H1.3 depletion in AML cells prompts H1.2 redistribution, chromatin remodeling and cell cycle defects

Abstract

Linker histone H1 variants play critical, yet distinct, roles in chromatin organization and gene regulation. However, their specificity in cancer cells is not well understood. Our previous studies have shown that acute myeloid leukemia (AML) patients who were H3K27me3HIST1-positive exhibited lower H1.3 expression and a favorable outcome. These results imply that the H1.3 linker histone may play an unknown role in AML progression. In this study, we investigated the function of the H1.3 variant in AML cells. Through chromatin mapping and transcriptomic analyses, we revealed that H1.3 was enriched in regions with a high GC content and colocalized with the repressive mark H3K27me3. This supports its role in chromatin compaction and transcriptional repression. Knockout of H1.3 induced specific changes in gene expression profiles and chromatin dynamics, characterized by a changed in H1.2 localization, which was redistributed from its usual chromatin regions to H1.3 regions. Consequently, we observed chromatin alterations associated with changes in gene programs affecting interferon-related signaling and cell cycle regulation. Overall, our study revealed a mechanistic connection between H1.3 variant imbalance, immune response activation, and cell cycle regulation, with implications for our understanding of epigenetic regulation in AML cells.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13072-026-00679-w.

Introduction

Chromatin is a fundamental macro nucleoprotein complex that organizes the eukaryotic genome, controlling the accessibility of genomic information and some crucial biological processes, including DNA replication, transcription, the cell cycle, and DNA repair [1–3]. Thus, chromatin dynamics are primarily influenced by epigenetic modifications, which alter its structure and function [4, 5]. The basic repeating unit of chromatin is the nucleosome, which contains 145–147 base pairs (bp) of DNA wrapped in a left-handed 1.65 turns around an octameric histone complex. This octamer is formed by two copies of the core histones H2A, H2B, H3, and H4, organized as an (H3-H4)₂ tetramer flanked by two (H2A-H2B) dimers [6, 7]. Nucleosomes are interconnected by short DNA segments (20–90 bp) and the linker histone H1, which binds to the entry-exit sites of the nucleosome and plays an essential role in chromatin compaction and higher-order folding [8–10].

The linker histone H1 constitutes the most diverse and heterogeneous histone family compared with the highly conserved core histone proteins. As core histones, linker histone H1 can be further divided into different variants, dependent on the organism, in a dependent or independent-replicative manner. In humans, there are seven somatic H1 variants, from H1.1 to H1.5, H1.0, and H1.10. The replication-dependent variants (H1.1–H1.5) are predominantly expressed at the entry of the S-phase of the cell cycle, while H1.0 and H1.10 are expressed independently of replication. Among these variants, H1.2 to H1.5 are predominantly present in all tissues, whereas H1.1 displays tissue-specific expression. The independent somatic variants are mainly observed in differentiated cells [11, 12].

The classical concept of the linker histone H1 has revolved around its role in stabilizing nucleosomes and promoting the condensation of higher-order chromatin structures [13, 14]. This histone family exhibits great sequence variability across species and variants and a key challenge is to determine whether this variability confers chromatin specificity to the variants. For many years, the lack of specific antibodies for chromatin immunoprecipitation sequencing (ChIP-seq) has limited our ability to explore the role of H1 variants in chromatin dynamics fully. Nevertheless, several studies have demonstrated that different H1 variants perform specific functions in both chromatin structure and the regulation of various cellular processes [13, 15–17], although our understanding remains limited.

In addition to their structural and regulatory functions, linker histones have been connected to various diseases [18], including cancer [1, 19, 20], highlighting their biological importance. In this context, we previously described a novel epigenetic biomarker, H3K27me3HIST1, in patients with acute myeloid leukemia (AML) [21]. This epigenetic biomarker is characterized by a focal enrichment of the repressive histone mark H3K27me3 within the major histone gene cluster (the HIST1 cluster). This epigenetic alteration affected 11 histone genes in the HIST1 cluster, including the single-copy linker histone H1.3. Consequently, AML patients who were H3K27me3HIST1-positive exhibited lower H1.3 expression and a favorable outcome [22]. These findings suggested an unexplored role for the H1.3 linker histone in AML progression.

In this study, we investigated the functional role of H1.3 in AML cells using a CRISPR-Cas9-mediated knockout (KO) model in the NPM1-mutated OCI-AML3 cell line. Our findings reveal that the loss of this linker histone variant leads to a reorganization of transcriptional programs, including downregulation of cell cycle genes and upregulation of interferon-stimulated genes, as well as local changes in chromatin accessibility. Our results indicate that upon H1.3 depletion, the linker histone variant H1.2 partially redistributes across the genome. Our work demonstrates that the absence of H1.3 triggers chromatin and transcriptional changes, highlighting the non-redundant role of H1 variants in leukemia biology and may account for the favorable outcome of H3K27me3HIST1-positive patients.

Materials and methods

Cell lines and culture conditions

Human cell lines (OCI-AML3 and derived clones) were grown in MEMα medium supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 U/mL streptomycin under standard culture conditions (37 °C, 5% CO2). Cell lines were regularly verified to be Mycoplasma-free using PCR detection and thawed to perform experiments within a 6-to-10-passage period.

For CRISPR-Cas9-mediated H1.3 KO we created a customized editing construct using an H1.3 guide RNA (gRNA-AAGAAGGCAGGCGCAACTGC), which was cloned into the px458 CRISPR-Cas9 expression plasmid (Addgene #48138). The plasmid (px458-H1.3-gRNA) was transfected into the target cells by electroporation using the Amaxa Kit V (program X-001) according to the manufacturer’s protocol. The following day, transfected cells were identified based on the presence of green fluorescent protein (GFP) and cloned by sorting one cell per well on 96-well plates using the BD FACS Aria III cell sorter. The loss of H1.3 was evaluated in the different clones by Western blot. Mutational profiles of the clones were obtained by Sanger sequencing of the region targeted by the gRNA (Eurofins Genomics).

For OCI-AML3 cells expressing a V5-tagged H1.3 we cloned the H1.3 open reading frame (ORF) into a doxycycline-inducible plasmid pLix-403 (Addgene #41395) using the GATEWAY strategy and transfected the pLix403-H1.3-V5 construct into OCI-AML3 cells by electroporation using the Amaxa Kit V (program X-001). Transfected cells were selected 48 h after electroporation with puromycin 2 µg/mL. Expression of the H1.3-V5 was induced with 1 µg/mL doxycycline and assessed by Western blot. Doxycycline (Sigma; #D5207) was resuspended at 10 mg/ml in DMSO and stored at -20 °C. Intermediate dilutions were prepared in a culture medium before being added to the cells.

Protein extraction and western blot

Cells were washed and resuspended with ice-cold PBS and lysed in 2x Laemmli buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5% β-mercaptoethanol, and 0.5% bromophenol blue) at 95 °C for 8 min. To further dissociate chromatin, lysates were sonicated using an ultrasonic homogenizer (Bioruptor™ Next Gen; Diagenode; 5 cycles, 30 s of sonication followed by 30 s of rest) and the soluble fraction was collected by centrifugation at 16.100 g for 15 min at 4 °C. The samples were loaded and separated under denaturing conditions using SDS-PAGE with the Mini-Protean^® Tetra System, Bio-Rad, at 100 V with electrophoresis buffer (25 mM Tris pH 8.3, 192 mM glycine, and 0.1% SDS). After separation, proteins were transferred to nitrocellulose membranes using a transfer buffer (25 mM Tris, pH 8.3, 192 mM glycine, and 10% ethanol) at 200 mA for 2 h. The membranes were blocked with either 5% milk or 1% BSA in TBS-T (Tween at 0.05%) for 1 h at room temperature (RT). Subsequently, the membranes were incubated with different primary antibodies overnight at 4 °C, followed by three 10-minute washes with TBS-T at RT. Then, the membranes were incubated with a secondary antibody (mouse or rabbit, 1/10000) for 1 h at RT and were revealed with chemiluminescent reagent (ECL 1:1)-ECL Prime™ or ECL Select™ (GE Healthcare) according to the detected protein, for 5 min and then developed using the Amersham Imager 680. Band quantification from the obtained images was performed using the ImageJ Lab software.

