEs of both observed following MLL-AF4 knockdown (Fig. 2f and PROM1 and TAPT1 in SEM and RS4;11 cells [18] (Fig. 3d, Supplementary Fig. 3b, c). Levels of cell surface CD133 e) indicative of a reduction in transcription factor binding. were similarly reduced following DOT1Li (Fig. 2g). Taken We have previously shown that interactions between together these data suggest that MLL-AF4 controls PROM1 KEEs and gene promoters are disrupted by DOT1Li, con- expression via deposition of H3K79me2/3. sistent with a loss of enhancer function [18]. Given that the PROM1 and TAPT1 KEEs contact both gene promoters PROM1 is regulated by H3K79me2/3 enhancers (Fig. 3a), we asked whether these interactions were dependent on H3K79me2/3. To test this, we performed To further understand the regulation of PROM1 by MLL- Capture-C following DOT1Li in SEM and RS4;11 cells. AF4 we explored the chromatin landscape at PROM1 by Strikingly, we observed clear reductions in the interactions ChIP-seq and ATAC-seq in SEM and RS4;11 cells and between the PROM1 promoter and the KEEs within both primary blast cells from an MLL-AF4 ALL patient, which PROM1 and TAPT1 in SEM (Fig. 4a) and RS4;11 cells we confirmed were CD133-positive (Supplementary (Fig. 4b). A statistical analysis of enhancer–promoter con- Fig. 1e, upper panels). Strikingly, in both cell lines (Fig. 3a) tact frequency confirmed significant decreases in interaction and the primary blasts (Fig. 3b) we observed regions between the KEEs in PROM1 and TAPT1 and both pro- marked with H3K27ac, H3K4me1, and H3K79me2/3 and moters (Fig. 4c, d), arguing that H3K79me2/3 is necessary peaks of accessibility within the gene body of PROM1 and for enhancer function. Given the decreased expression of TAPT1, indicative of intragenic enhancers (highlighted PROM1 following DOT1Li or MLL-AF4 knockdown, this regions). We have recently demonstrated that KEEs can be suggests that the loss of enhancer–promoter interactions in functionally important [18] and so we hypothesized that the absence of H3K79me2/3 may in part be responsible for these KEEs may be a common mechanism to upregulate the reduction in PROM1 transcription and subsequent loss PROM1 and TAPT1 in MLL-AF4 ALL. of cell surface CD133. One characteristic of an active enhancer is that it phy- sically contacts the promoter of the gene it regulates [42– PROM1 is polycomb-repressed in CD133− acute 45]. To test whether this was true for these putative leukemia enhancers, we performed next generation Capture-C to identify regions interacting with the promoters of PROM1 As observed, not all MLLr leukemias are CD133+ (Fig. 1a, and TAPT1. In both SEM and RS4;11 cells, the PROM1 and b), and the mechanism of differential PROM1 regulation in TAPT1 KEEs were found in close proximity with their gene leukemia is unknown. To investigate this, we analyzed promoter, demonstrated by a peak in interaction frequency THP1 cells, an MLL-AF9 AML cell line that does not over the enhancers (Fig. 3a). Interestingly, these interactions express PROM1/CD133 (Fig. 1a and Supplementary were reciprocal, with the PROM1 intragenic enhancer also Fig. 1a). In contrast to SEM cells, we observed no peak of contacting the TAPT1 promoter, and vice versa (Fig. 3a). MLL-N (which detects both MLL-AF9 and wild-type H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 99 a FGFBP1 100 kb PROM 1 TAPT1 c promoter enhancer FGFBP2 BC010030 284 gene body Capture-C from PROM1 546 Capture-C from TAPT1 110 PROM1 TAPT1 MLL-N 44 H3K79me3 d SEM 100 kb FGFBP1 FGFBP2 PROM 1 BC010030 TAPT1 48 H3K27ac ATAC peaks: ** * * ** 33 6 H3K4me1 DMSO ATAC 33 44 DOT1Li H3K4me3 SEM 75 H3K27ac 33 ATAC DMSO 75 DOT1Li 610 Capture-C from PROM1 40 736 Capture-C from TAPT1 ATAC-seq ** * ATAC-seq ** 30 60 DMSO 40 DOT1Li 78 20 MLL-N 20 10 RS4;11 reads per 10 million 28 H3K79me2 0 0 −125000−100000 −75000 −50000 −25000 0 −80000 −60000 −40000 −20000 0 28 H3K27ac 125 150 100 H3K27ac H3K27ac 18 DMSO H3K4me1 75 100 DOT1Li 50 26 50 ATAC 25 0 0 −125000−100000 −75000 −50000 −25000 0 −80000 −60000 −40000 −20000 0 b position relative to PROM1 TSS position relative to TAPT1 TSS 36 H3K79me3 e FGFBP1 FGFBP2 100 kb PROM 1 BC010030 TAPT1 78 H3K27ac MLL-AF4 ALL 46 DMSO RS4;11 ATAC 11 H3K4me1 46 DOT1Li 125 H3K4me3 11 reads per 10 million ATAC 80 ATAC-seq ATAC-seq 40 60 DMSO DOT1Li 40 20 20 0 0 −125000 −100000 −75000 −50000 −25000 0 −80000 −60000 −40000 −20000 0 position relative to PROM1 TSS position relative to TAPT1 TSS Fig. 3 PROM1 is regulated by H3K79me2/3-marked enhancers. a Model for the structure of the PROM1/TAPT1 locus, based on Capture-C tracks from the viewpoint (gray bars) of the PROM1 and Capture-C data. d ATAC-seq and H3K27ac ChIP-seq at PROM1 and TAPT1 promoter in SEM and RS4;11 cells. ChIP-seq tracks for MLL- TAPT1 in control (DMSO; orange/green) and EPZ-5676-treated N, H3K79me3, H3K27ac, H3K4me1, and H3K4me3, and ATAC-seq (DOT1Li; gray) SEM cells. Statistically significant decreases in at PROM1 and TAPT1 in SEM cells, and for MLL-N, H3K79me2, ATAC-seq peaks following DOT1Li are indicated by blue asterisks H3K27ac, and H3K4me1, and