Deficiency and haploinsufficiency of histone macroH2A1.1 in mice recapitulate hematopoietic defects of human myelodysplastic syndrome

Background Epigenetic regulation is important in hematopoiesis, but the involvement of histone variants is poorly understood. Myelodysplastic syndromes (MDS) are heterogeneous clonal hematopoietic stem cell (HSC) disorders characterized by ineffective hematopoiesis. MacroH2A1.1 is a histone H2A variant that negatively correlates with the self-renewal capacity of embryonic, adult, and cancer stem cells. MacroH2A1.1 is a target of the frequent U2AF1 S34F mutation in MDS. The role of macroH2A1.1 in hematopoiesis is unclear. Results MacroH2A1.1 mRNA levels are significantly decreased in patients with low-risk MDS presenting with chromosomal 5q deletion and myeloid cytopenias and tend to be decreased in MDS patients carrying the U2AF1 S34F mutation. Using an innovative mouse allele lacking the macroH2A1.1 alternatively spliced exon, we investigated whether macroH2A1.1 regulates HSC homeostasis and differentiation. The lack of macroH2A1.1 decreased while macroH2A1.1 haploinsufficiency increased HSC frequency upon irradiation. Moreover, bone marrow transplantation experiments showed that both deficiency and haploinsufficiency of macroH2A1.1 resulted in enhanced HSC differentiation along the myeloid lineage. Finally, RNA-sequencing analysis implicated macroH2A1.1-mediated regulation of ribosomal gene expression in HSC homeostasis. Conclusions Together, our findings suggest a new epigenetic process contributing to hematopoiesis regulation. By combining clinical data with a discrete mutant mouse model and in vitro studies of human and mouse cells, we identify macroH2A1.1 as a key player in the cellular and molecular features of MDS. These data justify the exploration of macroH2A1.1 and associated proteins as therapeutic targets in hematological malignancies. Electronic supplementary material The online version of this article (10.1186/s13148-019-0724-z) contains supplementary material, which is available to authorized users.


Background
The healthy hematopoietic stem cell (HSC) pool is maintained by genetically and epigenetically regulated gene expression [1,2]. When these systems go awry, aberrant hematopoietic stem and progenitor cells may emerge (reviewed in [3]), which are implicated in the development of a heterogeneous group of disorders called myelodysplastic syndromes (MDS). MDS are characterized by ineffective hematopoiesis, leading to variable peripheral blood cytopenias and risk of transformation to acute myeloid leukemia (AML) [4]. MDS can remain stable for many years with few symptoms or can rapidly progress into a more aggressive MDS subtype [5]; current treatment options, however, are limited and often ineffective. Studies have started to identify common mutations underlying HSC dysregulation in MDS, but more progress is needed to identify the molecules and pathways contributing to disease mechanisms.
Here, we report for the first time that loss of macroH2A1.1 expression in mice led to the dysregulation of myeloid differentiation. RNA-sequencing identified defective production of ribosomal mRNAs as a potential mechanism for disrupted HSC homeostasis following macroH2A1.1 depletion. Thus, macroH2A1.1 delineates its function in HSC homeostasis and differentiation in mice. Moreover, we found that macroH2A1 isoforms' mRNA levels are significantly decreased in MDS patients with a 5q deletion compared to other MDS groups and healthy individuals.

Results
Mice lacking macroH2A1.1 exhibit blood abnormalities under a steady state As macroH2A1.1 expression is decreased in U2AF1 S34F MDS patients and its knockdown in vitro perturbs erythroid and granulomonocytic differentiation [6], we sought to investigate macroH2A1.1 function in mice using a gene KO approach. We first generated a mouse line carrying a conditional macroH2A1 allele that permits selective elimination of one of the two macroH2A protein isoforms using either Cre or Dre recombinase (Fig. 1a).
Blood counts and biochemical tests identified a low hematocrit and reduced neutrophil counts in macro-H2A1.1 fl/− HET and macroH2A1.1 −/− KO mice compared to macroH2A1.1 fl/fl and WT littermates; an opposite trend was observed for lymphocyte and eosinophil cell counts (Additional file 1: Figure S2A and B). No difference in cholesterol or triglyceride levels was observed between groups, but decreased liver transaminase levels were observed in macroH2A1.1 fl/− HET and macroH2A1.1 −/− KO mice (Additional file 1: Figure  S2C). The blood count abnormalities are reminiscent of those observed in human patients with MDS. We also detected lower alanine transaminase (ALT) and aspartate transaminase (AST) levels in macroH2A1.1 fl/− HET, and to a greater extent in macroH2A1.1 −/− KO mice (Additional file 1: Figure S2C), which in humans is associated with higher all-cause mortality [24,25].