The primary antibodies used were: H1.0 (#56695 Santa Cruz Biotechnology), H1.2 (ab4086 Abcam), H1.3 (# ab203948 Abcam), H1.4 (#41328S Cell signaling), H1.5 (ab18208 Abcam), H1 (#39708 Active Motif), H3 (ab1791, Abcam) pRB (#81805 Cell Signaling), RB (#93095 Cell Signaling), E2F1 (#3742S Cell Signaling); Cyclin E (#HE12 Sc247 Santa Cruz); V5-tag (Abcam; ab15828), Actin (#A1978 Sigma) and Tubulin (#T8535 Sigma). The polyclonal goat anti-mouse or anti-rabbit immunoglobulins HRP were used as secondary antibodies (Dako; P0447, P0448, respectively).

Flow cytometry

Cell surface markers were monitored/assessed by Flow cytometry using a SymphonyA3 cytometer (BD Biosciences) and analyzed using FlowJo software Version 10. Antibodies used were CD11B-PE (Mac-1), 3/100 (Beckman Coulter #IM2581U); CD38-Pacific Blue, 5/100 (BioLegend #356628) DRAQ7™, 1/400 (Biostatus #DR71000).

Cell cycle synchronization by double thymidine blocking and cel cycle analysis

Cells were synchronized in the late G1/early S phase using a double thymidine block method. To initiate the synchronization process, 300.000 cells/mL were seeded four hours prior to treatment. The first thymidine block was started by the addition of thymidine (Sigma; #T9250-5G, 100 mM stock solution stored at -20 °C) at a final concentration of 2 mM for 12 h. Cells were then washed 3 times with PBS and cultured in normal cell maintenance media for 12 h, followed by a second 12-hour thymidine (2 mM) block. Cells were released from the double thymidine block by washing and collected at different time points. For proliferation assay, cells were count at T0, 8 h, 24 h and 48 h after cell release. For cell cycle analysis, cells were collected every 2 h, fixed with cold 70% ethanol and stained with FxCycle™ Violet Stain (Invitrogen™ #F10347). The cells were washed and resuspended in a FACS buffer solution and cell cycle analysis was performed using a BD-LSRII cytometer (BD Biosciences), allowing us to analyze the different phases of the cell cycle based on DNA content. For western blot analysis, cells were collected at 2, 4 and 8 h after the released and lysed in 2x Laemmli buffer as previously described.

RNA extraction and RT-qPCR

Total RNA was extracted using the RNeasy^® Mini Kit from Qiagen (74104). RNA elution was carried out using RNase-free water and RNA concentration was measured at 260 nm using a spectrophotometer (Nanodrop 2000, ThermoFisher Scientific). The A260/A280 ratio was used to assess RNA purity, with pure RNA having an A260/A280 ratio within the range of 1.9 to 2.1.

Total RNA was reverse-transcribed using the Transcription High Fidelity cDNA Kit (Roche, #5091284001), following the manufacturer’s protocol. For the polymerase chain reaction, SYBR Green Fast Master Mix from Bio-Rad was used, along with a set of specific gene primers (Supplementary Table S1). The mRNA expression was normalized to the housekeeping gene HPRT. Relative expression levels were determined using the ΔCT (Cycle threshold) method, where ΔCT = Ct gene - Ct HPRT, and the relative gene expression was calculated as 2^(-ΔCT).

RNA sequencing (RNA-seq)

Data generation Total RNA was extracted from the WT OCI-AML3 parental cell line and two H1.3 KOs (KO clone #1 and KO clone #2) using the RNeasy^® Mini Kit (Qiagen; 74104), as previously described. Each experimental condition includes three biological replicates. Libraries were prepared with rRNA depletion with a Kapa mRNA HyperPrep kit (Roche, #KR1352-v7.21) and sequenced using the Illumina NovaSeq6000 platform, 100-bp paired-end reads, at 50 million reads per sample at Genomics and Bioinformatics Platform Marseille (Marseille, France).

Data analysis

Alignment Sequencing quality control was determined using the FastQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and aggregated across samples using MultiQC (v1.7) [23]. Reads with a Phred quality score of less than 30 were filtered out and mapped to the human (Hg38) reference genome using Subread-align (v1.5.0) [24] with default parameters. Gene expression quantification was performed by counting mapped tags at the gene level using featureCounts [25].

Differential analysis Differentially expressed genes were identified using DESeq2 (v1.26.0) [26] and were selected according to the following criteria: Padj < = 0.05, absolute (FC) > 1.5. For IGV visualization, mapped reads were converted into Bigwig using the Deep Tools suite (v2.2.4) (bamCoverage --scaleFactor library_size_factor --binSize 5) [27].

Functional enrichment analysis The Gene Ontology (GO) analysis was performed using clusterProfiler with a Padj < = 0.05 and by considering a background. Then, the genes were split into two groups according to their fold-change (UP genes: (FC) > = 1.5, DOWN genes: (FC) < = 1.5). The GO categories chosen were Biological Process (BP) and KEGG pathways.

The gene set enrichment analysis (GSEA) was performed with different gene lists as input to the GSEA software (v4.3.2). For the H1.3 KO gene lists, we considered the differentially expressed genes between H1.3 KO and WT conditions with Padj < = 0.05 and (FC) = > 1.5) and ranked them according to their (FC) of all genes detected in the RNA-seq experiments, which were used for the ranked gene lists.

For the H1.3 knock-down (KD) and multi-H1 KD gene lists, we considered the H1.3 KD DEGs obtained from GSE12299 Microarray [15] and the dataset from GSE83277 RNA-seq [28] (Supplementary Table S2), respectively, with Pvalue < = 0.05 and (FC) > 1.4 and ranked them according to their (FC).

For the H1.3 KO specific inflammatory signature, we considered H1.3 KO DEG (Pvalue < = 0.05 and (FC) = > 1.4) common to the GOBP immune response, inflammatory response and defense response gene lists. Volcano plots, GO enrichment, and Cnetplot figures were plotted by https://www.bioinformatics.com.cn/en, a free online platform for data analysis and visualization. KEGG pathways were created with Pathway-based data integration and visualization (https://pathview.uncc.edu/) [29]. Genes ID were obtained from https://www.genetolist.com.