ATAC-seq at PROM1 and TAPT1 in (FDR < 0.05 from three biological replicates). Histograms show RS4;11 cells. Putative enhancer regions are highlighted in blue. b overlay of ATAC-seq and H3K27ac ChIP-seq read counts across the H3K79me3, H3K27ac, H3K4me1 and H3K4me3 ChIP-seq, and highlighted regions. e ATAC-seq in control (DMSO; orange) and ATAC-seq from bone marrow of a child with MLL-AF4 ALL. CD133 EPZ-5676-treated (DOT1Li; gray) RS4;11 cells. Histograms represent expression profile of these cells is shown in Supplementary Fig. 1e overlay of ATAC-seq read counts across the highlighted regions. (upper panels). Putative enhancer regions are highlighted in blue. c MLL) at PROM1 in THP1 cells or histone modifications the PROM1 promoter in place of a strong H3K4me3 peak, associated with enhancers (H3K4me1, H3K27ac, and whereas no enrichment for H3K27me3 was observed in H3K79me2/3) or promoters (H3K4me3 and H3K27ac) SEM cells (Fig. 5a, bottom panels). Indeed, we were able to (Fig. 5a). We have previously shown that MLL-AF9 acti- partially relieve the repression of PROM1 by treatment of vates target gene expression in THP1 cells by a similar THP1 cells with the EZH1/2 and EZH2-specific inhibitors mechanism to MLL-AF4 in SEM cells [17], so we asked UNC1999 and GSK126, respectively (Supplementary why it is unable to bind or activate PROM1. Deposition of Fig. 4a). These results suggest that PROM1 repression in H3K27me3 by EZH2, the enzymatic component of PRC2, THP1 cells is associated with PRC2-mediated H3K27me3, is one mechanism by which genes can be silenced, and is explaining the absence of CD133 on the cell surface. often targeted to developmentally regulated genes [46]. Interestingly, TAPT1, unlike PROM1, is expressed in Strikingly, a peak of H3K27me3 and EZH2 was visible at THP1 cells (Supplementary Fig. 1a) and the promoter is 100 L. Godfrey et al. a FGFBP1 FGFBP2 100 kb PROM 1 BC010030 TAPT1 b FGFBP1 FGFBP2 100 kb PROM 1 BC010030 TAPT1 284 284 Capture-C from PROM1 Capture-C from PROM1 284 284 + DOT1Li + DOT1Li 142 differential 142 differential -142 -142 546 Capture-C from TAPT1 546 Capture-C from TAPT1 RS4;11 546 546 + DOT1Li + DOT1Li SEM 273 273 differential differential -273 -273 44 28 H3K79me2 H3K79me3 48 28 H3K27ac H3K27ac 6 H3K4me1 18 H3K4me1 33 ATAC 26 ATAC c 200 PROM1 200 TAPT1 d 200 PROM1 200 TAPT1 EPZ (mean reads per fragment) EPZ (mean reads per fragment) enhancer 150 150 KEE 150 150 nonKEE 100 100 p value 100 100 ns 0.05 50 50 0.01 50 50 0.001 0.0001 0 0 <0.0001 0 0 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200 DMSO (mean reads per fragment) DMSO (mean reads per fragment) DMSO (mean reads per fragment) DMSO (mean reads per fragment) Fig. 4 H3K79me2/3 regulates enhancer–promoter interactions at RS4;11 cells. c, d Statistical analysis of Capture-C-measured changes PROM1 and TAPT1. a, b Capture-C from the PROM1 and TAPT1 in interactions between PROM1 and TAPT1 promoters and enhancers promoter (gray bars) in control (DMSO; black) and EPZ-5676-treated in SEM (c) and RS4;11 (d) cells, from three biological replicates. Each (DOT1Li; orange) SEM and RS4;11 cells. Differential tracks show the circle represents a KEE (H3K79me2/3-marked enhancer element; change in Capture-C signal (black: increases; red: decreases). Repre- purple) or non-KEE (enhancer not marked with H3K79me2/3, gray). sentative of three biological replicates. ChIP-seq tracks for Size of circle is inversely proportional to the significance of the change H3K79me2/3, H3K27ac, and H3K4me1, and ATAC-seq in SEM and in interaction. bound by MLL-AF9 and marked with active chromatin in CD133− AML and ALL cells. We also observed no features (Fig. 5a). In contrast to PROM1-expressing SEM binding of MLL-N at the PROM1 promoter or gene and RS4;11 cells (Fig. 3a), however, spreading of the MLL- body in MV4;11, SHI-1, and CCRF-CEM cells (Supple- FP and H3K79me2/3 into the gene body appears to be mentary Fig. 4d), but detected EZH2 at the promoter in the diminished, associated with reduced H3K27ac (Fig. 5a). latter two cell lines (Supplementary Fig. 4c), consistent This lack of a strong KEE region may in part explain the with a lack of PROM1 expression in these cells (Supple- reduced expression of TAPT1 in THP1 cells (Supplemen- mentary Fig. 1a) due to PRC2 repression. From these tary Fig. 1a). data we conclude that the presence of an MLL-FP or EZH2/ To assess whether PRC2 silencing is a common H3K27me3 at the PROM1 promoter may determine mechanism for PROM1 regulation in CD133-negative leu- whether it is expressed in leukemia cells. The reason why kemia and not restricted to MLL-AF9 AML, we analyzed PROM1 is PRC2 repressed in some cell types and expressed RCH-ACV (non-MLLr Pre-B-ALL) and ML-2 (MLL-AF6 in others, and whether the same mechanism drives AML) cell lines, which do not express PROM1 or present repression in CD133− primary ALL patient samples is CD133 on the cell surface (Fig. 1a and Supplementary unknown, but may reflect the developmental stage of the Fig. 1a). We observed a peak of H3K27me3 at the promoter cell of origin in different