MacroH2A1.1 haploinsufficiency or deficiency perturbs hematopoiesis following DNA damage
An expansion of phenotypically primitive HSCs is often observed in patients with MDS [3]. Murine HSCs are enriched in the bone marrow cell population that do not express mature hematopoietic cell lineage markers (Lin − ), such as B220, CD4, CD8, Gr-1, Mac-1, and Ter-119, but do express c-Kit and Sca-1, collectively termed LSK (Lin −-Sca-1 + c-Kit + ) cells [26]. HSC and early progenitors can be Histone H2A was used as a loading control. c macroH2A1.1 and macroH2A1.2 mRNA levels in adult mice tissues. RNA was extracted from the lung, liver, kidney, seminal gland, intestine, testis, brain, skeletal muscle, heart, and spleen of 3-month-old macroH2A1.1 fl/fl , macroH2A1.1 fl/− (HET), and macroH2A1.1 −/− (KO) mice, and analyzed by qRT-PCR using isoform-specific primers. N = 5 mice/group. **p < 0.01; ***p < 0.001 relative to macroH2A1.1 fl/fl further characterized within the LSK population based on the surface expression of SLAM (signaling lymphocyte activation molecule) markers CD150 and CD48 [27]. Given the enhanced susceptibility of macroH2A1.1 KO mice to radiation-induced death and cell damage, we tested the effects of haploinsufficiency and absence of the macroH2A1.1 on the hematopoietic compartment at a steady state and under DNA damaging conditions. At a steady state, total bone marrow cellularity (Fig. 2a) as well as the frequency of the LSK CD48 − CD150 + HSC did not vary significantly among macroH2A1.1 fl/fl , HET and KO mice (Fig. 2b). Because macroH2A1 isoforms regulate self-renewal and differentiation of induced pluripotent stem cells, embryonic stem cells and cancer stem cells [11,16,28], we hypothesized that macroH2A1.1 might be involved in regulating the hematopoietic compartment following genotoxic insult in a dose-dependent and time-dependent manner. We analyzed HSCs in the BM of all three macroH2A1.1 genotypes at 1 day (early time point) and 7 days (late time point) after systemic exposure to 600 (low dose) or 1200 rad (high dose) irradiation. Downregulation of c-Kit cell surface expression on functional HSCs has been reported in the situation of distress, such as 5-FU treatment and Myc gene deletion [29]. Since all Lin − Sca-1 + CD150 + cells in control animals are c-Kit + , the Lin − Sca-1 + CD150 + CD48 − marker combination can also be used to identify HSCs [29]. Since in irradiated mice, we also observed a dramatic reduction in c-Kit + cells (data not shown), we used the Lin − Sca- 1 + CD150 + CD48 − marker combination to identify HSC in WT, HET, and KO mice following different doses of irradiation in comparison to the same population in untreated mice. We observed that the macroH2A1.1 haploinsufficiency caused a significant increase in HSC frequency at late time point following low-dose (Fig. 2c) and high-dose ( Fig. 2d) irradiation compared to WT mice. Conversely, the macroH2A1.1 −/− mice showed a decrease in HSC frequency, caused by low (Fig. 2c, early time point) and high (Fig. 2d) doses of irradiation, compared to untreated and similarly treated WT mice. Altogether, these data demonstrate that the macroH2A1.1 isoform regulates HSC response to irradiation in a time-and dosedependent manner.