ChIP sequencing (ChIP-seq) and ATAC sequencing (ATAC-seq)

Data generation For ChIP-seq we used protocols as previously described in [30]. In brief, 5 × 10⁶ OCI-AML3 cells (from WT and H1.3 KO clones) and 20 × 10⁶ H1.3-V5 OCI-AML3 cells were fixed with 1% formaldehyde for 8 min. After successive washes, cells were lysed with Lysis Buffer 1 (50 mM HEPES, 1 mM EDTA, 140 mM NaCl, 0.75% NP40, 10% Glycerol, 0.25% Triton X100), and Lysis Buffer 2 (200 mM NaCl, 10 mM Tris-HCl pH8, 1 mM EDTA, 0.5 mM EGTA) and chromatin extracted in Sonication Buffer (0.5% Laurylsarcosine, 10 mM Tris-HCl pH8, 1mM EDTA, 0.5 mM EGTA, 100 mM NaCl, 0.1% NaDoc). After sonication using the Bioruptor^® Pico (Diagenode) to obtain DNA fragments with an average size of 300 bp, chromatin was immunoprecipitated overnight at 4 °C with magnetic beads pre-incubated with the following antibodies: H3K27me3 (Cell Signaling; #9733S), H3K9me3 (Active Motif; #39161), H1.3 (Abcam; ab203948), H1.2 (Abcam; ab4086), H1 (Active Motif; #39707), V5-tag (Abcam; ab15828) and rabbit-IgG (Cell signaling; #2729S), used as a control. Immunoprecipitations were washed with the following combinations of wash buffers: W1 (1% Triton X100, 0.1% NaDOC, 150 mM NaCl, 10 mM Tris-HCL pH8), W2 (0.5% NP40, 0.5% Triton X100, 0.5% NaDOC, 150 mM NaCl, 10 mM Tris-HCL pH8), W3 (0.7% Triton X100, 0.1% NaDOC, 250 mM NaCl, 10 mM Tris-HCL pH8), W4 (0.5% NP40, 0.5% NaDOC, 250 mM LiCl, 20mM Tris-HCL pH8, 1mM EDTA, W5 (0.1% NP40, 150 mM NaCl, 20 mM Tris-HCL pH8, 1 mM EDTA), W6 (20 mM Tris-HCL pH8, 1 mM EDTA) and DNA was purified with an I-Pure V2 Kit (Diagenode).

Libraries were generated using the MicroPlex Library Preparation Kit (Diagenode) following the manufacturer’s instructions and the quality and size distribution of library fragments were analyzed on a Bioanalyzer 2100 system (Agilent). Sequencing was performed using the Illumina Nextseq500, 75-pb paired-end reads with 15 millions reads per sample at the Theories and Approaches of Genomic Complexity platform (Marseille, France) (for H1.3-V5 and H3K27ac ChIP) and using the Illumina Novaseq6000 platform, 100-bp paired-end reads, at 50 million reads per sample at Genomics and Bioinformatics Platform Marseille (Marseille, France) for WT and H1.3 KO clones.

For ATAC-seq we used the Diagenode ATAC-seq kit; (#C01080001) following the manufacturer’s instructions. In brief, 5 × 10⁵ OCI-AML3 cells from WT and H1.3 KO (clones #1 and #2) were harvested and resuspended in 500 µL of cold PBS containing protease inhibitors. Nuclei extraction was performed by adding to the cell pellet 50 µL of cold ATAC Lysis Buffer 1 containing 2% Digitonin for 90 s. Nuclei were collected by centrifugation, and a transposition reaction was performed by adding 50 µL of transposition mixture (containing Tagmentation buffer, Tagmentase, 2% Digitonin, 10% Tween20, PBS and Nuclease-free water) for 30 min at 37 °C. DNA was isolated using supplied spin columns and eluted in 12 µL of DNA elution buffer. Transposed DNA fragments were amplified using 8 UDI for Tagmented libraries (Diagenode; #C01011035) and libraries were purified using AMPure XP beads (Beckman Coulter; #A63881).

Libraries were generated using the MicroPlex Library Preparation Kit (Diagenode) following the manufacturer’s instructions and the quality and size distribution of library fragments were analyzed on a Bioanalyzer 2100 system (Agilent). Sequencing was performed using the Illumina Novaseq6000 platform, 100 bp paired-end reads, at Genomics and Bioinformatics Platform Marseille (Marseille, France).

Data analysis

Alignment Sequencing quality control was determined using the FastQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Paired-end reads were trimmed and aligned to the human hg38 reference genome using Bowtie2 (v2.3.5.1) [31] with default options. Unmapped reads, duplicates, and low-quality alignments were filtered out using SAMtools (v1.9) [32] (view; -exclude-flags 4). Duplicate tags were removed using Picard Tools (v2.20.2) (https://broadinstitute.g.ithub.io/picard/) (MarkDuplicates; REMOVE_DUPLICATES = true), and mapped tags were processed for further analysis.

BedGraph and BigWigs The resulting BAM files were sorted, and input-subtracted, counts per million (CPM)-normalized signal tracks were generated in bedGraph and bigWig formats using deepTools [27] (bamCompare --operation subtract --normalizeUsing CPM -- scaleFactorsMethod None). BEDTools [33] was employed to quantify the average ChIP-Seq signal of the samples within specific genomic regions of interest (i.e., G-bands, TEs, and 100-kb bins). All resulting files were visualized using the Integrative Genomics Viewer (IGV; v2.5.2) [34]. BigWig files from multiple replicates were merged into a single file using a custom bash script (bigWigMerge). All resulting files were visualized using the Integrative Genomics Viewer (IGV; v2.5.2) [34].

Downstream analysis Heatmaps of histone H1 variants were generated using the R package pheatmap, with Euclidean distance and complete cluster method used for clustering both rows and columns. Scatter plots were produced using base R, and Spearman’s correlation was used to assess associations between histone H1 variants and PTMs. Box plots were also created using base R, and the Wilcoxon signed-rank test was applied to compare ChIP-seq signals between WT and H1.3 KO conditions within the same subset of regions. Correlation matrices were computed using the cor() function in R (method = “spearman”) and visualized with the corrplot package. ChIP-seq profiles of H1 variants at meta-repeats were constructed and visualized using deepTools (v3.5.1) [27] (commands: computeMatrix, plotHeatmap, plotProfile). All regions were scaled to the same length using the scale-regions mode.

Genome annotation and segmentations To evaluate H1 variants abundance, different chromatin segmentations were used. Eight groups of Giemsa bands (G-bands) were defined according to [35]. Briefly, G-bands were classified as G positive (Gpos25 to Gpos100, according to its intensity upon Giemsa staining), and G-negative (unstained), which were further divided into four groups according to their GC content (Gneg1 to Gneg4, from high to low GC content).

Peak calling and differential analysis High confidence binding sites were determined using MACS2 peak caller (v2.1.2) [36] in broad mode (broad-cutoff = 0.05 -q 0.01 --nomodel). Each sample’s fragment size was previously computed by deepTools (bamPEFragmentSize) and used through the --extsize option. Coordinates shared by at least 2 samples (i.e. coverage > = 2) were selectively retained and merged into unified coordinates (bedtools merge). Mapped tags were then counted within these consensus coordinates using featureCounts (v1.6.4) and normalized using DESeq2 (estimateSizeFactors) (v1.26.0) [26]. Finally, differential analysis was performed (DESeq, minReplicatesForReplace = Inf, betaPrior = T).

TE meta-repeat profiles and heatmaps For repetitive elements analyses, the TE transcripts (v2.1.4) annotation was used [37]. Annotation includes nine different classes of repeats, including: LINE, SINE, LTR, DNA, Satellite, Other (SVA), Unknown, RC (Rolling-Circle), and RNA. Repeats with unsure classification named with a ‘?’ at the end of the family or class name (e.g. SINE? ) were excluded from the analysis. Repeats overlapping problematic regions defined in ENCODE BlackList [38] were also excluded from the analysis. For analysis performed at the TE classes, family and group level (i.e., meta-repeat profiles and heatmaps), multi-mapping reads were used as recently incorporated TE copies have not diverged enough to efficiently assign a unique position.