leukemias. However, PROM1 of PROM1 in both RCH-ACV and ML-2 cells (Fig. 5b and repression does not reflect the relative expression of MLL Supplementary Fig. 4b), colocalizing with EZH2 in RCH- (KMT2A) and EZH2, as we observed no correlation between ACV cells (Fig. 5b and Supplementary Fig. 4c), indicating either transcript and PROM1 in ALL blasts (Supplementary that this may be a common mechanism to repress PROM1 Fig. 4e). H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 101 a FGFBP1 FGFBP2 100 kb PROM 1 BC010030 TAPT1 c FGFBP1 FGFBP2 100 kb PROM 1 BC010030 TAPT1 19 MLL-N 324 SEM 32 H3K79me2 562 THP1 Capture-C 31 H3K79me3 600 SEM 31 ATAC THP1 75 H3K27ac 474 THP1 5 H3K4me1 80 H3K4me3 d 562 Capture-C from PROM1 6 H3K27me3 31 EZH2 562 + DOT1Li 281 6 H3K27me3 differential SEM THP1 -281 474 Capture-C from TAPT1 b 48 MLL-N RCH-ACV 474 + DOT1Li 90 H3K27me3 237 69 EZH2 differential -237 e spatial interaction H3K4me3 CD133+ CD133- inhibition me me H3K27me3 me H3K79me2/3 ac H3K27ac MLL-FP MLL-FP MLL-FP MLL-FP MLL-FP MLL-FP MLL-FP MLL-FP MLL-FP PRC2 me me me ac me me ac ac me me me me me me me me me ac ac ac ac ac ac ac ac ac ac PROM1 TAPT1 ac me me INACTIVE PARTIALLY ACTIVE me PROM1 TAPT1 ACTIVE ACTIVE Fig. 5 PROM1 is polycomb-repressed in CD133− leukemia cells. a PROM1 and TAPT1 expression. Left: in CD133-positive MLLr cells, MLL-N, H3K79me2, H3K79me3, H3K27ac, H3K4me1, H3K4me3, the MLL fusion protein (MLL-FP) binds at the promoters of PROM1 H3K27me3 and EZH2 ChIP-seq, and ATAC-seq at PROM1 and and TAPT1 and spreads into the gene body. Recruitment of DOT1L TAPT1 in THP1 cells. H3K27me3 ChIP-seq at PROM1 and TAPT1 in results in elevated H3K79me2/3 levels, facilitating enhancer activity. SEM cells. b MLL-N, H3K27me3, and EZH2 ChIP-seq at PROM1 These KEEs come into proximity with the promoters of both genes, and TAPT1 in RCH-ACV cells. c Comparison of Capture-C tracks upregulating expression of PROM1 and TAPT1. Right: in CD133- from the promoter of PROM1 and TAPT1 in SEM (black) and THP1 negative MLLr cells, PRC2 binding at the promoter of PROM1 gen- (red) cells. Gray bars show the Capture-C viewpoint. Tracks are the erates a localized peak of H3K27me3 and disrupts MLL-FP binding at mean of three biological replicates. d Capture-C from the PROM1 and the promoter and gene body, repressing PROM1 expression. MLL-FP TAPT1 promoter in control (DMSO; black) and EPZ-5676-treated is still able to bind at the promoter of TAPT1, but for unknown reasons (DOT1Li; orange) THP1 cells. Differential tracks show the change in does not spread into the gene body. The lack of MLL-FP spreading Capture-C signal (black: increases; red: decreases). Tracks represent within PROM1 and TAPT1 prevents the formation of KEEs, so the the mean of three biological replicates. e Model for coregulation of expression of neither gene is upregulated. PROM1/TAPT1 H3K79me2/3 enhancers are inactive interact with the PROM1 and TAPT1 promoters. Indeed, in CD133− leukemia using Capture-C we observed no enrichment for interactions between the PROM1 promoter and the intragenic PROM1 As PROM1 is repressed in THP1 cells but TAPT1 is still or TAPT1 KEEs, in contrast to the high frequency of con- active we wanted to understand how the topological struc- tacts in SEM cells (Fig. 5c). The low frequency interactions ture of the locus compared with SEM and RS4;11 cells, observed spreading from the PROM1 promoter in where both genes are active and promoter–enhancer inter- THP1 cells are likely due to nonspecific contacts driven by actions are prevalent (Fig. 3a, c). The absence of active sequence proximity, something that is often observed in KEE marks such as H3K79me2/3, H3K4me1, and Capture-C experiments at inactive loci [21]. A similar effect H3K27ac, and accessible chromatin at PROM1 in was observed for interactions with the TAPT1 promoter, THP1 cells (Fig. 5a) is consistent with the absence of despite the fact that this gene is active (Fig. 5c). The lack of enhancer activity, so this region would not be expected to enhancer interactions is consistent with the reduced 102 L. Godfrey et al. expression of TAPT1 in THP1 relative to SEM and Finally, as in the CD133-negative leukemia cells (Fig. 5a, RS4;11 cells (Supplementary Fig. 1a). Whilst DOT1Li b), the PROM1 promoter was marked by H3K27me3 in treatment resulted in some decreases in promoter contacts CD34− FBM and FL, consistent with its repression by with local DNA sequences (Fig. 5d), there were no sig- PRC2 (Supplementary Fig. 5g). However, in adult BM nificant effects on interactions with enhancer regions cells, as for leukemic blasts (Supplementary Fig. 4d), there (Supplementary Fig. 4f), further demonstrating the lack of was no obvious correlation between the expression of MLL KEE behavior at this locus in THP1 cells. Taken together, (KMT2A), EZH2, and PROM1 (Supplementary Fig. 5d), these data suggest that PRC2 inactivation of PROM1 ren- suggesting that PROM1 status is not simply a product of the ders the enhancers found within PROM1 inactive, poten- relative levels of the two proteins. tially due to the inability