MacroH2A1.1 haploinsufficiency or deficiency in HSCs leads to a myeloid differentiation bias in transplantation assays
A myeloid differentiation bias is found in human patients with MDS [3]. To understand the downstream effects of macroH2A1.1 haploinsufficiency or deficiency on HSC/ hematopoietic progenitor cells (HPC) populations, we analyzed the BM from macroH2A1.1 fl/fl , macroH2A1.1 fl/− HET, and macroH2A1.1 −/− KO mice by flow cytometry. Higher frequencies of myeloid (Mac-1 + ) cells were found in macroH2A1.1 −/− KO mice compared to macroH2A1.1 fl/fl mice, and lower frequencies of B (B220 + ) cells were found in both macroH2A1.1 fl/− HET and in macroH2A1.1 fl/fl mice (Fig. 3a). To demonstrate that the cause of this myeloid bias observed in macroH2A1.1 −/− KO mice is intrinsic to the hematopoietic system, we performed an in vivo transplantation study. Here, flow cytometry analysis confirmed higher frequencies of myeloid cells and lower frequencies of B cells in the peripheral blood of mice receiving macroH2A1.1 fl/− HET or macroH2A1.1 −/− KO BM cells, compared to mice receiving macroH2A1.1 fl/fl BM cells ( Fig. 3b-d). The myeloid differentiation bias was also observed in HET mice in noncompetitive BM transplantation assay (Fig. 3c). Moreover, the analysis of the frequencies of myeloid progenitor sub-populations revealed a tendency increase in GMPs (Lin/IL7R/Sca-1 − c-Kit + CD34 + CD16/32 + ) and a tendency decrease in megakaryocyte-erythrocyte progenitors (MEPs; Lin − /IL7R/Sca-1 − c-Kit + CD34 − CD16/32 − ) in the BM of macroH2A1.1 fl/− HET and macroH2A1.1 −/− KO mice compared to macroH2A1.1 fl/fl mice (Fig. 3e). These data show that decreased macroH2A1.1 levels has a profound impact on HSC differentiation in the BM, resulting in a myeloid skewing similar to that observed in human MDS.

Downregulation of ribosomal protein genes after macroH2A1.1 depletion
To gain a mechanistic insight into the hematopoietic derangements associated with decreased macroH2A1.1 expression, we performed transcriptome-wide RNA sequencing (RNA-Seq) to identify differentially expressed genes (DEGs) between HPCs (CD150 − CD48 + LSK) isolated by cell sorting from macroH2A1.1 −/− KO and macroH2A1.1 fl/fl mice. Using a 1.5-fold change threshold, we identified 599 DEGs, of which 225 were upregulated and 374 were downregulated in macroH2A1.1 −/− compared to macroH2A1.1 fl/fl controls (Additional file 2: Table S1). KEGG analysis revealed that these genes significantly over-represented the ribosome pathway, protein processing in the endoplasmic reticulum, and the cysteine/methionine metabolism pathways (Fig. 4a). The ribosome pathway was the most significantly enriched pathway in our experiment (Fig. 4b), with 16 differentially expressed ribosomal proteins with their functional location within the large (Rpl) and small (Rps) subunits. Furthermore, we found the broad functions of chromatin modification/remodeling, transcription, redox cell metabolism, cytoskeleton homeostasis, and cellular response to DNA damage stimulus among the top 20 most represented biological processes (Gene Ontology, GO) in macroH2A1.1 −/− HPCs (Fig. 4c). As defective ribosome biogenesis has been reported in MDS [30], we next asked whether macroH2A1.1-deficiency-driven changes in ribosomal protein gene expression underpin pathology in hematopoietic cells. We stably knocked down (KD) macroH2A1.1 mRNA in human promyelocytic leukemia HL-60 and monocytic THP-1 cell lines grown in suspension [31], using lentivirally transduced shRNAs against the H2AFY gene. Upon efficient specific silencing of the macroH2A1.1 transcript (without altering the levels of macroH2A1.2 isoform) (Additional file 1: Figure S3A, B), we measured transcript levels for a subset of Rpl (19,29,38) and Rps (15a, 21) genes, whose expression displayed > twofold change in our RNA-Seq analysis (Additional file 2: Table S1). MacroH2A1.1 KD led to decreased Rpl19 (fivefold), Rpl29 (twofold), Rpl38 (fivefold), Rps15a (sixfold), and Rps21 (twofold) mRNA levels in both HL-60 and THP-1 cells (Additional file 1: Figure S3C, D), consistent with our RNA-Seq findings made in macroH2A1.1 −/− mouse HPCs (Fig. 4). In order to obtain evidence that macroH2A1.1 deficiency-dependent decreased ribosomal protein gene expression may lead to defective ribosome biogenesis and protein synthesis, we assessed the steady-state 47S pre-rRNA level in macroH2A1.1 −/− HPCs, and in human HL-60 and THP-1 cells KD for macroH2A1.1. Under these conditions, a~85% decrease in the relative amount of 47S pre-rRNA was observed in macroH2A1.1 −/− HPCs compared to macroH2A1.1 fl/fl HPCs (Fig. 5a). Similarly, ã 70% and~60% decrease in the relative amount of 47S pre-rRNA was observed in HL-1 and in THP-1 cells, respectively (Fig. 5b). In addition, we performed a protein synthesis inhibition assay using puromycin, an antibiotic that competes by acting as an analog of the three-terminal end of aminoacyl-tRNA, disrupting protein synthesis. To this purpose, we performed a [3H]-leucine incorporation assay upon treatment with puromycin (0, 0.5, 2 μg/ml) for 72 h in HL-60 and THP-1 cells; our findings show that protein synthesis was significantly inhibited in HL-1 and THP1 cells even in the absence of puromycin, and it was further inhibited upon treatment with 0.5 μg/ml puromycin (Fig. 5c). Two micrograms per milliliter puromycin was lethal in both cell lines (Fig. 5c).

