Chromatin accessibility promotes hematopoietic and leukemia stem cell activity
Nature Communications volume 11, Article number: 1406 (2020) Cite this article
Haematopoietic stem cells
Chromatin organization is a highly orchestrated process that influences gene expression, in part by modulating access of regulatory factors to DNA and nucleosomes. Here, we report that the chromatin accessibility regulator HMGN1, a target of recurrent DNA copy gains in leukemia, controls myeloid differentiation. HMGN1 amplification is associated with increased accessibility, expression, and histone H3K27 acetylation of loci important for hematopoietic stem cells (HSCs) and leukemia, such as HoxA cluster genes. In vivo, HMGN1 overexpression is linked to decreased quiescence and increased HSC activity in bone marrow transplantation. HMGN1 overexpression also cooperates with the AML-ETO9a fusion oncoprotein to impair myeloid differentiation and enhance leukemia stem cell (LSC) activity. Inhibition of histone acetyltransferases CBP/p300 relieves the HMGN1-associated differentiation block. These data nominate factors that modulate chromatin accessibility as regulators of HSCs and LSCs, and suggest that targeting HMGN1 or its downstream effects on histone acetylation could be therapeutically active in AML.
In eukaryotic cells, DNA is packaged into chromatin, a dynamic structure containing genomic DNA, histones, and nonhistone proteins. Chromatin regulation by chemical modification of DNA or histones modulates gene expression and is important for development, differentiation, and response to signals 1 . Cancer cells frequently acquire abnormalities in the same chromatin regulatory pathways, which promote malignancy and may impart targetable therapeutic dependencies 2 . Relatively less is known about how nonhistone nucleosome-associated proteins that modify chromatin compaction and accessibility contribute to transformation.
Chromatin accessibility affects gene expression by altering interactions of transcriptional regulatory factors with their targets. The linker histone H1 organizes nucleosomes into higher-order structures, increasing chromatin folding, and is associated with decreased transcription 3 , 4 . In contrast, the HMGN family of nucleosome binding proteins competes with histone H1 and promotes decompaction of chromatin, increased accessibility, and locally enhanced transcription 5 , 6 . HMGNs interact with nucleosomes independent of DNA sequence, via a 20 amino acid nucleosome binding domain that recognizes an acidic patch formed by the core histones H2A and H2B 7 . HMGN genomic positioning is not random; it preferentially co-localizes with regulatory marks at active promoters and affects nucleosome organization, DNase hypersensitivity patterns, and post-translational histone marks 8 , 9 . Despite these links between HMGNs and chromatin regulation, a mechanistic understanding of how HMGNs contribute to disease is not as well understood.
Hematopoiesis is a highly regulated process requiring both self-renewal and differentiation of hematopoietic stem cells (HSCs) to maintain diverse mature blood components. Disruption of normal hematopoiesis can lead to pathologic states, including leukemia 10 . One hallmark of leukemia is a block in normal differentiation from hematopoietic stem/progenitor cells (HSPCs) to their mature progeny. In acute promyelocytic leukemia (APL), “differentiation therapy” using all-trans retinoic acid and arsenic trioxide releases the differentiation block and is curative in most patients, but treatments to promote differentiation have not emerged clinically in most other leukemia subtypes 11 .
Amplification of chromosome 21, particularly the distal segment 21q22, is highly associated with acute leukemia. Germline amplification, such as in Down syndrome (trisomy 21), confers a markedly increased risk of myeloid and lymphoid leukemia 12 . In addition, somatic gain of chromosome 21 is associated with aggressive leukemias, such as iAMP21 B cell acute lymphoblastic leukemia that has multiple extra copies of 21q22 and a very poor prognosis 13 . In acute myeloid leukemia (AML) with complex karyotype, a subtype associated with poor outcomes, 21q22 is the most frequent region of high-level DNA copy gain 14 . We previously performed an RNA interference screen in murine B cell progenitors modeling Down syndrome and found that Hmgn1 was the amplified gene most critical to support hematopoietic colony forming activity 15 . In B cells, HMGN1 overexpression promotes global changes in transcription with selective amplification of lineage-specific survival pathways 16 . However, how 21q22/HMGN1 amplification affects HSPCs/myeloid differentiation or confers therapeutic vulnerability is not clear.
Here, we find that HMGN1 impairs normal myeloid differentiation in association with increased gene expression and H3K27 acetylation, particularly at promoters of genes that regulate HSPC identity and function. Moreover, HMGN1 overexpression promotes a clonal advantage in HSPCs in vivo and increases leukemia stem cell (LSC) activity in concert with AML oncogenes. Suggesting potential therapeutic relevance, the differentiation impairment by HMGN1 is dependent on the histone acetyltransferases (HATs) CBP and p300 and is reversible by HAT inhibition.
HMGN1 overexpression impairs myeloid differentiation
HMGN1 is highly expressed across human and mouse immature hematopoietic stem and progenitor populations but is markedly downregulated in differentiated myeloid cells such as neutrophils and monocytes (Supplementary Fig. 1a ) 17 . This is consistent with data from other tissues where downregulation of HMGN1 is linked with differentiation to specific lineages 18 . Furthermore, when examined microscopically, hematopoietic progenitors, and AML blasts have visibly “open” chromatin, which compacts during normal myeloid development or after AML treatments that restore myeloid differentiation 19 , 20 . This led us to hypothesize that HMGN1’s role in maintaining open chromatin might contribute to the differentiation block in AML.