of MLL-FP to bind, which consequently prevents H3K79me2/3 deposition and tran- scription factor binding. Inactivation of PROM1 also dis- Discussion rupts any interactions between PROM1 and the TAPT1 KEEs (and vice versa), suggesting that inactivation of We demonstrate here that MLL-AF4 directly regulates PROM1 via PRC2 repression impacts the chromatin struc- PROM1 leading to high expression of CD133 in a major ture and activity of the entire domain. subset of MLLr ALL patients, especially MLL-AF4 and MLL-ENL. MLL-AF4 drives transcriptional upregulation MLL and PRC2 regulate PROM1/CD133 expression of PROM1 via H3K79me2/3, which in turn promotes during lymphopoiesis enhancer function. Thus, this mechanism is most likely via aberrant recruitment of DOT1L, resulting in increased Finally, we sought to test whether the antagonistic rela- H3K79me2/3 levels and increased enhancer–promoter tionship between MLL and PRC2 is associated with interactions (Fig. 5e), although it is probable that elevated PROM1/CD133 regulation in normal B cell development. H3K79me2/3 also acts in part by promoting transcription We measured CD133 levels on normal fetal BM cells by elongation. Interestingly, MLL-AF4 appears to co-regulate flow cytometry (Supplementary Fig. 5a). As noted in nor- high level expression of the neighboring gene TAPT1 along mal cord blood and postnatal BM [6], CD133 was expres- with PROM1. When both genes are expressed, the PROM1 sed by the majority of early fetal HSPCs but was and TAPT1 KEEs interact with each other and with both significantly reduced on committed B progenitors (CD34+ promoters in an H3K79me2/3-dependent manner. Although CD19+CD10− PreProB and CD34+CD19+CD10+ ProB TAPT1 is expressed in the absence of PROM1 transcription, progenitors) and almost undetectable on CD34−CD19+ B its promoter does not interact with the PROM1 KEEs, likely cells (Supplementary Fig. 5a). Consistent with this, pub- driving the lower levels of expression of the TAPT1 gene lished single-cell RT-qPCR [36] and RNA-seq [34–36] data (Fig. 5e). Consistent with this, we observe a correlation of normal fetal and adult BM cells demonstrated a restric- between expression of PROM1 and high levels of expres- tion of PROM1 expression to immature CD34+ stem/pro- sion of TAPT1 in ALL patient samples (Supplementary genitor cells (Supplementary Fig. 5b, c). Indeed, human cell Fig. 1c). This correlation is not observed in normal hema- Atlas single-cell RNA-seq of adult BM cells [38] con- topoietic cells (Supplementary Fig. 5c, d), consistent with firmed restriction of PROM1 expression to CD34+ cells the idea that it is dependent on the MLL-FP- (Supplementary Fig. 5d). This loss of expression is generated KEEs. reflected by the chromatin structure at the PROM1 locus, One question which arises from our work is why CD133 where ATAC-seq [34, 36] revealed peaks of accessibility is expressed in some MLLr leukemias and not others. It is comparable to CD133+ SEM cells in stem and progenitor, possible that this simply reflects the initial CD133 status of but not mature, cells from both fetal and adult BM (Sup- the cell of origin, as many stem and early progenitor cells plementary Fig. 5e). are known to express CD133 while later progenitor and To ask whether wild-type MLL had a role in PROM1 more mature B cells do not [36, 40, 47, 48]. Another pos- expression in HSPCs, we purified CD34+ (CD133+; sibility is that the MLL-FP itself has the ability to bind and Supplementary Fig. 5a) and CD34− (CD133−) fetal BM cause aberrant upregulation of PROM1. Indeed, the obser- cells and analyzed MLL binding by ChIP-qPCR (Supple- vation that not all MLL-FPs are associated with high levels mentary Fig. 5f). Indeed, MLL binding was only detected at of CD133 suggests that the activity of the fusion partner the PROM1 promoter in the CD34+ cells and not CD34− may be important for its ability to bind at PROM1. We note (CD133−) cells, whilst TAPT1 and CDK6 were bound in that we do not observe binding of wild-type MLL at both progenitor and mature populations (Supplementary PROM1 in CD133− cells (Fig. 5a, b and Supplementary Fig. 5f). We confirmed these binding patterns in CD34+ Figs. 4d, 5f). Alternatively, differences in the chromatin and CD34− fetal liver (FL) cells (Supplementary Fig. 5f). state at PROM1 may render the environment permissive for H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 103 MLL-FP spreading only in certain target cell types, so (Supplementary Fig. 4a) suggests that, even in CD133− leukemias that initiate in a PROM1-repressed environment cells, there is some low-level ability to activate the locus may not be capable of reactivating the locus. For instance, that is normally suppressed. in CD133− leukemia cell lines such as THP1 and RCH- The mechanism of PROM1 regulation is of relevance ACV, the promoter of PROM1 is bound by EZH2 (the when considering the fetal cell of origin of infant MLL-AF4 enzymatic component of PRC2) and marked with ALL, which invariably originates in utero. CD133 is H3K27me3, suggesting that the gene is polycomb- expressed