Genomic distribution of annotated peaks Genomic distribution was performed by Galaxy software (http://usegalaxy.org/) using the ChIPseeker tool. Significant peak changes were filtered with a P value < 0.05 and (FC) > 1. We used the nearest gene research to identify the genes that would be impacted by the difference in accessibility.

Gene Ontology analysis Gene Ontology (GO) analysis was performed using clusterProfiler with a Padj < 0.05 and by considering a background. Enriched GO terms were determined using only the genes that exhibited significant modulation (Pvalue < = 0.05). GO figures were done using https://www.bioinformatics.com.cn/en a free online platform for data analysis and visualization.

Multiomics analysis

RNA-seq and ATAC-seq. Differential analyses data from OCI-AML3 bulk RNA-seq and ATAC-seq were merged using a custom script. Briefly, DESeq2 output peaks from ATAC-seq differential analysis were annotated using both Homer (v4.11) [39] (annotatePeaks.pl) and ChIPseeker (v1.34.0) [40] (annotatePeak) annotation tools. Only TSS peaks were retained for further analysis. Bulk RNA-seq and ATAC-seq data were then merged regarding their gene names. ATAC-seq and RNA-seq log2 fold-changes were plotted against each other using ggplot2 (v3.4.1). Genes were highlighted using filtering thresholds regarding their relative p-values and log2 fold-changes. Only the top 10 candidates per quadrant were labelled using ggrepel (v0.9.3) (geom_text_repel).

ChIP-seq, RNA-seq and ATAC-seq The relationship between data from OCI-AML3 bulk RNA-seq, ATAC-seq, and ChIP-seq within 100-kb bins was assessed using a custom R script. First, bins were ranked by the log₂ fold-change in H1.2 ChIP-seq signal (KO vs. WT), and the top 10% (N = 2,946) were selected. To assess DEG enrichment in H1.2-associated regions, the 1,139 identified DEGs were mapped to 100-kb genomic bins using BEDTools. We quantified the number of bins containing at least one DEG in both the top bins and a subset of randomly selected bins (N = 2,946), and the association between H1.2-enriched bins and the presence of DEGs was evaluated using bar plots and a chi-square test. Statistical significance was further confirmed by a permutation test with 10,000 random reshufflings of bin labels. To assess ATAC-seq enrichment within 100-kb bins, the average genome-wide ATAC-seq signals in both WT and H1.3 KO conditions were mapped to the 100-kb bins using BEDTools. We then compared the average ATAC-seq abundance in top versus random bins using box plots and evaluated the statistical significance of the differences between WT and H1.3 KO conditions with a Wilcoxon signed-rank test.

Statistical analysis

Statistical analyses were done using R software (version 2.15.2) (The Comprehensive R Archive Network. http://www.cran.r-project.org/) and Prism 9 (Graph Pad Software) and the significance of the differences between groups was determined by unpaired t-Test, Mann–Whitney test or exact Fisher test. Data was presented as the median ± SEM.

Results

H1.3 depletion induces changes in gene expression related to cell cycle and inflammatory signals in OCI-AML3 cells

In order to study the role of H1.3 in AML, we used CRISPR-Cas9 approach to knock out (KO) H1.3 in the NPM1-mutated cell line, OCI-AML3 (Fig. 1A). We selected two H1.3 KO clones with a modified genome leading to the loss of H1.3 protein (Fig. 1B) and a moderate decrease of H1.3 mRNA (Supplementary Fig. S1A&B). Despite the diversity of H1s, several studies have shown that eliminating a single variant did not alter overall levels of total H1 due to compensatory effects by the other H1 variants [41, 42]. To study whether the absence of H1.3 caused a change in H1 variants, we checked H1 variant abundance in our H1.3 KO clones in comparison to OCI-AML3 cells. We did not detect any significant changes in the expression of the other variants, except for H1.0, which was increased, both at the mRNA (Supplementary Fig. S1B) and protein levels (Fig. 1B). Similar upregulation of the replication-independent H1.0 variant has been reported in breast cancer cell lines following knockdown of H1.2, H1.4, or multiple H1 variants (multi-H1KD), without being able to avoid consequences of H1 variant depletions [15, 28]. We then evaluated the effect of H1.3 depletion on gene expression, by RNA sequencing (Supplementary Fig. S1C). Differential expression analysis, comparing WT and H1.3 KO cells. identified 1,139 differentially expressed genes (DEGs; at least Padj < = 0.05, absolute (FC) > 1.5), including 770 genes upregulated in KO conditions and 369 genes in the WT condition. (Fig. 1C; Supplementary Table S3). Upregulated genes in the KO conditions exhibited enriched gene signatures related to defense against viruses and type I interferon signaling. Conversely, the upregulated genes in the WT condition showed enrichment in the cell cycle, particularly in mitotic nuclear division and mitotic sister chromatid degradation (Fig. 1D). Gene set enrichment analysis (GSEA) confirmed the enrichment of the inflammatory response, highlighting the interferon-gamma and -alpha response in the H1.3 KO, while the WT condition was enriched for cell cycle-related signatures (Fig. 1E). Notably, compared to the H1.3 KO conditions, the WT cells showed an enrichment of the DREAM complex and E2F target signatures (Supplementary Fig. S2A). Analysis of the overlap of the most representative signatures shows that 6% of the overexpressed genes in the H1.3 KO conditions were involved in the gamma and alpha interferon response, while 9% of the underexpressed genes were involved in the cell cycle (Fig. 1F; Supplementary Table S4). To validate the specific effect of H1.3 deletion on gene expression, we analyzed published transcriptomic data from an H1.3 KD of a human breast cancer cell line (22) (Supplementary Table S2). As with H1.3 KO AML cells, H1.3 KD breast cancer cells demonstrated a down-regulation of cell-cycle effectors involving E2F targets and DREAM complex signatures (Supplementary Fig. S2B). The interferon-gamma and alpha response signatures were not retrieved in the H1.3 KD breast cancer cell. However, when we tested our H1.3 KO inflammatory signature (Supplementary Table S4B) in the multi-H1 KD breast cancer model [28], we observed an enrichment of our inflammatory signature in the multi-H1 KD model (Supplementary Fig. S2C), suggesting a common pattern whereby depletion of H1 variants influences inflammatory responses in different cell types.

Fig. 1.

H1.3 depletion induces changes in gene expression related to cell cycle and inflammatory signals in OCI-AML3 cells. A Scheme of clone selection after CRISPR-Cas9 editing downstream of the ATG of H1.3 in the OCI-AML3 cell line. B Western blot analysis of the H1 histone variant protein expression in KO1 and KO2 clones using H1.3, H1.0, H1.2, H1.4 and H1.5 specific antibodies, with total H1 and H3 used as loading controls. C Volcano plot displaying differentially expressed genes (DEGs) between WT OCI-AML3 (n = 3) and the 2 H1.3 KO clones (n = 3 per clones) considering Padj < 0.05, | log2(fold change) | ≥ 0.585. Genes upregulated in the WT condition are colored blue, and those upregulated in the H1.3 KO conditions are colored red. D Gene ontology (GO) enrichment and KEGG pathway analysis of upregulated genes in WT (blue) and upregulated genes in H1.3 KO clones (red). E Gene set enrichment analysis (GSEA) of WT and H1.3 KO DEGs against interferon-gamma and defense responses, cell cycle and signatures. F Venn diagrams showing overlap between DEGs in H1.3 KO cells and functionally annotated gene sets. Left: overlap between genes overexpressed (UP) in KOs and interferon-stimulated genes (ISGs). Right: overlap between downregulated genes (DN) in KOs and genes involved in cell cycle checkpoints

Taken together, these results highlight a role for H1.3 in controlling cell-cycle progression and reveal that its absence stimulates interferon signaling and inflammatory response, representing the first documented link between histone H1.3 and these pathways.