To interrogate the role of 21q22 amplification and HMGN1 in myeloid differentiation, we immortalized primary hematopoietic progenitors in an ex vivo culture system that facilitates analysis of immature myeloid cells and their progeny during synchronized differentiation 21 . In mice, Hmgn1 is located on chromosome 16 and is trisomic in several models of Down syndrome, including Ts1Rhr 22 , which triplicates 31 genes orthologous to a segment of human chr21q22 that is recurrently amplified in AML. We transduced bone marrow from wild-type (WT), Ts1Rhr, and HMGN1-OE mice (a transgenic model only overexpressing human HMGN1, at 2–3 times the level of the endogenous protein 15 , 16 , 23 ) with a retrovirus expressing an estrogen receptor (ER)-HoxB8 fusion protein, which maintains cells as immature progenitors in the presence of estradiol (E2). Upon removal of E2 and in the presence of interleukin 3, wild-type cells undergo synchronized differentiation to mature myeloid cells (CD11b+ GR-1+) over 6–7 days. In contrast, cells from the Ts1Rhr or HMGN1-OE models had delayed myeloid differentiation, as measured by later acquisition of CD11b and GR-1 (Fig. 1a , upper panel). Ts1Rhr and HMGN1-OE progenitors did not acquire mature myeloid cell morphology at day 4 (Fig. 1b ) nor did HMGN1-OE progenitors upregulate reactive oxygen species (ROS) production during differentiation to the same degree as wild-type cells (Fig. 1c ). This suggests that the HMGN1-associated differentiation abnormality was functionally relevant and not simply a change in cell surface marker expression. Ts1Rhr and HMGN1-OE undifferentiated progenitors also had an increased growth rate in the presence of E2 compared to wild-type cells (Fig. 1a , lower panel).
Fig. 1: HMGN1 overexpression impairs myeloid differentiation.
a Analysis of myeloid differentiation (upper) and proliferation (lower) in wild type, Ts1Rhr, and HMGN1-OE myeloid progenitors. Differentiation was measured after withdrawal of E2 as percentage of cells expressing CD11b and GR-1. b Representative morphology of progenitors at baseline, 2, and 4 days after withdrawal of E2. c Analysis of reactive oxygen species (ROS) production during differentiation. Plot is represented as relative fluoresce units (RFU) normalized to wild type at time 0. d Myeloid differentiation with or without shRNA knockdown of Hmgn1 in Ts1Rhr cells, Data compared by two-sided t test between control and Hmgn1 hairpins. Western blot for HMGN1 96 h after induction of the indicated shRNAs. e (Upper) Schematic representation of HMGN1-WT and HMGN1-SEmut (nucleosome binding domain (NBD) mutant harboring S20E and S24E mutations). NLS nuclear localization signal, RD relaxation domain. Proliferation (middle) and differentiation (lower) assays were measured as in panel a) in wild-type cells expressing the indicated proteins (EV empty vector). Western blot reflects expression of FLAG-HMGN1-WT and FLAG-HMGN1-SEmut. In all panels, graphs represent n = 3 biologically independent samples, except panel e, which is biologically independent duplicates. Conditions compared by two-sided t test. Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size image
To validate the role of HMGN1 in myeloid differentiation, we performed knock-down and overexpression experiments in these model systems. First, we found that HMGN1 was necessary for the aberrant differentiation phenotype in the 21q22 triplication model, because shRNA knockdown of Hmgn1 in Ts1Rhr cells promoted increased differentiation (Fig. 1d ). Next, we transduced WT progenitors with retroviral constructs to express either wild-type HMGN1 (HMGN1-WT) or a mutant unable to bind to nucleosomes (HMGN1-SEmut) 16 . Overexpression of HMGN1, but not the HMGN1-SE mutant, caused increased proliferation and impaired differentiation after removal of E2 (Fig. 1e ).
We next asked if modulating HMGN1 was consequential in human AMLs. Among 216 human cancer cells lines analyzed in Project Achilles v2.4.3, a genome-wide RNA interference screen 24 , there were 18 derived from patients with AML. shRNAs targeting HMGN1 were selectively depleted in AML cells compared to all other cancer cell types, suggesting a lineage-specific dependency on HMGN1 in AML (Supplementary Fig. 1b ). AML cell lines trended toward higher expression of HMGN1 compared to other lineages (Supplementary Fig. 1c ). However, it is important to note that most of the AML cell lines tested were diploid for 21q22/HMGN1 and none have known mutations in HMGN1, yet they still exhibited relative HMGN1 dependency (Supplementary Fig. 1d ), raising the possibility that HMGN1 is important for AML even in cases without chr21 amplification. To confirm these findings by an orthogonal technique, we used CRISPR/Cas9 to deplete HMGN1 from AML cell lines. Consistent with the Project Achilles data, loss of HMGN1 was associated with decreased proliferation and/or induction of differentiation (Supplementary Fig. 2 ).
HMGN1 promotes expression of stem cell and leukemia genes
Our next goal was to determine how HMGN1 overexpression affected the transcriptome and proteome of myeloid cells. We first determined the “point of no return” after removal of E2 beyond which wild-type cells were fully committed to terminal myeloid differentiation. Thirty-six hours after removal of E2 was the first timepoint where differentiation could no longer be reversed by re-addition of E2 (Supplementary Fig. 3 ). By 96 h after removal of E2 there were obvious differences in proliferation between wild-type and HMGN1-OE cells (Fig. 1a ). Therefore, we performed RNA-sequencing and proteome analysis via tandem mass tagging (TMT) at 0, 36, and 96 h after the withdrawal of E2 from wild-type and HMGN1-OE cultures. Transcriptome analysis was normalized to “spiked-in” synthetic RNAs to control for possible differences in total RNA content per cell 25 . We also performed an Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) to measure chromatin openness, and multiplexed, indexed T7 chromatin immunoprecipitation sequencing (Mint-ChIP) 26 to profile histone modifications associated with HMGN1 overexpression (Fig. 2a ).