during normal fetal hematopoiesis primarily in repressed. hematopoietic stem cells (HSCs) and early progenitors What determines whether MLL-FP or PRC2 becomes (multipotent progenitors (MPPs) and lymphoid-primed stably bound at PROM1? Polycomb and Trithorax group MPPs) [36, 40]. This is consistent with a permissive proteins (such as MLL) are known to be antagonistic [49], environment for MLL-AF4 binding at PROM1, if the leu- so stable PRC2 binding at the promoter may prevent MLL- kemia was initiated in any of these cell types. However, FP association, and vice versa (both can recognize unme- infant MLL-AF4 ALL shares some characteristics with a thylated CpG islands [49]). In this model, in CD133− more developed B cell progenitor [20] such as fetal BM leukemias PRC2 wins this competition, rendering PROM1 PreProB progenitors [36], suggesting this could be the cell repressed. In contrast, in MLLr CD133+ leukemias MLL- of origin [33, 54]. In these cells PROM1 is much more FP binding may be preferentially stabilized over PRC2 weakly expressed and may be in the process of PRC2 and therefore PROM1 is activated. In past work, we have repression, but MLL-AF4 induction could antagonize this noted that the spreading of MLL-AF4 binding into the process by competing for binding at the promoter. Notably, gene body, as observed at PROM1, is associated with par- the chromatin environment at PROM1 remains accessible in ticularly high levels of MLL-AF4-mediated transcription PreProB progenitor cells, despite the low expression level, [17], but it was not known whether spreading initiated at the and therefore may be permissive for MLL-AF4 binding. promoter. The fact that when PROM1 is repressed the Thus, the balance between Trithorax and polycomb group MLL-FP is also absent from the body of the gene (even proteins at PROM1 upon leukemia initiation may render though PRC2 is not bound here) suggests that spreading CD133 either highly expressed or absent. may nucleate from the promoter. It should be noted that One important question is whether CD133 could be a non-MLLr leukemias can also be CD133+ (see Fig. 1b and useful therapeutic target. Since CD133 is expressed on both [40, 50, 51]), as are many stem cells [36, 40], in the absence fetal and adult HSCs, it may be difficult to target CD133+ of an MLL-FP to drive expression. We identified MLL blasts without impacting normal HSCs [40]. One approach binding at the PROM1 promoter in normal fetal CD34+ that has been proposed is to combine CD133 targeting with stem and progenitor cells (Supplementary Fig. 5f), arguing a second marker found on ALL blasts such as CD19 [4]. An that wild-type MLL plays a role in PROM1 expression, and “AND gate” strategy that requires dual target engagement there are likely other oncogenic mechanisms for PROM1 for activity would spare normal HSCs. However, such a upregulation. strategy is contingent on the inability of the leukemia cells Our work here suggests that early stem/progenitor cell to downregulate expression of either marker. Of note, we expression of PROM1/CD133 eventually becomes repres- show here that continuous expression of CD133 is essential sed by PRC2 in more mature cells. This raises an important for the growth of SEM and RS4;11 cells, as it functions question—do MLL-FPs such as MLL-AF4 overcome this at least in part by promoting cell cycle progression and repressive environment and aberrantly activate PROM1/ suppressing apoptosis. This suggests that CD133+ leuke- CD133 or does the transformation event occur in a pro- mia cells may be addicted to CD133 expression genitor cell that still displays PROM1/CD133 expression? (whilst CD133− blasts tolerate its absence). However, some Our data cannot fully answer this question, but an emerging long-term repopulating ALL cells have been shown to be model in the field suggests that PRCs are responsive to, CD133+/CD19−, complicating the use of an “AND gate” rather than instructive for, gene repression [49, 52, 53]. strategy in this way [51]. In the large subset of MLLr ALL What this might indicate is that PROM1/CD133 is normally patients where PROM1/CD133 is co-expressed with CD19, repressed by various developmental signals (e.g., binding of CD133/CD19 dual AND gate targeting could be a pro- repressive transcription factors) that eventually lead to PRC- mising targeted therapeutic approach in difficult-to-treat mediated repression. If MLL-FPs are able to bind to infants, children, and adults. To be truly effective in a wide PROM1 in a cell type that has not completely repressed the range of patients, combination therapy will be required, locus, this could lead to a disruption of the normal repres- with other approaches such as BCL2 inhibition [55, 56] or sive machinery and reactivation of the gene. The fact that targeting other MLLr-specific cell surface receptors, such as inhibiting PRC2 activity partially reactivates the locus NG2 [57]. 