H1.3 depletion modifies chromatin accessibility in AML cells

As H1.3 participates in chromatin compaction [43, 44], we studied the impact of its absence on chromatin accessibility using Assay for Transposase-Accessible Chromatin with sequencing (ATAC-seq) in our cellular model. Overall, we did not find significant difference in the distribution of accessible regions across the genome between H1.3 KO and WT conditions (Supplementary Fig. S3A). However, differential peak analysis revealed differences in chromatin accessibility regions upon H1.3 depletion. These were characterized by an increase in chromatin accessibility in 214 regions under the H1.3 KO condition and a decrease in 78 regions (Supplementary Fig. S3B; Table S5).

Next, to assess the potential relationship between the gain or loss of open chromatin regions identified by ATAC-seq and changes in gene expression, we compared differential ATAC peaks with DEGs. We separated the peaks into two groups: intragenic peaks, which include transcription start sites (TSSs) and may capture chromatin accessibility associated with transcription initiation or regulation within gene bodies, and intergenic peaks, which constitute a set of potential distal regulatory elements. We categorized the genes into 4 groups based on the direction of significant changes in ATAC-seq and RNA-seq signals (Supplementary Table S6). We then focused our analysis on genes displaying either more open chromatin together with upregulation of the gene expression, or, at the opposite, a more closed chromatin accompanied by a gene expression downregulation (Supplementary Table S7). We revealed that differences in accessibility in intragenic regions were more strongly associated with differentially expressed genes (DEGs) than differences found in intergenic regions (Fig. 2A). Notably, TRIM68 was consistently downregulated and associated with reduced chromatin accessibility at the TSS, whereas WT1-AS was upregulated andshowed increased accessibility at the TSS in the H1.3 KO conditions (Fig. 2B).

Fig. 2.

H1.3 depletion impacts chromatin accessibility. A Venn-diagrams showing the overlap between differential ATAC-seq peaks and differentially expressed genes (DEGs) in H1.3 KO cells compared with H1.3 WT cells. ATAC peaks were categorized according to their genomic context: transcription start site (TSS) + intragenic regions and intergenic regions. B Coverage plots of ATAC and RNA-seq signals per condition (WT and H1.3 KO) associated with TRIM68 and WT1-AS genes. Grey boxes indicate regions with significantly different signals

Taken together our analysis suggests modest correlations between changes in chromatin accessibility and gene expression induced by H1.3 KO, which may indicate that H1.3 has other genomic functions.

H1.3 shows a distinct genomic distribution and its loss alters H1.2 localization

To further elucidate the genomic function of H1.3 in AML cells, we performed ChIP-seq to localize H1.3 in OCI-AML3 cells, using both endogenous H1.3 and ectopic V5-tagged H1.3 in an inducible V5-tagged H1.3 expressing OCI-AML3 cell line (Supplementary Fig. S4A). We confirmed the expected “H1 valley” around the TSS for endogenous H1.3 and V5-tagged H1.3 (Supplementary Fig. S4B), as previously reported for other H1 variants [17, 45, 46]. Next, we characterized the repercussions of H1.3 KO on the genomic localization of H1.2 and total H1 along with the distribution of repressive histone marks (H3K27me3 and H3K9me3). H1.2 was chosen as a representative of replication-dependent H1 variants due to its abundant expression in all cell types. An antibody that recognizes all H1 variants was used to study the general behavior of H1 variants. To measure and classify the H1 variants within the genome, we used the G-bands classification based on GC content, where GC-rich bands are linked to euchromatin and transcriptional activity, while GC-poor bands correlate with heterochromatin and repressive features [35]. Analysis of H1 distribution showed that both endogenous H1.3 and H1.3-V5-tagged were enriched in GC-rich regions, whereas H1.2 (and total H1) was predominantly found in GC-poor regions (Fig. 3A; Supplementary Fig. S4C). This result confirms the specificity and differential nature of the genomic localization of the two histone linkers, H1.2 and H1.3. Interestingly, upon H1.3 KO, the enrichment of H1.2 in GC-poor regions was reduced, suggesting a redistribution of H1.2 in the absence of H1.3 (Fig. 3B). To better characterize this redistribution, we analyzed the correlation between the endogenous H1.3 and H1.2 localization in WT and H1.3 KO conditions in 100-kb bins. In WT cells, H1.2 and H1.3 exhibited a modest correlation in their distribution (rs = 0.463), and H1.2 localization was mildly affected by H1.3 KO (rs = 0.878), thereby confirming the specificity of action of the two H1 variants (Fig. 3C). Noticeably, the genomic localization of H1.2 in the H1.3 KO condition exhibited a stronger correlation with H1.3 localization than in the WT condition (rs = 0.748 in H1.3 KO vs. rs = 0.463 in WT) (Fig. 3C). This observation suggested the possibility of redistribution of H1.2 to occupy sites previously occupied by H1.3. Interestingly, and despite an increase in H1.0 expression, the loss of H1.3 did not affect H1.0 abundance on GC-regions (Supplementary Fig. S4D), its limited correlation with H1.3 (Supplementary Fig. S4E), or its genomic distribution (Supplementary Fig. S4F). This highlights the unique way in which H1.3 and H1.2 are connected at the genomic level.

Fig. 3.

H1.3 has a distinct genomic distribution, and its loss affects H1.2 localization. A Heatmap and cluster analysis of the input-subtracted ChIP-Seq abundance (scaled) of H1 variants within Giemsa bands (G-bands) from WT or H1.3 KO OCI-AML3 cells. ChIP was performed with Specific antibodies for H1.2, H1.3, V5 or total H1 were used. The Y-axis annotation indicates which group the G-band belongs to, as indicated in the legend. The relative GC content of the different G-bands is indicated. Data used for this analysis corresponds to the averaged values obtained from two independent ChIP-seq biological replicates. B Scatter plots of H1.2 (from WT or H1.3 KOs) and of H1.3 (from WT) input-substracted ChIP-Seq abundance and GC content at each individual G band color-coded as in A. Pearson’s correlation coefficient is shown as well as the p-value. C Scatter plots of H1.2 (from WT or H1.3 KOs) and of H1.3 (from WT) and input-substracted ChIP-Seq abundance within 100-kb bins of the human genome. Spearman’s correlation coefficient is shown as well as the p-value. D Spearman’s correlation plots between the repressive histone marks (H3K9me3 and H3K27me3) and the histone variants (H1.2 and H1.3) input-subtracted ChIP-Seq signal in WT and H1.3 KO cells within 100-kb bins. E Spearman’s correlation coefficients between all the ChIP-Seq features analyzed in 100-kb bins. Only correlations with p < 0.01 were considered (colored circles in the correlation matrices)

Next, we compared the association of H1.2 and H1.3 with the two repressive marks, H3K9me3 and H3K27me3. As previously reported in other tissues [44, 47], in AML cells, H1.2 correlated identically with both H3K27me3 and H3K9me3 (rs = 0.675 and rs = 0.66, respectively) (Fig. 3D). By contrast, H1.3 correlated poorly with the repressive H3K9me3 mark (rs = 0.242) while being more correlated with H3K27me3 (rs = 0.669) (Fig. 3D), suggesting that H1.3 is associated preferentially with H3K27me3 along the genome. Interestingly, and consistent with a potential H1.2 relocalization to H1.3 sites, H1.2 gained in association with H3K27me3 (rs = 0.771 in H1.3 KO vs. rs = 0.675 in WT) while losing in association with H3K9me3 upon H1.3 KO (rs = 0.5651 in H1.3 KO vs. rs = 0.66 in WT) (Fig. 3D). The correlative analyses, summarized in a Spearman correlation graph (Fig. 3E), illustrate our various observations. H1.3 does not colocalize with H1.2; however, its deletion induces the displacement of H1.2 to sites that are free of H1.3. Furthermore, the results support the similarity of the localization of ectopic H1.3 labelled with V5 to that of endogenous H1.3. Finally, it shows that the behavior of total H1 also reflects that of H1.2.