Fig. 2: HMGN1 promotes chromatin accessibility and confers stem cell and leukemia-associated transcriptional and chromatin phenotypes.
a Schematic of the experimental approach for ATAC-seq, RNA-seq, Mint-ChIP chromatin profiling, and TMT proteomics during myeloid differentiation. b Gene set enrichment analysis (GSEA) of RNA-seq in undifferentiated wild-type and HMGN1-OE progenitors showing enrichment of the Brown_myeloid_cell_development_up gene set in wild-type cells. Heatmap shows mean expression of the leading edge genes from the gene set in GSEA in biological duplicates of wild-type and HMGN1-OE cells at the indicated time points, expressed as log2 fold change (FC) relative to WT_0h. c GSEA of wild-type and HMGN1-OE progenitors showing enrichment of genes in the Eppert_LSC_signature gene set in HMGN1-OE cells. Heatmap of leading edge genes as in panel b. d Heatmap of expression of Hox cluster genes in wild-type and HMGN1-OE cells during differentiation, expressed as in panel b. e Chromatin accessibility tracks representing distance from nearest peak center of all differential peaks in ATAC-seq (left), and metagene plots of ATAC-seq differential peaks and all shared peaks in wild-type and HMGN1-OE cells (right, n = 3 biologically independent replicates). f Gene tracks showing ATAC-seq reads at the HoxA7 and HoxA9 loci at baseline and 96 h in wild-type and HMGN1-OE progenitors (top). Heatmap of ATAC-seq signals at HoxA and HoxB genes in wild-type and HMGN1-OE progenitors, expressed as log2 FC relative to the mean WT_0h value (bottom). g Metagene profiles of H3K27 acetylation surrounding promoters in wild-type and HMGN1-OE cells at 0 and 96 h of differentiation (left). H3K27ac levels were also measured by western blotting (middle) and by intracellular flow cytometry (right, normalized to WT 0 h, n = 3 biological replicates, genotypes compared by two-sided t test). Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size image
First, we confirmed that differential RNA expression was correlated with proteomic changes at all three time points of differentiation, suggesting that the HMGN1-associated transcriptional changes were also reflected in protein abundance (Supplementary Fig. 4a, b ). Gene set enrichment analysis (GSEA) of RNA expression in immature progenitors (in E2) suggested that myeloid differentiation programs were reduced in HMGN1-OE cells compared to wild-type cells (Fig. 2b , Supplementary Data 1 ). Furthermore, HMGN1-OE progenitors were more transcriptionally similar to HSCs or AML LSCs 27 when compared to wild-type progenitors (Fig. 2c , Supplementary Data 2 ). This transcriptional difference is notable given that the cells are morphologically similar (Fig. 1b ) and that wild-type, Ts1Rhr, and HMGN1-OE undifferentiated cells have a comparable cell surface phenotype, most similar to that of a normal granulocyte–monocyte progenitor (GMP, Lineage-CD117+Sca1-CD34+CD16/32+, Supplementary Fig. 5a ). Among the most upregulated genes in HMGN1-OE cells were the HoxA and HoxB families. This difference was present in the undifferentiated state and persisted during differentiation (Fig. 2c, d , Supplementary Data 3 ). Hox proteins are homeobox transcription factors that are well-known regulators of hematopoietic stem and progenitor cell (HSPC) function 28 and have been implicated as oncogenes in AML 29 . We validated increased expression of HoxA7 and HoxA9 in HMGN1-OE cells by Q-RT-PCR (Supplementary Fig. 5b ).
ATAC-seq revealed that HMGN1 overexpressing myeloid progenitors had globally increased chromatin accessibility compared to wild-type cells (Fig. 2e , Supplementary Fig. 5c ). This increased ATAC-seq signal was particularly evident in the HoxA and HoxB clusters that were identified as targets of higher RNA expression in HMGN1-overexpressing cells (Fig. 2f ). We hypothesized that increased chromatin accessibility might also lead to changes in histone marks, consistent with what has been observed with HMGN1 expression in fibroblasts 30 . Therefore, we profiled chromatin marks during myeloid differentiation in the context of HMGN1 overexpression using Mint-ChIP. Mint-ChIP enables quantitative measurements of histone modifications by normalization to total histone H3 26 . Overexpression of HMGN1 was associated with global increased levels of H3K27 acetylation (H3K27ac), a marker of active transcription, at transcription start sites (TSSs) compared to wild-type cells, and this increase was maintained during induction of myeloid differentiation (Fig. 2g ). Higher global H3K27 acetylation was validated by western blotting and flow cytometry across multiple time points during differentiation (Fig. 2g ).