MacroH2A1 isoforms' transcript levels are decreased in the BM of MDS patients carrying a 5q deletion or U2AF1 S34F mutation
We first compared the BM expression levels of macroH2A1.1 or macroH2A1.2 between sub-groups of MDS patients with different genetic abnormalities and healthy controls (n = 5). MDS patients (n = 24 total) were categorized as either having a normal karyotype (NK; n = 4); a 5q deletion (del(5q); n = 15), of which seven also additional cytogenetic changes; or a non-del5q abnormal karyotype (AK; n = 5) (Additional file 3: Table S2). As expected, del(5q) patients expressed significantly lower macroH2A1.1 and macroH2A1.2 mRNA levels compared to healthy controls (p < 0.05), due to the loss of H2AFY located at 5q31.1 within the commonly-deleted 5q region (Fig. 8a, b). By contrast, NK and AK MDS patients showed normal macroH2A1.1/macroH2A1.2 mRNA levels ( Fig. 8a, b). Aberrant H2AFY splicing causing specifically reduced macroH2A1.1, but not macroH2A1.2, expression has been associated with the U2AF1 S34F mutation in MDS [6]. However, only three MDS patients in this cohort carried this mutation, and they were not del(5q) carriers   Table S2). When comparing these U2AF1 mutants versus healthy subjects, we detected a decreased, albeit not significant, macroH2A1.1 but not macroH2A1.2 mRNA expression (Fig. 8c, d), according to previous studies [6]. These data provide evidence that macroH2A1.1 mRNA tends to decrease in the BM of MDS patients carrying the U2AF1 S34F mutation and that both macroH2A1.1 isoforms' transcript levels are decreased in the BM of MDS patients who are del(5q) carriers.