104 L. Godfrey et al. Data availability Compliance with ethical standards All high-throughput data have been deposited in the Gene Conflict of interest PV and TAM are founding shareholders of OxStem Oncology (OSO), a subsidiary company of OxStem Ltd. The Expression Omnibus (GEO) under accession number other authors declare that they have no conflict of interest. GSE135026. Accession numbers for datasets used from previous publications can be found in Supplementary Publisher’s note Springer Nature remains neutral with regard to Table 3. jurisdictional claims in published maps and institutional affiliations. Code availability Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as The ChIP-seq and ATAC-seq sequence mapping pipeline long as you give appropriate credit to the original author(s) and the can be accessed at https://github.com/Hughes-Genome- source, provide a link to the Creative Commons license, and indicate if Group/NGseqBasic/releases. Capture-C analysis code is changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless available from https://github.com/Hughes-Genome-Group/ indicated otherwise in a credit line to the material. If material is not CCseqBasicF/releases. included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted Acknowledgements TAM, LG, NTC, I-JL, RT, JRH, and SR were use, you will need to obtain permission directly from the copyright funded by Medical Research Council (MRC, UK) Molecular Hae- holder. To view a copy of this license, visit http://creativecommons. matology Unit grant MC_UU_12009/6, MC_UU_00016/6, and MR/ org/licenses/by/4.0/. M003221/1. AR was supported by a Bloodwise Clinician Scientist Fellowship (grants: 14041 and 17001), Wellcome Trust Clinical Research Career Development Fellowship (216632/Z/19/Z), Lady Tata References Memorial International Fellowship, and EHA-ASH Translational Research Training in Hematology Fellowship. SOB was funded by the 1. Liou GY. CD133 as a regulator of cancer metastasis through the Department of Paediatrics and Alexander Thatte Fund, University of cancer stem cells. Int J Biochem Cell Biol. 2019;106:1–7. Oxford. DJHFK is funded by a CIHR Postdoctoral Fellowship. IR is 2. Zhu L, Finkelstein D, Gao C, Shi L, Wang Y, Lopez-Terrada D, supported by the NIHR Oxford BRC, by a Bloodwise Program Grant et al. Multi-organ mapping of cancer risk. Cell. (13001) and by the MRC Molecular Haematology Unit 2016;166:1132–46.e7. (MC_UU_12009/14). PV via the Molecular Haematology Unit 3. Chen H, Lin J, Shan Y, Zhengmao L. The promotion of nano- (MC_UU_12009) and the National Institute for Health Research particle delivery to two populations of gastric cancer stem cells by (NIHR) Oxford Biomedical Research Centre (BRC) Programme. We CD133 and CD44 antibodies. Biomed Pharmacother. would like to acknowledge the WIMM Flow Cytometry Facility which 2019;115:108857. is supported by the MRC HIU, MRC MHU (MC_UU_12009), NIHR 4. Li D, Hu Y, Jin Z, Zhai Y, Tan Y, Sun Y, et al. TanCAR T cells Oxford BRC, Kay Kendall Leukemia Fund (KKL1057), John Fell targeting CD19 and CD133 efficiently eliminate MLL leukemic Fund (131/030 and 101/517), the EPA fund (CF182 and CF170), and cells. Leukemia. 2018;32:2012–6. by the WIMM Strategic Alliance awards G0902418 and 5. Schultz KR, Pullen DJ, Sather HN, Shuster JJ, Devidas M, Bor- MC_UU_12025. We thank the High-Throughput Genomics Group at owitz MJ, et al. Risk- and response-based classification of child- the Wellcome Trust Centre for Human Genetics (funded by Wellcome hood B-precursor acute lymphoblastic leukemia: a combined Trust Grant Reference 090532/Z/09/Z); the MRC WIMM Centre for analysis of prognostic markers from the Pediatric Oncology Group Computational Biology (CCB), Radcliffe Department of Medicine, (POG) and Children’s Cancer Group (CCG). Blood. University of Oxford; and Jelena Telenius for the use of her pipelines. 2007;109:926–35. B-ALL work in PM’s lab is financially supported by the European 6. Pui CH, Carroll WL, Meshinchi S, Arceci RJ. Biology, risk Research Council (CoG-2014-646903), the Spanish Ministry of stratification, and therapy of pediatric acute leukemias: an update. Economy and Competitiveness (SAF2016-80481-R), the Fundación J Clin Oncol. 2011;29:551–65. Uno entre Cienmil, the Fundación Leo Messi. PM also acknowledges 7. Sanjuan-Pla A, Bueno C, Prieto C, Acha P, Stam RW, Marschalek structural support from the Obra Social La Caixa-Fundaciò Josep R, et al. Revisiting the biology of infant t(4;11)/MLL-AF4+ Carreras. The human fetal material was provided by the Joint MRC/ B-cell acute lymphoblastic leukemia. Blood. 2015;126:2676–85. Wellcome Trust Grant 099175/Z/ 12/Z Human Developmental Biol- 8. Milne TA. Mouse models of MLL leukemia: recapitulating the ogy Resource (http://hdbr.org). We gratefully acknowledge the kind human disease. Blood. 2017;129:2217–23. generosity of patients, their parents, and staff at Great Ormond Street 9. Pieters R, De Lorenzo P, Ancliffe P, Aversa LA, Brethon B, Hospital, London. Biondi A, et al. Outcome of infants younger than 1 year with acute lymphoblastic leukemia treated with the interfant-06 protocol: results from an international phase III randomized study. J Clin Author contributions LG, NTC, SOB, AR, and TAM conceived the Oncol. 