Although there was a global increase in chromatin accessibility and H3K27ac across the genome in undifferentiated cells, not all loci were affected equally, suggesting the possibility of additional specificity to the HMGN1 effect. Therefore, we performed an integrated analysis of RNA-seq, ATAC-seq, and Mint-ChIP, focusing on genes at the intersection of the most differentially expressed with the largest changes in accessibility and histone marks at their promoters. HoxA3, HoxA7, and HoxA9 ranked as those with the most significant overlapping differences between wild-type and HMGN1-OE cells (Fig. 3a, b , Supplementary Fig. 5d ). Loss of the repressive histone mark H3K27 trimethyl (H3K27me3), which is an expected reciprocal change in the setting of increased H3K27 acetylation, was also detected in the HoxA cluster (Fig. 3a ). We also observed dose-dependent binding of HMGN1 to chromatin at the HoxA7 and HoxA9 genes, in both HMGN1-OE compared to wild-type cells and in progenitors retrovirally overexpressing wild-type HMGN1 compared to a nucleosome-binding incompetent mutant (Fig. 3c ). Notably, nearly identical positions in the orthologous human HoxA cluster were previously implicated as a region of concentrated epigenetic regulatory activity in AML 31 .
Fig. 3: Chromatin and expression changes of Hox genes linked to HMGN1 are associated with the histone acetyl transferases (HATs) Cbp and p300.
a Gene tracks showing H3K27ac and H3K27me3 Mint-ChIP reads at the HoxA7 and HoxA9 loci at baseline and during differentiation in wild-type and HMGN1-OE progenitors. b Venn diagram overlapping RNA-seq, H3K27ac Mint-ChIP, and ATAC-seq sets of significant differences in HMGN1-OE vs. wild-type progenitors at baseline (log2FC > 1.2, p < 0.05). c ChIP-PCR in wild-type and HMGN1-OE myeloid progenitors (left) and wild-type progenitors overexpressing either myc-tagged HMGN1-WT or HMGN1-SEmut (right) for relative HMGN1 binding in the HoxA7 and HoxA9 loci. n = 2 or 3 biologically independent samples, as indicated. d Heatmap of expression of CBP (left) and p300 (right) target genes that are members of the Wong_Adult_Tissue_Stem_Cell gene set and enriched in HMGN1-OE progenitors, expressed as log2 fold change (FC) relative to WT_0 h. e ChIP-PCR in wild-type and HMGN1-OE myeloid progenitors for relative H3K27ac, H3K27me3, Cbp, and p300 binding in the HoxA7 and HoxA9 loci. n = 3 biologically independent samples. f ChIP-PCR in Ts1Rhr progenitors with Hmgn1 knockdown by shRNA for relative H3K27ac (top) and the HATs Cbp and p300 (bottom) in the HoxA7 and HoxA9 loci. n = 3 biologically independent samples, each measured once (conditions with three data points) or twice (six data points). All statistical comparisons are by two-sided t test. Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size image
HMGN1 is known to facilitate HAT activity in local chromatin environments by increasing accessibility to chromatin-modifying enzymes 30 . We found that many targets of the HATs CBP and p300 in murine hematopoietic cells 32 were upregulated in HMGN1-OE progenitors, including several genes important in HSCs and leukemia (e.g., HoxA and Hox B family, Meis1, and Msi2) (Fig. 3d ; left, CBP targets; right, p300 targets). We confirmed these associations using ChIP-PCR, which validated the Mint-ChIP findings of increased H3K27ac and decreased H3K27me3 at the HoxA7 and HoxA9 loci and also demonstrated enriched binding of CBP in the HoxA7 locus and CBP and p300 in the HoxA9 locus in HMGN1-OE cells (Fig. 3e ). We also confirmed that HMGN1 levels correlated with H3K27ac, and CBP and p300 binding in HoxA7 and HoxA9 loci in Ts1Rhr myeloid progenitors with and without knockdown of Hmgn1 (Fig. 3f , Supplementary Fig. 5e ). Together, these data suggest that the effects of HMGN1 in myeloid progenitors are most prominent at genes that are important in myeloid differentiation, HSC function, and leukemia, including but not limited to, Hox transcription factors.
HMGN1 promotes clonal expansion and HSC activity in vivo
Given its striking effects on differentiation and stem cell programs, we next asked how HMGN1 overexpression affects hematopoiesis and HSPC function in vivo. First, we analyzed HSPCs in HMGN1-OE and wild-type mice at 3 and 12 months of age. No significant differences were observed in young mice. However, older HMGN1-OE animals had higher frequency of immature, lineage-negative (Lin−) cells, Lin-CD117/c-Kit+ (LK) myeloid progenitors, and expansion of GMPs (Fig. 4a ).
Fig. 4: HMGN1 promotes expansion of hematopoietic precursors and provides clonal competitive advantage in vivo.
