The National Center of Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (ncbi.nlm.nih.gov/geo) was searched and microarray expression data (GSE68643) and RNA-Seq data (GSE73457) were downloaded for comparing MLL-AF9 retrovirus infected and wild-type (WT) mouse bone marrow cells. Unprocessed data sets for microarray (.cel files) and RNA-Seq (.sra files) were used for further analysis.

DEGs were identified using GEO2R (ncbi.nlm.nih.gov/geo/geo2r/) from a microarray dataset (GSE68643). The probe was annotated as an official gene symbol by the corresponding annotation files (GPL16570). For RNA-Seq data (GSE73457), sra files were converted into the fastq format using the SRA Toolkit version 2.9.1 (ncbi.nlm.nih.gov/Traces/sra/?view=software). Reads were aligned to the mouse Ensemble (GRCm38.p6) reference genome using HISAT2 version 2.0.4 (ccb.jhu.edu/software/hisat2/index.shtml) (9). Aligned reads were counted using HTSeq version 0.5.4p3 (htseq.readthedocs.io/en/release_0.10.0/) and summarized at the gene level guided by the gene annotation file in GTF format from Ensembl (ftp://ftp.ensembl.org/pub/release-94/gtf/mus_musculus) (10). The expression levels of genes was calculated using cufflinks version 2.2.0 (cole-trapnell-lab.github.io/cufflinks/) and normalized to reads per kilobase per million (RPKM) (11). DESeq2version 1.18.1 (bioconductor.org/packages/release/bioc/html/DESeq2.html) was applied to analyze the differential expression of genes (12). Only those genes with a log₂ fold change (FC)>1 and Benjamini and Hochberg adjusted P<0.05 were recognized as significantly differentially expressed in microarray and RNA-Seq data. Pearson's correlation coefficient of log2FC was used to measure common DEGs reliability between two expression datasets (GSE68643 and GSE73457).

The Database for Annotation, Visualization and Integrated Discovery (DAVID) database was used to detect GO categories and KEGG pathways with significant over-representation in DEGs compared with the whole genome (13). The significantly enriched biological processes and KEGG were identified as Benjamini-Hochberg adjusted P<0.05.

To identify the binding sites of MLL-AF9 and WT MLL, ChIP-Seq data obtained using antibodies against the N-terminus of MLL-1 and the C-terminus of AF9 in various GEO datasets (GSE89336, GSE79899, GSE54344 and GSE83671) were downloaded. Model-based Meta-analysis of ChIP (MM-ChIP) software (http://liulab.dfci.harvard.edu/MM-ChIP/MMChIP-1.0.tar.gz) was applied in the cross-study integrative analysis of MLL and AF9 ChIP-Seq data (14). The MLL and AF9 binding peaks were overlapped to identify considerable MLL-AF9-binding peaks using the mergePeaks function from Hypergeometric Optimization of Motif EnRichment (HOMER) version 4.8 (homer.ucsd.edu/homer/) software (15). MLL binding peaks without AF9 signal were defined as MLL WT binding sites.

The ChIP-Seq data of histone modifications downloaded from the GEO database are summarized in Table I. Using dPCA (www.biostat.jhsph.edu/dpca), the present study analyzed differential levels of H3K4me3, H3K27ac and H3K79me2 at WT MLL and MLL-AF9 binding sites in MLL-AF9 leukemia cells and normal hematopoietic cells. Differential principal components (dPCs) with high signal-to-noise ratio (SNR) were considered reliable dPCs to report. The cut-off SNR value (SNR>5) was based on a previous study (8). dPCA calculated the false discovery rate (FDR) and log₂ FC of the ChIP-Seq binding signal of each dPC. Differential sites of reliable dPCs were defined at a 5% FDR level. Genome regions with differential ChIP-Seq binding signals were annotated by HOMER software.

The gene expression data were downloaded from the UCSC genome browser database (https://xenabrowser.net/datapages/?cohort=TCGA%20TARGET%20GTEx&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443), which contained RNA-seq data from three large cohorts, such as TCGA (https://portal.gdc.cancer.gov/), TARGET (https://ocg.cancer.gov/programs/target) and GETx (https://commonfund.nih.gov/GTEx/).

Differential expression between AML samples with MLL-AF9 translocation and normal blood samples was performed using DESeq2 version 1.18.1 (bioconductor.org/packages/release/bioc/html/DESeq2.html). All of the input parameters were default values.