2019;25:2246–56. experimental design; LG, NTC, SOB, I-JL, SR, NE, CC, SI, GB, and 10. Meyer C, Burmeister T, Groger D, Tsaur G, Fechina L, Renneville TAM carried out experiments; LG, NTC, SOB, SR, TJ, JRH, RT, HG, A, et al. The MLL recombinome of acute leukemias in 2017. and DJHFK analyzed and curated the data; LG, NTC, SOB, SR, PM, Leukemia. 2018;32:273–84. IR, AR, TAM interpreted the data; LG, NTC, IR, AR, and TAM wrote 11. Bursen A, Schwabe K, Ruster B, Henschler R, Ruthardt M, the original manuscript; LG, NTC, SOB, IL, PV, PM, HG, IR, AR, Dingermann T, et al. The AF4.MLL fusion protein is capable of and TAM contributed to reviewing and editing the manuscript. OH, inducing ALL in mice without requirement of MLL.AF4. Blood. PA, PV, IR, AR, and TAM provided supervision and funding. 2010;115:3570–9. H3K79me2/3 controls enhancer–promoter interactions and activation of the pan-cancer stem cell. . . 105 12. Benedikt A, Baltruschat S, Scholz B, Bursen A, Arrey TN, Meyer 30. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a bioconductor B, et al. The leukemogenic AF4-MLL fusion protein causes P- package for differential expression analysis of digital gene TEFb kinase activation and altered epigenetic signatures. Leuke- expression data. Bioinformatics. 2010;26:139–40. mia. 2011;25:135–44. 31. Harvey RC, Mullighan CG, Wang X, Dobbin KK, Davidson GS, 13. Ballabio E, Milne TA. Molecular and Epigenetic Mechanisms of Bedrick EJ, et al. Identification of novel cluster groups in pediatric MLL in Human Leukemogenesis. Cancers. 2012;4:904–44. high-risk B-precursor acute lymphoblastic leukemia with gene 14. Milne TA, Martin ME, Brock HW, Slany RK, Hess JL. Leuke- expression profiling: correlation with genome-wide DNA copy mogenic MLL fusion proteins bind across a broad region of the number alterations, clinical characteristics, and outcome. Blood. Hox a9 locus, promoting transcription and multiple histone 2010;116:4874–84. modifications. Cancer Res. 2005;65:11367–74. 32. Andersson AK, Ma J, Wang J, Chen X, Gedman AL, Dang J, et al. 15. Mueller D, Bach C, Zeisig D, Garcia-Cuellar MP, Monroe S, The landscape of somatic mutations in infant MLL-rearranged Sreekumar A, et al. A role for the MLL fusion partner ENL in acute lymphoblastic leukemias. Nat Genet. 2015;47:330–7. transcriptional elongation and chromatin modification. Blood. 33. Agraz-Doblas A, Bueno C, Bashford-Rogers R, Roy A, Schneider 2007;110:4445–54. P, Bardini M, et al. Unraveling the cellular origin and clinical 16. Bernt KM, Zhu N, Sinha AU, Vempati S, Faber J, Krivtsov AV, prognostic markers of infant B-cell acute lymphoblastic leukemia et al. MLL-rearranged leukemia is dependent on aberrant H3K79 using genome-wide analysis. Haematologica. 2019;104:1176–88. methylation by DOT1L. Cancer Cell. 2011;20:66–78. 34. Corces MR, Buenrostro JD, Wu B, Greenside PG, Chan SM, 17. Kerry J, Godfrey L, Repapi E, Tapia M, Blackledge NP, Ma H, Koenig JL, et al. Lineage-specific and single-cell chromatin et al. MLL-AF4 spreading identifies binding sites that are distinct accessibility charts human hematopoiesis and leukemia evolution. from super-enhancers and that govern sensitivity to DOT1L Nat Genet. 2016;48:1193–203. inhibition in leukemia. Cell Rep. 2017;18:482–95. 35. Quek L, Otto GW, Garnett C, Lhermitte L, Karamitros D, Stoilova 18. Godfrey L, Crump NT, Thorne R, Lau IJ, Repapi E, Dimou D, B, et al. Genetically distinct leukemic stem cells in human CD34- et al. DOT1L inhibition reveals a distinct subset of enhancers acute myeloid leukemia are arrested at a hemopoietic precursor- dependent on H3K79 methylation. Nat Commun. 2019;10:2803. like stage. J Exp Med. 2016;213:1513–35. 19. Mak AB, Nixon AM, Moffat J. The mixed lineage leukemia 36. O’Byrne S, Elliott N, Rice S, Buck G, Fordham N, Garnett C, (MLL) fusion-associated gene AF4 promotes CD133 transcrip- et al. Discovery of a CD10-negative B-progenitor in human fetal tion. Cancer Res. 2012;72:1929–34. life identifies unique ontogeny-related developmental programs. 20. Lin S, Luo RT, Ptasinska A, Kerry J, Assi SA, Wunderlich M, Blood. 2019;134:1059–71. et al. Instructive role of MLL-fusion proteins revealed by a model 37. Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck of t(4;11) pro-B acute lymphoblastic leukemia. Cancer Cell. WM 3rd, et al. Comprehensive integration of single-cell data. 2016;30:737–49. Cell. 2019;177:1888–902.e21. 21. Davies JO, Telenius JM, McGowan SJ, Roberts NA, Taylor S, 38. Hay SB, Ferchen K, Chetal K, Grimes HL, Salomonis N. The Higgs DR, et al. Multiplexed analysis of chromosome con- human cell Atlas bone marrow single-cell interactive web portal. formation at vastly improved sensitivity. Nat Methods. Exp Hematol. 2018;68:51–61. 2016;13:74–80. 39. Buenrostro JD, Giresi PG, Zaba LC, Chang HY, Greenleaf WJ. 22. Greil J, Gramatzki M, Burger R, Marschalek R, Peltner M, Transposition of native chromatin for fast and sensitive epige- Trautmann U, et al. The acute lymphoblastic leukaemia cell line nomic profiling of open chromatin, DNA-binding proteins and SEM with t(4;11) chromosomal rearrangement is biphenotypic nucleosome position. Nat Methods. 