a Quantification, as percentage of total bone marrow mononuclear cells, of hematopoietic precursor subpopulations comparing young (3 months) vs. older (12 months) wild-type (n = 5 at 3 months, n = 10 at 12 months) and HMGN1-OE mice (n = 5 at 3 months, n = 6 at 12 months). b Schematic of the experimental approach for competitive bone marrow transplantation experiments. c Flow analysis of blood samples comparing CD45 isoforms from mice bled monthly after initial transplantation (first Tx, n = 10 mice) and subsequent serial re-transplantation (second Tx, n = 8 mice; and third Tx, n = 6 mice). d Flow analysis of HSPC populations in bone marrow samples 4 months after initial transplantation (first Tx, n = 5 independent animals) and subsequent serial re-transplantation (second Tx, n = 8 independent animals; and third Tx, n = 6 independent animals). Data are represented as comparison of CD45.1 (wild-type cells) vs. CD45.2 (HMGN1-OE cells). e Quantification of G0 cells within LSK and LK populations comparing wild-type and HMGN1-OE bone marrow cells in competition in vivo, n = 10 independent animals. f Analysis of CD11b by flow cytometry in CD117+ BM progenitors plated in liquid culture for 5 days (n = 5 independent animals). In all panels, genotypes compared by two-sided t test, in panel (d), *p < 0.0001. Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size image
To further explore the role of HMGN1 in hematopoiesis, we performed serial competitive bone marrow transplantation between HMGN1-OE (marked by CD45.2) and CD45.1(STEM) wild-type marrow (Fig. 4b ). CD45.1(STEM) is a C57BL/6 congenic strain with a single point mutation that confers recognition by CD45.1-specific monoclonal antibodies and that lacks the competitive disadvantage inherent in the B6.SJL-Ptprc(a)Pepc(b)/BoyJ-CD45.1 strain 33 . HMGN1-OE (CD45.2) transplanted cells had an increasing representation among mature blood cells over three serial bone marrow transplantations, indicating a progressive HSPC competitive advantage (Fig. 4c ). Four months after the initial transplantation, mice were sacrificed for analysis of CD45.2 (HMGN1-OE) and CD45.1 (wild type) representation in bone marrow populations (Fig. 4b ). In the primary transplant, we did not see a significant difference in the overall mature peripheral blood cell CD45.2/CD45.1 ratio after 4 months (Fig. 4c , first Tx). However, in bone marrow at the same time point, an advantage was present for HMGN1-OE hematopoietic progenitors, including in early multipotent progenitors (MPPs) (Fig. 4d ). In secondary and tertiary transplants (second Tx and third Tx), the HMGN1-OE HSPC competitive advantage progressively increased in all primitive and mature hematopoietic populations, including in the peripheral blood, suggesting a long-term HSC advantage associated with HMGN1 overexpression (Fig. 4c, d ).
To understand mechanisms underlying the HMGN1-OE competitive advantage and to determine if the epigenome changes we observed during ex vivo myeloid differentiation were also seen in intact primary hematopoiesis, we analyzed transcriptomes and histone marks from sorted HMGN1-OE and wild-type HSPCs that were in competition in vivo. GSEA of differentially expressed genes in LK cells revealed gene sets associated with mitochondrial activity enriched in HMGN1-OE cells (Supplementary Fig. 6a, b , Supplementary Data 4 ). Mitochondrial biogenesis is triggered with cell cycle induction in HSCs exiting quiescence and is linked to proliferation 34 , 35 . We also observed decreased expression of programs associated with HSC quiescence 36 and increased expression of genes upregulated in cycling HSCs 37 in HMGN1-OE HSPCs (Supplementary Fig. 6c, d , Supplementary Datas 5 and 6 ). Therefore, we tested whether HMGN1 overexpression was associated with decreased quiescence in HSPCs in vivo. We found that G0 quiescent subpopulations were reduced in both LSK and LK stem/progenitor cells from HMGN1-overexpressing animals (Fig. 4e ). These data indicate that HMGN1 overexpression may provide a competitive advantage to HSPCs associated with mitochondrial biogenesis and proliferation.
In most cases where alterations in cell cycle regulators provide a short-term HSPC proliferative advantage, the long-term consequence is premature exhaustion of HSC self-renewal in vivo 38 . However, HMGN1-OE HSPCs had both reduced quiescence and increased repopulating activity in serial transplantation, suggesting they had sustained self-renewal. In support of that hypothesis, HMGN1-overexpressing HSPCs had increased expression of markers of HSC activity, such as HoxA7 and HoxA9, and simultaneously decreased expression of genes associated with myeloid differentiation (Supplementary Fig. 6e , Supplementary Data 7 ). Expression differences were also associated with changes in H3K27ac at TSSs of genes identified by GSEA (Supplementary Fig. 6f ). Together, the gene expression and epigenomic changes in HMGN1-OE bone marrow HSPCs are similar to those we observed in progenitors studied by ex vivo myeloid differentiation (Fig. 2 ), supporting the in vivo relevance of HMGN1 effects. Similarly, we found that c-Kit+ myeloid progenitors from HMGN1-OE bone marrow had delayed differentiation compared to wild-type cells when cultured in vitro (Fig. 4f ). Collectively, these data support the role of HMGN1 in promoting clonal expansion in vivo associated with expression changes in genes involved in stemness, quiescence, and differentiation programs.
HMGN1 cooperates with AML oncogenes to increase LSC activity
HMGN1 overexpression impaired myeloid differentiation and conferred a competitive advantage to HSPCs but did not cause leukemia on its own (up to 18 months of observation). Therefore, we asked if HMGN1 overexpression cooperated with AML oncogenes in our established assays of myeloid differentiation. We first measured differences in HMGN1-OE or wild-type myeloid progenitor growth and differentiation in vitro in the presence of several AML oncogenes. Some (BCR-ABL and FLT3-ITD), were unaffected by HMGN1 overexpression (Supplementary Fig. 7a ). Others, such as AML-ETO9a, MLL-AF9, PML-RARα, NRAS G12V, GATA2 R396W, NPM1mut (exon 12 insertion 39 ), and MOZ-TIF2 cooperated with HMGN1 as seen by enhanced myeloid differentiation delay compared to either alteration alone (Fig. 5a , Supplementary Fig. 7a ). In addition, AML-ETO9a, NPM1mut, and MOZ-TIF2 cooperated with HMGN1 overexpression to increase colony numbers, suggesting enhanced self-renewal capacity. Increased colony formation by AML-ETO9a plus HMGN1-OE cells was maintained after multiple replating events (Fig. 5b , Supplementary Fig. 7b ). This enhanced self-renewal was associated with a lower frequency of differentiated cells in HMGN1-OE plus AML-ETO9a colonies (Supplementary Fig. 7b ).