Overall, these results emphasize the non-random genomic localization of H1.3 and its association with repressive histone marks that can be partially replaced by H1.2. This indicates the specificity of the action of histone-binding variants.

H1.3 deletion remodels H1.2 localization in regions where gene expression is deregulated

To further investigate the effect of H1.3 deletion on H1.2 localization, we focused on the top 10% bins enriched in H1.2 signal upon H1.3 KO in comparison to WT (Fig. 4A). We showed that the gain in H1.2 ChIP-seq signal upon H1.3 KO was significant in this top 10% bins in comparison to randomly selected bins (Fig. 4B). This was also true for the more stringent 1, 3, or 6.5% top H1.2-enriched bins (Supplementary Fig. S5A). In addition, we observed that the H1.3 ChIP-seq signal was markedly higher in the top 10% bins compared with the random bins (Fig. 4C). This suggests that chromatin regions acquiring H1.2 in the H1.3 KO condition were previously enriched in H1.3, but not H1.2. This result supports the view that H1.2 redistributes in a compensatory manner upon H1.3 depletion. Additionally, both H3K27me3 and H3K9me3 marks showed alterations in the top and random bins upon H1.3 KO, suggesting broader chromatin alterations beyond H1.2 redistribution (Supplementary Fig. S5B&C). Next, we tested whether regions that increased H1.2 abundance upon H1.3 depletion also changed their chromatin accessibility. We observed that these bins decreased chromatin accessibility in the H1.3 KO condition; however, this decrease was not specific to the top 10% of bins, as we also observed the same effect in random bins (Fig. 4D).

Fig. 4.

H1.3 deletion remodels H1.2 localization in regions where gene expression is deregulated. A Scatter plots of H1.2 input-subtracted ChIP-Seq abundance from WT or H1.3 KO cells within 100-kb bins of the human genome. Spearman’s correlation coefficient is calculated for the total of bins (N = 29,452; labeled light blue) or for the top 10% bins with the highest increase of H1.2 in KO compared to WT (N = 2,946; labeled dark blue). For the selection of the top 10% H1.2 bins, the highest KO and WT variations in H1.2 absolute RPKM values were taken into account without limitation for RPKM > 0 or RPKM < 0, respectively. B Boxplot showing H1.2 ChIP-seq signal in the top 10% bins, comparing the WT and the H1.3 KO conditions. An equivalent number of randomly selected bins is shown as a control (random). C Boxplot showing the basal H1.3 ChIP-seq signal (from WT) in the top 10% H1.2 enriched bins (in B) in comparison to randomly selected bins, indicating that bins that gained H1.2 upon H1.3 KO were previously occupied by H1.3. D Boxplot showing the ATAC-seq signal in the top 10% H1.2-enriched and randomly selected bins in WT and H1.3 KO conditions, showing increased accessibility in top bins, but general reduced chromatin accessibility upon H1.3 KO. E A chi-squared test of independence was performed to test the association between the top 10% H1.2-enriched bins and DEGs. Out of 2,946 bins, 243 overlap with DEGs, which is higher than would be expected if the bins were chosen at random. A permutation test is shown in Supplemental Figure S6A. DEGs within the top H1.2-enriched bins represent the 36% of all downregulated genes and the 14% of all upregulated genes upon H1.3 KO. F GO term enrichment analysis of DEGs located within the top 10% H1.2-enriched bins. Upregulated genes are enriched for interferon signaling pathways, represented in red, while downregulated genes are associated with cell cycle regulation, represented in blue. G IGV snapshots of input-subtracted ChIP-Seq H1.3 (from WT) and H1.2 (from WT and H1.3 KO) signal at two representative DEGs, which are located within the top 10% H1.2-enriched bins. The grey boxes highlight chromatin region where H1.2 is redistributed upon H1.3 KO

We then assessed whether these regions, characterized by increased H1.2 abundance and previously occupied by H1.3, were associated with changes in gene expression. We selected the top 10% of bins with the highest increase in H1.2 signal when comparing H1.3 KO versus WT conditions, as previously described, and crossed them with the DEGs. Of the 2,946 top 10% of bins, 243 (8.25% of bins) overlapped with DEGs, which was more than would be expected by chance (as 1,139 DEGs represent 4.03% of the total number of annotated genes in the genome), indicating an association between changes in histone variant occupancy and gene expression (Fig. 4E). Statistical significance of this association was further confirmed by a permutation test with 10,000 iterations (Supplementary Fig. S6A), indicating a strong association between changes in histone variant occupancy and gene expression. More precisely, 132 of these bins were associated with downregulated genes (132 genes, 35.8% of all downregulated genes), and 106 bins were associated with upregulated genes (101 genes, 13.6% of all upregulated genes). Moreover, 5 bins were associated with both upregulated and downregulated genes (Supplementary Fig. S6B; Table S8). In summary, the top 10% bins enriched in H1.2 upon H1.3 KO were enriched in DEGs, especially downregulated genes, supporting the idea that H1.2 redistribution may be associated with gene repression, possibly due to its stronger repressive potential compared to H1.3 in this cellular context. To explore the biological relevance of these changes, we performed GO enrichment analysis on DEGs located in the top H1.2-enriched bins. Notably, several of those genes were already highlighted in our global RNA-seq analysis (Supplementary Table S3). Specifically, upregulated DEGs located within these bins were significantly enriched in interferon signaling, while downregulated DEGs were enriched in cell cycle regulation (Fig. 4F; Supplementary Table S9). To provide representative examples of these chromatin changes at specific loci, we visualized ChIP-seq signals using IGV. We selected IRF9, an interferon-stimulated gene upregulated in the H1.3 KO condition, and BIRC5, a key regulator of the cell cycle that is downregulated upon H1.3 depletion (Fig. 4G). Additional examples of these chromatin changes at specific loci are provided in (Supplementary Fig. S7), illustrating the redistribution of H1.2 at H1.3 sites when H1.3 is depleted.

Taken together, these results suggest that H1.3 depletion in AML cells promotes the redistribution of H1.2 in regions previously occupied by H1.3. Besides, this redistribution is associated with changes in gene expression, characterized by more downregulated genes than upregulated genes, and affecting pathways related to interferon signaling and cell cycle regulation, either directly or indirectly.