2013;10:1213–8. and responsive to interleukin-7. Br J Haematol. 1994;86:275–83. 40. Bueno C, Velasco-Hernandez T, Gutierrez-Aguera F, Zanetti SR, 23. Ohyashiki K, Ohyashiki JH, Sandberg AA. Cytogenetic char- Baroni ML, Sanchez-Martinez D, et al. CD133-directed CAR acterization of putative human myeloblastic leukemia cell lines T-cells for MLL leukemia: on-target, off-tumor myeloablative (ML-1, -2, and -3): origin of the cells. Cancer Res. toxicity. Leukemia. 2019;33:2090–125. 1986;46:3642–7. 41. Guenther MG, Lawton LN, Rozovskaia T, Frampton GM, Levine 24. Chen S, Xue Y, Zhang X, Wu Y, Pan J, Wang Y, et al. A new SS, Volkert TL, et al. Aberrant chromatin at genes encoding stem human acute monocytic leukemia cell line SHI-1 with t(6;11)(q27; cell regulators in human mixed-lineage leukemia. Genes Dev. q23), p53 gene alterations and high tumorigenicity in nude mice. 2008;22:3403–8. Haematologica. 2005;90:766–75. 42. Williamson I, Lettice LA, Hill RE, Bickmore WA. Shh and ZRS 25. Jack I, Seshadri R, Garson M, Michael P, Callen D, Zola H, et al. enhancer colocalisation is specific to the zone of polarising RCH-ACV: a lymphoblastic leukemia cell line with chromosome activity. Development. 2016;143:2994–3001. translocation 1;19 and trisomy 8. Cancer Genet Cytogenet. 43. Chen H, Levo M, Barinov L, Fujioka M, Jaynes JB, Gregor T. 1986;19:261–9. Dynamic interplay between enhancer-promoter topology and gene 26. Roy A, Cowan G, Mead AJ, Filippi S, Bohn G, Chaidos A, et al. activity. Nat Genet. 2018;50:1296–303. Perturbation of fetal liver hematopoietic stem and progenitor cell 44. Brown JM, Roberts NA, Graham B, Waithe D, Lagerholm C, development by trisomy 21. Proc Natl Acad Sci USA. Telenius JM, et al. A tissue-specific self-interacting chromatin 2012;109:17579–84. domain forms independently of enhancer-promoter interactions. 27. Wilkinson AC, Ballabio E, Geng H, North P, Tapia M, Kerry J, Nat Commun. 2018;9:3849. et al. RUNX1 is a key target in t(4;11) leukemias that contributes 45. Monahan K, Horta A, Lomvardas S. LHX2- and LDB1-mediated to gene activation through an AF4-MLL complex interaction. Cell trans interactions regulate olfactory receptor choice. Nature. Rep. 2013;3:116–27. 2019;565:448–53. 28. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, 46. Ringrose L, Paro R. Epigenetic regulation of cellular memory by et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. the Polycomb and Trithorax group proteins. Annu Rev Genet. 2013;29:15–21. 2004;38:413–43. 29. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, 47. Menendez P, del Canizo MC, Orfao A. Immunophenotypic et al. The sequence alignment/map format and SAMtools. characteristics of PB-mobilised CD34+ hematopoietic progenitor Bioinformatics. 2009;25:2078–9. cells. J Biol Regul Homeost Agents. 2001;15:53–61. 106 L. Godfrey et al. 48. Menendez P, Caballero MD, Prosper F, Del Canizo MC, Perez- 54. Malouf C, Ottersbach K. The fetal liver lymphoid-primed multi- Simon JA, Mateos MV, et al. The composition of leukapheresis potent progenitor provides the prerequisites for the initiation of t products impacts on the hematopoietic recovery after autologous (4;11) MLL-AF4 infant leukemia. Haematologica. 2018;103: transplantation independently of the mobilization regimen. e571–4. Transfusion. 2002;42:1159–72. 55. Benito JM, Godfrey L, Kojima K, Hogdal L, Wunderlich M, 49. Klose RJ, Cooper S, Farcas AM, Blackledge NP, Brockdorff N. Geng H, et al. MLL-rearranged acute lymphoblastic leukemias Chromatin sampling-an emerging perspective on targeting poly- activate BCL-2 through H3K79 methylation and are sensitive to comb repressor proteins. PLoS Genet. 2013;9:e1003717. the BCL-2-specific antagonist ABT-199. Cell Rep. 50. Wuchter C, Ratei R, Spahn G, Schoch C, Harbott J, Schnittger S, 2015;13:2715–27. et al. Impact of CD133 (AC133) and CD90 expression analysis 56. Khaw SL, Suryani S, Evans K, Richmond J, Robbins A, Kur- for acute leukemia immunophenotyping. Haematologica. masheva RT, et al. Venetoclax responses of pediatric ALL 2001;86:154–61. xenografts reveal sensitivity of MLL-rearranged leukemia. Blood. 51. Cox CV, Diamanti P, Evely RS, Kearns PR, Blair A. Expression 2016;128:1382–95. of CD133 on leukemia-initiating cells in childhood ALL. Blood. 57. Lopez-Millan B, Sanchez-Martinez D, Roca-Ho H, 2009;113:3287–96. Gutierrez-Aguera F, Molina O, Diaz de la Guardia R, et al. 52. Riising EM, Comet I, Leblanc B, Wu X, Johansen JV, Helin K. NG2 antigen is a therapeutic target for MLL-rearranged Gene silencing triggers polycomb repressive complex 2 recruit- B-cell acute lymphoblastic leukemia. Leukemia. 2019;33: ment to CpG islands genome wide. Mol Cell. 2014;55:347–60. 1557–69. 53. Sneppen K, Ringrose L. Theoretical analysis of Polycomb- 58. Orlando DA, Chen MW, Brown VE, Solanki S, Choi YJ, Olson Trithorax systems predicts that poised chromatin is bistable and ER, et al. Quantitative ChIP-seq normalization reveals global not bivalent. Nat Commun. 2019;10:2133. modulation of the epigenome. Cell Rep. 2014;9:1163–70.