Fig. 5: HMGN1 overexpression enhances LSC activity in AML-ETO9a-transformed leukemias.
a Analysis of myeloid differentiation by flow cytometry for CD11b and GR-1 in GFP+ wild-type and HMGN1-OE myeloid progenitors transduced with empty vector (EV) or AML-ETO9a-expressing retrovirus, each co-expressing GFP. n = 2 biologically independent cultures. b Serial replating assays of wild-type and HMGN1-OE cells transduced with AML-ETO9a or EV. n = 3 biologically independent replicates. c Analysis of myeloid differentiation by flow cytometry for CD11b and GR-1 in myeloid progenitors from Ts1Rhr (3 copies of Hmgn1) or Ts1Rhr_HMGN1+/− (2 copies of Hmgn1 but 3 copies of the other 30 genes in the Ts1Rhr triplication), each transduced with AML-ETO9a or EV. n = 2 biologically independent replicates. d Schematic of the experimental approach for AML-ETO9a in vivo leukemia studies. e Survival of primary recipients of wild-type or HMGN1-OE marrow transduced with AML-ETO9a or empty vector (EV). f Flow cytometry of HSPC phenotypes within GFP+ cells in spleens from moribund recipients of wild-type + AML-ETO9a or HMGN1-OE + AML-ETO9a transplants, n = 3 independent animals. g Quantification of H3K27ac levels by flow cytometry in the LT-HSC subpopulation of moribund animals’ leukemias, n = 8 independent animals for wild-type + AML-ETO9a; n = 5 independent animals for HMGN1-OE + AML-ETO9a. h LTC-IC assays showing number of colony formation units (CFU) obtained from leukemic spleens from either AML-ETO9a wild-type or HMGN1-OE transplants, n = 3 independent animals. i Quantification of GFP+ levels and cell surface markers for subpopulations within GFP+ cells in the peripheral blood 4 weeks after secondary transplantation of 100,000 GFP+ splenocytes of the indicated genotypes, n = 8 independent animals. *p < 0.0001 comparing lineage negative or CD11b+/GR-1+ populations between genotypes. j Survival after limiting dilution secondary transplantation of wild-type + AML-ETO9a or HMGN1-OE + AML-ETO9a leukemia cells at the indicated cell doses (100 k n = 8 independent animals; 10 k n = 6; 1 k n = 6, each representing two independent leukemias of each genotype). In all cases, samples compared by two-sided t test. Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size image
We confirmed these findings in two additional independent primary cell models. First, we found that loss of one copy of Hmgn1 in Ts1Rhr progenitors (Ts1Rhr_HMGN1+/−; three copies of the 21q22 genes, except only two copies of Hmgn1 16 ) restored myeloid differentiation and that the differentiation block by AML-ETO9a was greater in Ts1Rhr compared to Ts1Rhr_HMGN1+/− cells (Fig. 5c ). Second, in colony assays, depletion of Hmgn1 in Ts1Rhr cells by deletion of one copy (Ts1Rhr_HMGN1+/−) or by shRNA decreased the self-renewal activity associated with AML-ETO9a (Supplementary Fig. 8 ). Cooperation between HMGN1 and these specific fusion oncogenes is interesting given that AML-ETO promotes HSPC self-renewal and leukemia via CBP/p300 acetylation 40 and MOZ is a HAT that depends on HAT activity to inhibit senescence 41 . For these reasons, we hypothesized that AML-ETO9a and HMGN1 overexpression would cooperate in vivo.
We enriched CD117+ myeloid progenitors from 8-week-old wild-type or HMGN1-OE bone marrow and transduced the cells with a retrovirus expressing AML-ETO9a linked to GFP or with a GFP-only control, and then transplanted equivalent numbers of transduced cells into irradiated recipients (Fig. 5d ). Two months after injection, we sacrificed a subset of animals from each group and analyzed the relative frequency of hematopoietic progenitor and mature populations. We did not observe any significant differences between recipients of AML-ETO9a-transduced wild-type and HMGN1-OE bone marrow (Supplementary Fig. 9a ). However, upon culturing CD117+ cells under conditions that promote myeloid differentiation, we observed moderately increased expansion of stem/progenitor populations and decreased production of CD11b+ mature myeloid cells in HMGN1-OE cultures (Supplementary Fig. 9b ).
Beginning at approximately 100 days after transplantation, recipients of AML-ETO9a-transduced marrow began to display signs of systemic illness. Moribund animals had elevated frequency of GFP+ cells in the peripheral blood, bone marrow, and in the spleen, accompanied by splenomegaly (Supplementary Fig. 9c ). Although the overall survival was not significantly different in the primary transplant recipients of HMGN1-OE plus AML-ETO9a compared to wild-type plus AML-ETO9a (Fig. 5e ), we tested for phenotypic differences in the diseases. HMGN1-OE leukemias were enriched for LSK-like cells and particularly for the LSK subset having a multipotent progenitor (MPP)-like phenotype (Lin− CD117+ Sca-1+ CD150+ CD48−, Fig. 5f ). This suggested that HMGN1-overexpressing leukemias had a larger proportion of primitive, stem-like cells compared to wild-type leukemias. Moreover, similar to what we had observed in HMGN1-OE ex vivo immortalized progenitors and HSPCs in vivo, HMGN1-OE plus AML-ETO9a leukemias had increased H3K27 acetylation, particularly within the subset of leukemia cells that had an HSC-like phenotype (Fig. 5g ).