Conclusions
Here, we show that the histone variant macroH2A1.1, which is downregulated in a subset of patients with MDS, profoundly affects the survival and differentiation of murine HSCs/HPCs in vivo and human leukemic cell lines in vitro. We provide evidence that macroH2A1.1 is required for normal ribosomal protein gene expression in HPCs. Without proper expression of this histone variant, abundant cell death occurred in the hematopoietic cells. Uncovering the relationship between macroH2A1.1 depletion and defective ribosomal biogenesis in HSCs/ HPCs provides a critical link between this epigenetic regulator and the molecular pathologies typical of MDS.
Our findings are in line with a recent report from Yip et al. [6]. These authors analyzed the role of the mutation S34F in the splicing factor U2AF1, which is frequent in MDS: U2AF1 S34F altered mRNA splicing of many transcripts and among those H2AFY (coding for macroH2A1.1 and macroH2A1.2). Only 3 out of 29 patients in our cohort displayed U2AF1 S34F mutation; in those individuals, we observed a not statistically significant tendency toward a decrease in macroH2A1.1, but not macroH2A1.2 mRNA level: analyses on larger cohorts are required to corroborate these data [6].
H2AFY is physically localized on the long arm of chromosome 5, implying a connection with the most common karyotypic abnormality in MDS-del(5q). As expected, both macroH2A1.1 and macroH2A1.2 mRNA levels were found decreased in the BM of del(5q) MDS patients, compared to healthy subjects.
Various transcriptional and epigenetic mechanisms contribute to HSC maintenance and stepwise differentiation to produce distinct hematopoietic lineages [42,43], and macroH2A1.1 has only recently emerged as a novel epigenetic regulator of hematopoiesis. Silencing macroH2A1.1 expression in human progenitor cells alters erythroid and granulomonocytic differentiation, and the reintroduction of macroH2A1.1 in U2AF1 mutant BM cells rescues cell death and their differentiation potential [6]. Consistently, we found that in lineage-committed HL-60 (neutrophils) and THP-1 (monocytes) human myelomonocytic leukemia cell lines, KD for macroH2A1.1 led to massive reduction in proliferation and death, and that macroH2A1.1 haploinsufficiency, and to a lesser extent, its full deficiency, tends to decrease the frequency of CMPs and MEPs, and to increase GMPs in mice. Expanded GMPs characterize high-risk MDS [44]. Compared to in vitro shRNA-mediated silencing or full in vivo KO, we believe that the features of our newly generated macroH2A1.1 haploinsufficient mouse might be more akin to the pathological changes observed in MDS patients. In fact, we also detected a myeloid bias in mice that lack the full dose of macroH2A1.1: the hematopoietic cell-autonomous enrichment of myeloid Mac-1+ cells parallels the myeloid differentiation bias that is commonly observed in other MDS murine models and in human MDS. MDS is characterized on one hand by dysplastic myeloid expansion, with myeloid cells that have reduced ability to differentiate, and on the other hand by neutropenia, thrombocytopenia, and anemia [45][46][47]: we propose macroH2A1.1 −/− mouse as a useful tool for further mechanistic studies on MDS. By characterizing the phenotype of macroH2A1.1-insufficient mice, we report also another primary in vivo role for macroH2A1.1 in hematopoiesis: our RNA-Seq analysis of HPCs from macroH2A1.1 −/− mice identified a substantial depletion of many transcripts [small ribosomal proteins (Rps) and large ribosomal proteins (Rpl)] involved in ribosome assembly, and this was confirmed in HL-60 and THP-1 myeloid cell lines. Moreover, we found that macroH2A1.1 −/− HPCs and HL-60/THP-1 KD for macroH2A1.1 display decreased levels of 47S pre-rRNA, a long primary transcript that is the precursor of three of the four ribosomal RNAs: 18S, 5.8S, and 28S. Within ribosome structure, the 18S rRNA assembles with 33 Rps to form the 40S ribosomal subunit or small subunit, while the 5S, 5.8S, and 28S rRNAs associate with 47 Rpl to assemble the 60S or large subunit. 47S pre-rRNA and Rps/ Rpl levels might thus be connected events leading to defective ribosome biogenesis in macroH2A1.1 −/− HPCs. Although the causal role for ribosomal biogenesis in HSC maintenance is not fully understood, impaired ribosome biogenesis-induced nuclear stress, for instance, due to hemizygosity for genes encoding ribosomal proteins, is associated with the development of clinical entities collectively known as "ribosomopathies" which include several bone marrow failure syndromes, including MDS [48]. We speculate that macroH2A1.1 depletion in the hematopoietic system might have two deleterious effects. First, it alters HSC homeostasis through defective ribosomal production [30]: adult HSCs display low levels of global protein synthesis relative to HPCs and genetic perturbations that alter the dynamics of protein synthesis impair HSC function [49]. Our data suggesting that macroH2A1.1-dependent impairment in ribosome biogenesis relates to impaired HSC differentiation are consistent with Signer et al., who demonstrated that reduced ribosome function in Rpl24 mice reduced protein synthesis of 30% in HSCs and impaired HSC function [50].
Second, given the extensive Rpl and Rps expression heterogeneity in the hematopoietic system [51], macroH2A1.1 −/− might disrupt the regulation of hematopoietic lineage-specific ribosomal proteins that might be involved in lineage differentiation. Predicting the phenotypes of perturbed Rpl and Rps expression patterns remains challenging and is an area that warrants future research.
In summary, our study shows that a loss of macroH2A1.1, which affects a subset of MDS patients, has a critical role in the defective hematopoiesis and perturbed ribosomal biogenesis that are central to MDS pathology.
By combining clinical data with a discrete mutant mouse model and in vitro studies of human cells, we identify macroH2A1.1 as a key determinant of the cellular and molecular features of MDS. These data justify the exploration of macroH2A1.1 and associated proteins as therapeutic targets in hematological malignancies.