H1.3 loss is linked to transcriptional activation of transposable elements (TEs) and H1.2 relocation to H1.3-enriched elements

H1 linker histones have been implicated in regulating transposable elements (TEs) across different systems [17, 43, 48]. Here, we examined the impact of H1.3 depletion on TEs expression. Differential expression analysis revealed changes in multiple TE families, including LINEs, RCs, DNA elements, and mainly from LTRs, as shown in the MA plot, with several of these TEs being upregulated in H1.3 KO cells (Fig. 5A; Supplementary Table S10). To explore whether this effect was associated with changes in histone H1 variant distribution, we examined the abundance of H1 variants (H1.2 and H1.3) across the different TE element classes and families in WT and H1.3 KO OCI-AML3 cells using our ChIP-seq data. In WT cells, H1.3 was highly enriched at the SVA class, while H1.2 was not enriched at SVA but was more abundant at RNA elements (Fig. 5B). Neither variant was considerably enriched at DNA, RC, or LINE classes (Fig. 5B&C; Supplementary Fig. S8A). Interestingly, in the absence of H1.3, the H1.2 variant showed a redistribution, with increased enrichment at SVA families in H1.3 KO cells (Fig. 5B&C), and decreased at some SINE Alu elements (Supplementary Fig. S8B). This redistribution of H1.2 in H1.3 KO cells was confirmed by meta-repeat profile analysis (Fig. 5D). Upon H1.3 depletion, H1.2 was also enriched at some satellite families (acro, telo, etc.) where H1.3 was abundant (Fig. 5B&C). This may suggest that H1.3 was relevant at some TEs and H1.2 takes its role upon H1.3 KO. This finding may explain why the deregulation of TEs is mild and no deregulation of SVA or satellites was observed (Fig. 5A).

Fig. 5.

H1.3 loss is linked to transcriptional activation of transposable elements and H1.2 relocation to H1.3-enriched elements. A MA plot showing differential expression of transposable elements (TEs) in WT and H1.3 KO OCI-AML3 cells. Significantly deregulated TEs (Padj < 0.05) are highlighted and colored according to their TE class. B, C Heatmap and cluster analysis of the average input-subtracted ChIP-seq abundance of H1 variants in WT and H1.3 KO cells across TE classes in (B) and TE families in (C). In panel C, the TE class to which each TE family belongs is indicated by color in the first column. D Meta-repeat profile of input-subtracted ChIP-seq signal of H1 variants in WT and H1.3 KO cells at SVA repeats and their 3Kb-flanking regions. All SVA copies were scaled to a uniform length for visualization. In the heatmaps, each row corresponds to a single SVA element (separated in SVA_A to SVA_F elements), ordered based on the H1 binding profile in each condition

H1.3 KO alters G1/S transition in synchronized leukemic cells and induces a more mature phenotype

To investigate the biological effect of these deregulations, we assessed the impact of H1.3 KO on cell proliferation and differentiation. We first conducted proliferation assays and demonstrated, by measuring the cell division rate over 24 h, that the two KO clones had a slower cell division rate compared to WT cells. This reduction in the doubling rate resulted in slower cell growth over 96 h of culture (Fig. 6A). Next, to precise the effect of H1.3 depletion on cell cycle progression, we synchronized the cells at the G1/S boundary by double-thymidine block followed by cell cycle release [49]. Cellular counts after the thymidine block revealed a significant difference 8 h after the release suggesting an alteration in progression through the cell cycle in KO conditions (Fig. 6B). To precise the effect of H1.3 depletion on cell cycle progression, we monitored the cell cycle progression by DNA content staining (Fx violet) at 0, 4 and 8 h after release of the double-thymidine block. As expected, T0 was enriched in cells in G1 phase, T4 in S phase, and T8 in G2/M phase. A difference in cell cycle progression between WT and KO was observed at T4, characterized by an increase in the H1.3 KO cells in G1 at the expense of the S phase compared to WT cells, suggesting an alteration of the G1/S transition in the absence of H1.3 (Fig. 6C). Consistent with the observed difference in proliferation, we demonstrated overexpression of three inhibitors essential for the proper progression of the cyclin-dependent kinase cycle, namely CDKN1A, CDKN2A, and CDKN2B, in H1.3 KO cells (Fig. 6D). Then, we investigated the effect of H1.3 deletion on cell cycle regulator protein expression after thymidine block. We showed that E2F1, a transcription factor driving S-phase entry, was significantly reduced in H1.3 KO cells at T0 and T4, suggesting an impaired activation of the G1/S program (Fig. 6E). Accordingly, the phosphorylated form of RB (pRB) and the Cyclin E (CCNE1), a marker of the G1/S transition, decreased in KO cells, witnessing a default in the G1/S transition observed at 4 h after Thymidine release. Since we had previously demonstrated that H1-3 knockdown (KD) promotes granulocyte differentiation in AML cell lines treated with all-trans retinoic acid (ATRA) [22], we evaluated the effects of complete H1-3 knockout on ATRA-induced differentiation. We observed higher expression of the two differentiation markers, CD11b and CD48, in H1.3 KO cells both before and after ATRA treatment (Fig. 6F).

Fig. 6.

The H1.3 deletion affects the G1/S transition of the cell cycle and differentiation. A Proliferation assay of WT (blue) and KO clones (red). On the left panel, the cells were diluted to 0.25 × 10⁶/ml after each 24 h period for 48 h. Each point represents one count from 4 biological replicates. Statistical significance was calculated using a t and Wilcoxon test considering that the theoritical mean was the WT one, * p < 0.05. On the right, cell concentration was assessed by counting them every 24 h. Data represent mean ± s.e.m. of n = 4 biological replicates. Statistical significance was estimated using a two-way Anova test, ** p < 0.01, *** p < 0.001. B Proliferation after synchronization. Cells were synchronized with double thymidine block and cell concentration was measured at T0, 8 h, 24 h and 48 h after cell release. Data represent mean ± s.e.m. of the growth of 2 biological replicates. Statistical significance was estimated using a two-way Anova test, ** p < 0.01, *** p < 0.001. C Cell-cycle analysis after double-thymidine block. The left panel shows typical Fx-Violet fluorescence profiles for the two conditions at different time points (2, 4, 6 and 8) after release. The right panel shows a comparison of the evolution of cells in the different phases of the cell cycle between WT and H1.3 KO clone at T0, T4, and T8 . Data represent mean ± s.e.m. of n = 6 biological replicates. Statistical significance was estimated using an unpaired two-tailed t-test, * p < 0.05; **p < 0.005. D Analysis of cyclin inhibitor expression. Genes were quantified by RT-qPCR in WT (n = 6), H1.3 KO1 (n = 3) and H1.3 KO2 (n = 3) OCI-AML3 cells. mRNA levels were normalized to the average of the housekeeping gene HPRT. Data represent mean ± s.e.m. Statistical significance was estimated using an unpaired two-tailed Mann Whitney test * p < 0.05; **p < 0.005; ns = no significance. E Western blot analysis of cell cycle regulators RB, phospho-RB (pRB), E2F and Cyclin E (CCNE1) at 2, 4, and 8 h after thymidine block release in WT and H1.3 KO cells. Bottom panel shows the quantification of the blots, n ≥ 2 biological replicates. Statistical significance was estimated using a Welch’s t-test, * p < 0.05. F Cell surface marker analysis of CD38 and CD11b by FACS of WT (blue) and H1.3 KO (red) clones with either DMSO or ATRA (2µM). Intensity of fluorescence and percentage of positive cells are presented

Altogether, our results show that loss of H1.3 alters the cell cycle progression while promoting a more mature phenotype, suggesting an important role of H1.3 expression level in AML cells.