The expansion of phenotypically immature leukemia cells led us to hypothesize that HMGN1 overexpressing leukemias might be associated with increased LSC activity. Consistent with this prediction, in long-term culture initiating cell (LTC-IC) assays, an in vitro measure of HSPC activity that also correlates with survival in human AML 42 , HMGN1-OE plus AML-ETO9a leukemias had increased colony formation capacity (Fig. 5h ). To measure LSC activity in vivo, we performed limiting dilution transplantation of leukemias into secondary recipients. When we examined the peripheral blood 4 weeks after transplantation, while all animals had a large AML-ETO9a/GFP+ transplanted cell fraction (mean ~80%), we observed an expanded lineage-negative primitive hematopoietic population averaging 59% of the GFP+ mononuclear cells selectively in HMGN1-OE plus AML-ETO9a recipients (Fig. 5i ). In contrast, recipients of equivalent cell doses of wild-type plus AML-ETO9a secondary transplants did not have evidence of this circulating undifferentiated population and nearly all the GFP+ cells had a CD11b+ GR-1+ mature myeloid surface phenotype (Fig. 5i ). HMGN1-OE plus AML-ETO9a transplants caused fatal leukemia within 75 days in 8/8 secondary recipients that received 100,000 GFP+ cells, 8/8 that received 10,000 cells, and 6/8 that received 1000 cells, which correlates with an estimated leukemia-initiating cell frequency of 1:721 (95% CI 1:304–1:1714) (Fig. 5j ). At the same time point, no recipients of wild-type plus AML-ETO9a splenocytes at the same doses developed fatal leukemia. Thus, HMGN1 overexpression increased LSC frequency and capacity to transplant a fatal HSPC-like immature leukemia in AML-ETO9a cells.
Inhibition of HATs antagonizes HMGN1 effects
We next explored therapeutic implications of HMGN1-associated myeloid phenotypes. Given the striking increase in H3K27ac at HAT binding sites within loci of genes important for HSPCs and leukemia (Fig. 2 ), we evaluated the effect of genetic or chemical HAT inhibition in primary cell models. We hypothesized that reversing H3K27 hyperacetylation might counteract the HMGN1-associated differentiation block. First, we used a CRISPR-Cas9 approach to target Cbp or Ep300 in wild-type and HMGN1-OE myeloid progenitors and confirmed decreased expression (Fig. 6a ). Upon removal of E2, whereas neither Cbp nor Ep300-targeted sgRNAs affected wild-type cell differentiation, disruption of either gene promoted differentiation in HMGN1-OE cells (Fig. 6b ). Targeting of Cbp or Ep300 also decreased the aberrant global hyperacetylation of H3K27 in HMGN1-OE progenitors (Fig. 6c ). The pro-differentiating effect of HAT targeting in HMGN1-overexpressing cells was corroborated by morphologic assessment (Fig. 6d ).
Fig. 6: Targeting histone acetyltransferases CBP/p300 reverses HMGN1-associated myeloid differentiation abnormalities.
a Q-RT-PCR of Cbp and Ep300 expression after infection with CRISPR guides targeting each gene (n = 4 biologically independent guides) or control non-targeting guides (n = 2 biologically independent guides). b Effects of sgRNAs targeting Cbp (left panel) or Ep300 (right panel) compared to controls upon differentiation of wild-type and HMGN1-OE progenitors after E2 withdrawal. Lines are mean of biologically independent guide-expressing cultures (WT and OE sgControl = 2 guides, WT sgCbp and sgEp300 n = 2 guides each in biological duplicate, OE sgCbp = 2 guides, OE sgEp300 = 2 guides, one in biological duplicate). c Flow cytometry analysis of H3K27ac in wild-type and HMGN1-OE myeloid progenitors expressing sgRNAs targeting Cbp and Ep300 compared to controls. n = 3 or 4 biologically independent samples, as indicated. d Morphological assessment of Cbp and Ep300 sgRNA effects on myeloid differentiation 4 days after E2 withdrawal. Data are representative of 4 biologically independent experiments. e Effects of HAT inhibitors 0.5 µM C646 and 0.5 µM A-485 (compared to inactive analog A-486) on differentiation after withdrawal of E2 measured by flow cytometry for CD11b and GR-1. n = 3 biologically independent samples. f Morphologic assessment of C646 and A-485 effects on myeloid differentiation. Data are representative of three biologically independent experiments. g ChIP-PCR in wild-type and HMGN1-OE myeloid progenitors after 96 h of E2 withdrawal with or without C646 for relative H3K27ac at the HoxA7 and HoxA9 loci. n = 3 biologically independent samples. h Effects of HAT inhibitors 0.5 µM C646 and 0.5 µM A-485 on proportion of the indicated subpopulations of AML-ETO9a leukemic splenocytes derived from HMGN1-OE or WT progenitors. n = 3 biologic replicates. *p = 0.003 OE control vs. C646, p = 0.002 OE control vs. A-485; **p = 0.04 OE control vs. C646; #p = 0.0001 OE control vs. C646, p = 0.0001 OE control vs. A-485; ##p = 0.03 OE control vs. C646, p = 0.03 OE control vs. A-485; @p = 0.06 OE control vs. C646, p = 0.01 OE control vs. A-485. In all panels, samples compared by two-sided t test. Data are presented as mean values ± SD. Source data are provided as a Source Data file.