Clinical and laboratory data of MDS patients
Bone marrow aspirates, clinical information, and routine laboratory data were collected from 24 patients diagnosed with MDS according to the revised World Health Organization criteria [52] and prior to commencement of treatment, at the University Hospital Brno. Basic MDS patients and cytogenetic characteristics are shown in Additional file 3: Table S2. U2AF1 mutation analyses were performed as described previously [6].

Human bone marrow cell isolation and sample processing
Red blood cells were depleted using ACK lysing buffer (NH 4 Cl 150 mM, KHCO 3 , 10 mM, Na 2 EDTA 0.1 mM, pH 7.2). White blood cells were further processed for RNA isolation. Total RNA was isolated using TriReagent (MRC, USA) according to the manufacturer's instructions and the quality of RNA was assayed with an Agilent 2100 Bioanalyser (RNA 6000 Nano Assay; Agilent Technologies, USA).

Animal models
Mice lacking macroH2A1.1 were generated as follows: a 12 kb segment of the murine H2AFY gene (introns 5-8) was subcloned from a BAC by recombineering into p15A-HSV tk-DTA-amp. A lacZ-neo cassette [53], flanked by loxP and rox sites at the 5′ and 3′ ends, respectively, was inserted into the intron between exons 6a and 6b, also by recombineering [54]. Another rox site was inserted upstream of exon 6a and another loxP site inserted downstream of exon 6b so that Dre/rox recombination [55] would remove exon 6a and the lacZ-neo cassette, and Cre/loxP recombination would remove exon 6b and the lacZ-neo cassette; thus, Cre recombination will eliminate macroH2A1.1 expression. Southern blotting of genomic NheI-digested DNA from individual ES-cell-derived clones with a 3′ probe was used to identify homologous recombinants (Additional file 1: Figure S1). A 12.3-kb DNA fragment corresponds to the wild-type macroH2A1.1 locus; integration of the loxP-flanked neomycin cassette 3′ of exon 6b introduced an additional NheI site, thus increasing the size of the NheI DNA fragment to 16.2 kb in the targeted allele (Additional file 1: Figure S1). Cre-mediated recombination resulted in a 3.9kb NheI DNA fragment recognized by the 3′ probe, which is diagnostic of the macroH2A1.1 allele. The targeting of the macroH2A1.1 allele was performed by electroporation of A9 ES cells, which were then injected into C57BL/6 eight cell-stage embryos. The targeted macroH2A1 fl/fl mice were crossed to deleter HPRT-Cre mice (129S1/Sv-Hprt tm1(CAG-cre)Mnn /J), purchased from Jackson Laboratories, USA, to remove the loxP-flanked neomycin cassette and generate macroH2A1.1 fl/− mice (heterozygous, HET), respectively. Mice heterozygous for the macroH2A1.1 allele were further crossed to deleter Cre mice to generate the macroH2A1.1 −/− (knockout, KO) mice, respectively. All mice used were obtained after eight generations of back crossing on a C57Bl/6 genetic background. Mice were bred and maintained at the EMBL Mouse Biology Unit, Monterotondo, or at Plaisant Srl (Rome, Italy), in accordance with current Italian legislation (article 9, 27 January 1992, number 116) under a license from the Italian Health Ministry. The congenic C57BL/6-Ly5.1 mice were purchased from Charles River Laboratory. C57BL/6-Ly5.1/2 recipients were generated by intercrossing C57BL/6-Ly5.1 and C57Bl/6-Ly5.2 mice (Harlan, Italy).

Statistical analyses
Data are shown as means ± standard error of the mean (SEM). Groups were compared with either Student's t test or the non-parametric Mann-Whitney U test, as appropriate, using GraphPad Prism Software (version 5.00 for Windows, San Diego, CA, USA): significance was p ≤ 0.05. Survival analyses of mice employed the Kaplan-Meier estimator.