Bone marrow transplantation studies were performed as previously described (16). Briefly, C57BL/6 female mice (12 week old; Shanghai SLAC laboratory Animal Co., Ltd.) were used as donors and recipients. A total of 22 recipients mice were housed in specific pathogen-free (SPF) conditions under a 12 h light-dark cycle at 22±1°C, humidity of 50–60% with free access to food and water. The mice were randomly divided into control and MLL-AF9 model groups (n=11 in each group; weight, ~20 g). The donor mice were sacrificed and lineage negative hematopoietic progenitors were obtained from femurs and tibias using Lineage Cell Depletion mouse kit and LS MACS columns (Miltenyi Biotec, Inc.) and cultured in RPMI-1640 medium (Hyclone; GE Healthcare Life Sciences) supplemented with 20% fetal bovine serum (Gibco; Thermo Fisher Scientific, Inc.), 100 ng/ml stem cell factor, 10 ng/ml interleukin (IL)-3 and 10 ng/ml IL-6 (R&D Systems, Inc.). Retrovirus was produced following infection of 293T cells (The Institution of Biochemistry and Cell Biology, Chinese Academy of Sciences) with retroviral construct pMIG-FLAG-MLL-AF9 (Addgene plasmid 71443; Addgene, Inc.) along with the helper plasmid pCL-ECO (Addgene plasmid 12371; Addgene, Inc.) and the virus supernatants was harvested 48 h after transfection, filtered through a 0.45 µm syringe filter. The lineage negative hematopoietic progenitors were infected by spinfection (657 g for 2 h at 32°C; 4 µg/µl polybrene). pMIG-FLAG-MLL-AF9 was a gift from Professor Daisuke Nakada (Addgene plasmid no. 71443; http://n2t.net/addgene:71443; RRID, Addgene_71443) (17). pCL-Eco was a gift from Professor Inder Verma (Addgene plasmid no. 12371; http://n2t.net/addgene:12371; RRID, Addgene_12371) (18). Infection efficiency was shown to be 20–25% as assessed by GFP expression using flow cytometry (BD FACSVerse; BD Biosciences; Becton, Dickinson and Company). Recipient C57BL/6 mice were transferred from SPF conditions to the RS 2000 biological irradiator (Rad Source Technologies, Inc). The X-ray-irradiation lasted 5 min and 52 sec with a total dose rate of 7.5 Gy. After irradiation, the animals were placed back to their cages under SPF conditions. The infected cells were intravenously injected via the tail vein into X-ray-irradiated recipient C57BL/6 mice 48 h after infection (10⁶ transduced cells per mouse). All recipient mice were housed in a specific pathogen-free environment and fed autoclaved water (pH 2.0). Retroviral construct pMIG-FLAG-MLL-AF9 co-expressed MLL-AF9 along with a green fluorescent protein (GFP) reporter gene, thus GFP-positive cells indicated effective MLL-AF9 virus infection. Peripheral blood was collected into Eppendorf tubes from the mouse tail with EDTA anticoagulant, incubated with red blood lysis buffer (Beijing Solarbio Science & Technology Co., Ltd.) and subjected to flow cytometry analysis. The recipients were evaluated for the percentage of GFP-positive cells in peripheral blood 3 weeks after injection using flow cytometry (BD FACSVerse; BD Biosciences; Becton, Dickinson and Company). The control recipients received the same dose of radiation and were intravenously injected with normal lineage negative hematopoietic progenitors via the tail vein. The percentage of GFP-positive cells increased with time in MLL-AF9 model mice.

Total RNA samples harvested from the whole bone marrow cells of sacrificed MLL-AF9 mice and WT C57BL/6 mice were extracted using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.). cDNA was synthesized using Takara PrimeScript RT Master Mix (cat. no. RR036A; Takara Biotechnology Co., Ltd.) according to the manufacturer's protocols. The RT reaction was performed as follows: 15 min at 37°C followed by an incubation at 85°C for 5 sec. qPCR was performed with SYBR Premix Ex Taq (cat. no. DRR041A; Takara Biotechnology Co., Ltd.) using an Applied Biosystems 7500 Real-Time PCR system (Thermo Fisher Scientific, Inc.) following manufacturer's protocols. The thermocycling conditions were as follows: Initial denaturation at 95°C for 30 sec, followed by 40 cycles of 95°C for 5 sec and 60°C for 30 sec. Gene expression was normalized to the levels of mouse GAPDH mRNA. Relative mRNA expression was calculated using the 2^−ΔΔCq method (19). Primer sequences are listed in Table II.

Kaplan-Meier survival analysis and log-rank test were performed to determine the overall survival of MLL-AF9 mice and WT C57BL/6 mice. Leukemic burden in peripheral blood and splenomegaly between groups were compared using Student's t-test. Statistical analyses were performed using statistical software R (version 3.3.3; www.r-project.org/) and survival package. P<0.05 was considered to indicate a statistically significant difference.

The present study identified DEGs from two independent datasets and mapped them to GO and KEGG pathways to identify significantly enriched functional terms. Histone modification changes in MLL-AF9 binding sites were further profiled based on public ChIP-Seq datasets. The potential direct targets of MLL-AF9 were identified by comparing alterations in histone modifications with changes in gene expression. The differential expression of potential direct targets of MLL-AF9 was validated in the public AML project TARGET and by RT-qPCR in an MLL-AF9 mouse model. A flowchart of presenting the study design is shown in Fig. 1.