Full size table
Drug treatment assays
Cell lines were plated in a 96-well dish at a concentration of 2000 cells per 160 μl of media with drugs or vehicle (DMSO). Compounds were added to the cells in a serial threefold dilution. After incubation for 72 h at 37 °C, viability was determined by a 3-(4.5-dimehtylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma) assay according to manufacturer’s instructions. Briefly, 20 μl of 5 mg/ml MTT was added to each well and incubated for 2 h at 37 °C followed by addition of 100 μl of MTT lysis buffer and overnight incubation at 37 °C. Absorbance values were measured using a SpectraMax M3 plate reader at 570 and 630 nm. Viability curve values were analyzed by GraphPad Prism 8 (GraphPad Software, Inc.) by nonlinear regression analysis. Detailed information for the compounds is in Table 2 .
Table 2 List of compounds and other reagents.
Full size table
For in vitro differentiation, cells were treated with compounds or vehicle at doses and times indicated. Cells were harvested and washed with PBS before staining and then incubated with the antibodies indicated for 30 min in the dark at 4 °C followed by a final wash in PBS before flow cytometry analysis. In the case of murine peripheral blood samples, we initially lysed the samples twice with red cells lysis buffer prior to staining. For analysis of bone marrow or spleen samples, samples were lysed once. Antibodies for staining are indicated in Table 1 . Gating for HSPC subpopulations were defined as indicated in Table 3 .
Table 3 Markers of hematopoietic subpopulations.
Full size table
For analysis of H3K27ac levels by flow cytometry, cells were harvested and washed twice with PBS followed by fixation with 4% paraformaldehyde in PBS for 15 min on ice. Cells were pelleted by centrifugation for 5 min at 300g and resuspended in cold 70% EtOH added drop by drop while vortexing and then incubated overnight at −20 °C. Next, cells were washed twice with PBS and once with PBS/1% FBS. Cells were permeabilized for 20 min with 0.25% Triton-X100 in PBS/1% FBS followed by incubation with primary at 1:500 dilution for 30 min and then secondary antibody at 1:1000 dilution for 2 h at 4 °C in the dark. For analysis of cell cycle distribution, cells were initially stained with surface markers as described above and then permeabilized and fixed as indicated for H3K27ac analysis above. Next, cells were washed twice with PBS/2% FBS and stained with 1 μM Hoechst 33342 in HBSS (Hank’s balanced salt solution) supplemented with 20 mM HEPES pH 8 and 10% FBS. After 30 min of incubation at 37 °C, Pyronin Y was added to a final concentration of 1 μg/ml incubated for an additional 30-minutes. Cells were analyzed using a CytoFlex flow cytometer (Beckman Coulter) in most cases. For analysis of HSPCs, cells were analyzed using a FACS Canto II. In all cases, data was processed using FlowJo X software. For sorting of LK and LSK progenitor populations, cells were incubated with the antibodies indicated on each case and sorter in a FACS Aria II SORP. Table 1 includes complete information on antibodies.
Cellular ROS measurement
For measurement of ROS production during differentiation, we used the cell permeable reagent 2′,7″-dichlorofluorescein diacetate (DCFDA, Life technologies D399) following the manufacturer’s instructions. Briefly, cells were harvested and incubated with 25 μM DFCDA in culture media for 30 min at 37 °C and then analyzed by flow cytometry for fluorescence shift caused by DFCDA oxidation to DCF.
Cytospins for cellular morphology
Twenty thousand cells were washed and resuspended in 100 μl of PBS and centrifugation was performed at 10×g for 3 min on a CytoSpin 3 centrifuge (Shandon). Slides were air-dried and then stained for 5 min in May-Grunwald solution followed by a 2 min wash in PBS. Next, slides were stained for 15 min in a 1:20 dilution in Giemsa and washed briefly 2–3 times in H2O. Slides were air-dried before visualization.
Lentiviral and retroviral particles were produced in 293T cells using standard procedures with Lipofectamine 2000 (Invitrogen, 11668027). For retroviral production, 293T cells in a 10 cm plate were transfected with 12 μg of pECO-PAC together with 12 μg of the viral construct of interest and viral supernatant was harvested 48 h later. In the case of lentivirus production, 293T cells were transfected with 10.8 μg pPAX2, 2.4 μg pVSVg, and 10.8 μg of the viral expression plasmid of interest. Viral supernatant was harvested 48- and 72-h post transfection and concentrated by ultracentrifugation at 23,000g for 2 h at 4 °C. For infection, 200,000 actively dividing cells were spinfected in the presence of 1.5 ml of viral suspension and polybrene (final concentration 8 μg/ml) at 2000g for 90 min at 32 °C. To knock-down Hmgn1 in stably infected cells, shRNAs were induced with 0.5 μg/ml doxycycline for 96 h before performing assays.
cDNA and shRNA expression and CRISPR/Cas9 gene targeting
For constitutive expression of HMGN1-WT and HMGN1-SEmut (mutant protein unable to bind to nucleosomes 16 ) we used a pMSCV backbone, adding an in-frame FLAG tag using the primers indicated in Table 4 . For modulation of HMGN1 expression in Ts1Rhr cells by shRNA, we used the doxycycline-inducible tet on-pLKO system targeting the sequences indicated in Table 4 . For the CRISPR–Cas9 approach, we initially generated Flag–Cas9 stable cell lines by lentiviral infection of pCRISPRV2-FLAG-CAS9 (Addgene #52961) as described above, followed by selection with 2 μg/ml puromycin. Next, for targeting of either HMGN1 (human AML cell lines), or Ep300 or Cbp (murine myeloid progenitors) we used the pLKO5-RFP657 vector (Addgene #57824). Oligonucleotide sequences are in Table 4 .
Table 4 List of primers.