Discussion

The linker histone H1 family has recently emerged as an important regulator of chromatin architecture and transcription in various biological processes, including cancer. Building on our previous work demonstrating that downregulation of H1.3 contributes to the pathogenesis of AML [22], this study defines the chromatin functions of H1.3 in AML cells. Our work reveals that, in many respects, H1.3 shares several characteristics with other H1 variants. For example, it exhibits characteristic depletion at the transcription start sites of highly expressed genes (‘H1 valley’) and enrichment at the ends of gene bodies [35, 46, 50, 51]. However, our results also highlight notable contextual differences that underscore the distinctive role of H1.3 in AML. A key feature is the genomic distribution of H1.3. We observed that H1.3 is associated with regions marked by H3K27me3, which is consistent with its role in chromatin compaction and transcriptional regulation [44, 52] and concordant with recent studies reporting enrichment of linker histones in Polycomb domains [53]. However, the chromatin distribution of H1.3 in AML differs significantly from those reported in breast [54], kidney [46], and murine embryonic stem cells [50], suggesting that the H1.3 chromatin functions are highly dependent on cell type. One of the characteristics of our AML model is that H1.3 shows only a weak correlation with H1.2 in terms of chromatin occupancy. While H1.2 is preferentially localized in constitutive heterochromatin that is GC-poor and enriched in H3K9me3 [47, 55, 56], H1.3 is enriched in GC-rich regions that are usually associated with open chromatin [35]. This finding underscores the complexity of GC-rich domains, which can harbor both active genes and repressed Polycomb targets [57, 58], and highlights the functional diversity among H1 variants.

Previous studies have suggested that there is compensatory adaptation between H1 variants depending on which variants are expressed [59]. In our AML model, loss of H1.3 does indeed trigger partial compensatory changes, but these are neither complete nor functionally equivalent. Contrary to previous observations that the depletion of one H1 variant induces the upregulation of others [41, 42], the inactivation of H1.3 in AML does not increase the expression of replication-dependent H1 variants. In fact, we observe increased expression of the cell cycle-independent H1.0 variant, which is consistent with some observations reported in breast cancer cell lines [15]. However, despite H1.0 overexpression, its genomic distribution remained unchanged and whether H1.0 contributes functionally to chromatin maintenance in this context remains open. Even more striking is the redistribution of H1.2 when H1.3 is depleted. H1.2 moves from its preferred low-GC, H3K9me3-rich regions to areas typically occupied by H1.3, including those marked by H3K27me3. Notably, loci that gain H1.2 tend to correspond to genes whose expression is altered in H1.3 KOs, particularly repressed genes. This implies that H1.2 replacement may contribute to transcriptional deregulation. Therefore, changes in the occupancy of H1 chromatin variants may have direct transcriptional consequences, reinforcing the idea that H1 chromatin variants are not functionally interchangeable and that their regulatory effects depend on their binding preferences and interactions with chromatin features. However, regions that are highly occupied by H1.3, become occupied by H1.2 upon H1.3 KO, and are enriched in genes that are downregulated, either directly by H1.2 recruitment or indirectly.

The major transcriptional programs affected by H1.3 loss include interferon signaling, cell-cycle regulation, and transposable element (TE) control. Similar to other systems where H1 depletion activates immune pathways through derepression of satellites and TEs [17, 60], we find that H1.3 deficiency leads to activation of specific TE families via incomplete repression by H1.2 at loci formerly occupied by H1.3. While some TEs, particularly certain LTR elements, are derepressed and likely trigger a viral-mimicry-driven interferon response, LINE and SINE elements remain unaffected, indicating a selective H1.3 contribution to the repression of a distinct subset of TEs that H1.2 cannot fully compensate for. Furthermore, despite increased enrichment of H1.2 within this TE family, we observed no deregulation of SVA following the loss of H1.3. Since histone H1 variants have previously been shown to influence splicing [46], it is possible that the redistribution of H1.2 regulates TE splicing, producing non-functional TE isoforms and compensating for transcriptional derepression. This would prevent the overall expression of TEs. This also suggests that neither interferon response genes nor cell cycle signature genes need to be direct targets of H1 recruitment or absence, which makes it difficult to interpret the effects of H1 variant depletion. Further detailed studies are required to elucidate whether the absence or further recruitment of different H1 variants directly affects the expression or splicing of particular genes.

The loss of H1.3 and the altered distribution of H1.2 that results from this have implications for cell-cycle control. While H1.3 has been linked to oncogenic activity in pancreatic cancer and tumor suppression in ovarian cancer [61], our results indicate that, in AML, H1.3 promotes leukemic proliferation. At the cellular level, the loss of H1.3 led to a decrease in AML cell proliferation rates, and our molecular data point to a defect in the G1/S transition upon H1.3 inactivation. We observed the repression of DREAM and E2F complex target genes, which regulate this cell-cycle checkpoint [62, 63] as well as the upregulation of CDK inhibitors (CDKN1A, CDKN2A and CDKN2B). There have also been reductions in the reduced levels of CCNE1 and E2F1 proteins, and a decreased RB phosphorylation, which are consistent with earlier findings of cell-cycle repression upon H1.3 dysregulation [15], as well as reports implicating H1.2 in RB-mediated cell-cycle control [16, 64]. In addition, impaired replication initiation or replicative stress has been observed during depletion of other H1 protein variants [65, 66]. Recently, this replicative stress has been attributed to increased chromatin accessibility and altered topoisomerase II activity as the result of H1 protein [67]. This highlights the significance of H1 variants in preserving the chromatin architecture necessary for precise DNA replication and entry into the cell cycle.

In conclusion, our findings reveal the specific chromatin features of H1.3, and its interactions with other members of the H1 family. They also demonstrate the role of H1.3 in the cell cycle and its involvement in repressing transposable elements (TEs) in AML cells. Our work supports a model in which H1 variants play tissue- and context-specific roles in cancer, providing a basis for investigating H1.3 as a potential AML therapy. It is important to note that our studies were performed in NPM1mut AML cells, for which the interplay between NPM1mut and H1.3 remains unclear. Understanding how NPM1mut interacts with H1.3 will therefore be crucial in order to interpret the chromatin and transcriptional phenotypes observed, particularly in AML cases with low H1.3 expression.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

**CTQ: ** Investigation; Methodology; Formal analysis; Visualization; Writing—original draft; Writing—review and editing. **LNG: ** Investigation; Methodology; Formal analysis; Visualization; Writing—review and editing. **NSP: ** Investigation; Methodology; Formal analysis; Visualization; Writing—review and editing. **MR: ** Investigation; Validation, Visualization.** PHA** Formal analysis; Visualization**JV: ** Data curation, Formal analysis; Visualization. **AJ: ** Methodology; Formal analysis; Supervision; Funding acquisition; Writing—review and editing. **ED: ** Conceptualization; Supervision; Funding acquisition; Investigation; Methodology; Project administration; Writing—review and editing.All authors read and approved the final version of the manuscript.

Funding

This work was supported by the Ligue Nationale Contre le Cancer ; the Institut National de la Santé et de la Recherche Médicale (INSERM) ; the Centre National de la Recherche Scientifique (CNRS); the Canceropôle Provence Alpes Côte d’Azur, the Institute for Cancer and Immunology (Aix-Marseille University), by a grant from l’Institut National du Cancer (PRT-K16-071 to E.D.), which also funded LNG ; a European Union’s Horizon 2020 Research and Innovation Program (Marie Skłodowska-Curie grant # 813091 ARCH, age-related changes in hematopoiesis (that funded CT), by the Spanish Ministry of Science, Innovation and Universities and FEDER, EU (grant number PID2023-146239OB-I00/AEI/10.13039/501100011033).

We also acknowledge the Generalitat de Catalunya Suport Grups de Recerca AGAUR (grant number 2017-SGR-597).

Data availability

ATAC-seq datasets were deposited in the GEO database under accession number GSE302140,CHIP-seq datasets were deposited in the GEO database under accession number GSE302143,RNA-seq datasets were deposited in the GEO database under accession number GSE302144.

Declarations

Competing interests

The authors declare no competing interests.