To identify DEGs associated with MLL-AF9, two independent datasets (GSE68643 and GSE73457) were analyzed and murine MLL-AF9 leukemia cells were compared with normal hematopoietic cells to detect common genes across these datasets. The analyses revealed that 413 genes were consistently identified as being differentially expressed (adjusted P<0.05) with ≥2-fold differential expression between the groups (Fig. 2A). The log2FC of common DEGs exhibited a high correlation (Pearson's correlation coefficient value=0.78) between these two independent experiments. Of these, 57 and 356 genes were upregulated and downregulated, respectively (Fig. 2B). There were various genes previously demonstrated to be associated with MLL-AF9, including the homeobox A (HOXA) cluster genes, Meis homeobox 1, β3 integrin and runt related transcription factor 2. These results suggested that the datasets used in the present study were suitable for MLL-AF9 gene analysis.

The common DEGs were then subjected to enrichment analysis in DAVID. The common DEGs in blood were enriched in biological processes of ‘immune system process’, ‘inflammatory response’, ‘immune response’, ‘signal transduction’, ‘innate immune response’, ‘adaptive immune response’, ‘chemotaxis’, ‘positive regulation of ERK1 and ERK2 cascade’, ‘positive regulation of T cell proliferation’, ‘positive regulation of cell migration’, ‘embryonic skeletal system morphogenesis’, ‘negative regulation of inflammatory response’, ‘positive regulation of angiogenesis’ and ‘response to lipopolysaccharide’ (Fig. 3A). KEGG enrichment analysis revealed that DEGs were enriched in pathways including the ‘nuclear factor-κB (NF-κB) signaling pathway’, ‘tumor necrosis factor (TNF) signaling pathway’, ‘cytokine-cytokine receptor interaction’, hematopoietic cell lineage’, ‘sphingolipid signaling pathway’, ‘T cell receptor signaling pathway’, ‘transcriptional misregulation in cancer’ and ‘measles’ (Fig. 3B).

Using MM-ChIP for integrative peaks analysis and HOMER software for overlapping peaks. The present study identified 10,464 MLL-AF9 occupied regions and 11,722 WT MLL occupied regions. There was no significant difference in genomic distribution between the MLL-AF9 and MLL WT peaks (Fig. 4A). However, MLL-AF9 peaks exhibited a higher signal than the MLL WT peaks (Fig. 4B).

To systematically investigate the aberrant histone modifications involved in MLL-AF9, ChIP-Seq data of histone modifications in MLL-AF9 leukemia cells and normal hematopoietic progenitor cells were downloaded from NCBI GEO datasets, including H3K4me3, H3K79me2 and H3K27ac. dPCA was applied to explore differential chromatin patterns at the MLL-AF9 peaks and MLL-WT peaks (Table I). The top dPCs in MLL-AF9 exhibited SNR >5, while the SNR in dPCs of MLL-WT peaks displayed SNR <5 (Fig. 5). Thus, only significant histone modifications differences were observed in MLL-AF9 binding sites. At the 5% FDR level, dPCA reported 4,166 differential histone modifications sites.

A total of 390 differential histone modifications sites were associated with 220 MLL-AF9 DEGs, accounting for 53.3% of all DEGs. As H3K4me3, H3K27ac and H3K79me2 are known as active transcription activities, the same directional changes in all three histone modifications were associated with an increase in gene expression. As presented in Fig. 6, a specific epigenetic signature was identified and the expression difference of 14 genes was highly associated with the log₂FC of H3K4me3, H3K27ac and H3K79me2 binding signals in promoter-transcription start site regions when comparing MLL-AF9 leukemia cells and normal hematopoietic cells. These 14 genes were identified as potential direct targets of MLL-AF9.

The present study validated the epigenetic-regulated genes by comparing the gene expression profiles of AML samples with MLL-AF9 translocation and normal blood samples. The comparisons indicated that 8 of the 14 epigenetic-regulated genes were confirmed to be differentially expressed in AML patients with MLL-AF9 translocation (Fig. 7).

The present study constructed a murine AML model by transducing lineage negative hematopoietic progenitors with a retroviral vector encoding human MLL-AF9 and transplanting the infected cells into 7.5 Gy X-ray-irradiated C57BL/6 mice via the tail vein (Fig. 8A). The recipient mice developed leukemia with an average life span of ~6 weeks (Fig. 8B). The transplanted mice exhibited similar leukemic burden by analysis of GFP+ cells in peripheral blood (Fig. 8C). Leukemic mice were sacrificed and splenomegaly was evaluated when the tumor burden in peripheral blood was ≥90% (Fig. 8D). The expression levels of potential direct target genes of MLL-AF9 were validated by RT-PCR, and the results supported the microarray and RNA-Seq DEG results (GSE68643, GSE73457; Fig. 8E).