Loss of H2A.X results in changes in chromatin condensation during maturation
To better understand the role of the variant histone H2A.X in erythropoiesis, we examined the erythroid compartment in bone marrow of H2A.X KO mice [24]. These mice exhibit macrocytic anemia with decreased numbers of reticulocytes, indicative of reduced production of RBCs; these phenotypic changes are accompanied by increased cell size (mean corpuscular volume, MCV) compared to WT mice (Additional file 2: Fig. S1a, b). Cell numbers for other hematopoietic lineages are not significantly affected (Additional file 2: Fig. S1a), indicating that H2A.X loss specifically impacts the erythroid lineage. Next, we quantified erythroblast intermediates in bone marrow (Additional file 2: Fig. S1c) [25] and found an increase in total Ter119+ cells compared to WT (Additional file 1: Fig. S1d), consistent with an increased erythropoietic drive in response to anemia. In addition, H2A.X KO mice have an increased proportion of more immature (basophilic) erythroblasts in the bone marrow and a corresponding decrease in the proportion of the most mature (orthochromatic) erythroblasts (Additional file 1: Fig. S1e), indicating potential defects in terminal erythroid maturation. Taken together, these data demonstrate that loss of H2A.X leads to dysregulated terminal maturation and an inability to maintain steady-state levels of erythrocyte production.
Terminal maturation of erythroid precursors is characterized by a substantial decrease in nuclear size prior to enucleation, associated with chromatin condensation. Increased MCV results in defects in nuclear maturation, as seen in megaloblastic anemias due to folate or B12 deficiencies [26]. Thus, the increased MCV of erythroblasts in H2A.X KO mice may reflect improper nuclear maturation. Consistent with this, we find that basophilic erythroblasts exhibit a higher nuclear:cytoplasmic (N:C) ratio compared to WT (Additional file 2: Fig. S1f), suggesting that nuclear maturation in H2A.X KO erythroid cells is impaired. Taken together, these data indicate that histone H2A.X plays a role in normal erythropoiesis, and may be required for normal nuclear maturation and chromatin condensation.
Prevalence of H2A.X and phosphorylated H2A.X in terminal erythroid maturation
Phosphorylation of histone H2A.X at S139 (γ-H2A.X) is an early signal for DNA repair pathways in response to DNA breaks. To determine if γ-H2A.X is present during erythropoiesis, we performed antibody staining on cells isolated from adult murine bone marrow and subjected the cells to imaging flow cytometry as previously described (Additional file: Fig. S2) [25, 27]. We observed γ-H2A.X during terminal erythroid maturation, with the most abundant signal in basophilic erythroblasts (Fig. 1A, B). Intriguingly, this stage of erythropoiesis corresponds with increased resistance to cell death and the start of chromatin condensation.
To further investigate H2A.X in erythropoiesis, we examined histone H2A.X protein expression in two human in vitro models of erythroid differentiation: primary human CD34+ progenitor cells and the HUDEP-2 erythroid cell line. Proliferating HUDEP-2 cells appear to model basophilic erythroblasts, both by morphology and by gene expression; under maturation culture conditions, the cells gradually exhibit morphologies characteristic of later stages of erythropoiesis. For the purposes of this study, we examined HUDEP-2 cells at day 4 and day 7 of maturation, corresponding to a transition of the majority of the culture from BasoE to PolyE (Fig. 1C) [28]. For the primary cell cultures, we utilized an established protocol involving synchronization by selection of CD36+ erythroid progenitors [29]. In these cultures, morphology observed at days 6, 8, and 10 after synchronization corresponds with BasoE, PolyE and OrthoE, respectively (Fig. 1D) [29].
Through immunoblotting, we found that H2A.X protein levels increase during terminal erythroid maturation in both maturation models (Fig. 1E, F). Furthermore, the immunoblots indicate that γ-H2A.X dynamics are substantially different. However, while HUDEP-2 cells exhibit consistent levels of γ-H2A.X across the time course, primary cells show initially high levels that decrease dramatically after the basophilic erythroblast stage. Together these data indicate that H2A.X and γ-H2A.X exhibit distinct dynamics during erythropoiesis and may play a role in nuclear maturation.
In addition to its well-documented role in DNA repair, γ-H2A.X has been implicated as a signal for apoptosis when accompanied by phosphorylation of the nearby C-terminal Y142 (H2A.X pY142) [22, 23]. Given the occurrence of γ-H2A.X during erythroid maturation, we determined if H2A.X pY142 was also present by immunoblotting (Fig. 1E, F). H2A.X pY142 was present in both HUDEP-2 and primary cell cultures at varying levels throughout maturation (Fig. 1E, F). Interestingly, H2A.X pY142 was only evident at D6 of maturation in the primary cultures, corresponding to the basophilic erythroblast stage of maturation (Fig. 1F). These data suggest that H2A.X may be dynamically phosphorylated at both S139 and Y142 during erythroid maturation.
Loss of H2A.X perturbs the expression of erythroid-specific genes
To determine the role of H2A.X in normal transcriptional dynamics associated with erythroid maturation, we generated H2A.X KOs in HUDEP-2 cells using a CRISPR/Cas9 protocol (Fig. 2A, B) [30, 31]. When grown in expansion media conditions, loss of H2A.X did not affect proliferation of HUDEP-2 cells compared to wild type (WT) (Fig. 2C). In contrast, in culture conditions that promote erythroid maturation, the loss of H2A.X resulted in cells that failed to condense their nuclei to the same degree as observed in WT cells (Fig. 2D). In addition, H2A.X KO cells did not accumulate hemoglobin during the first 4 days of maturation to the same extent as WT (Fig. 2E).
To further understand the effect of loss of H2A.X in erythroid cells, we performed RNA-Seq at D0 and D6 of maturation in WT and H2A.X KO HUDEP-2 subclones, and identified differentially expressed genes. At D0, 492 genes were upregulated and 77 genes downregulated in KO cells compared to WT, consistent with previous reports of a repressive role for H2A.X in the expression of subsets of genes [21]. By D6 of maturation, 480 genes were upregulated and 285 genes were downregulated (Fig. 3A–D, Additional file 1: Table S1). Gene ontology (GO) analysis on the DEGs revealed that “hemopoiesis” is populated in D0 DEGs upregulated in H2A.X KO erythroblasts, and pathways related to heme biosynthesis in D0 downregulated DEGs. By D6, however, no hematopoietic differentiation pathway was indicated by the analyses in either upregulated or downregulated DEGs in H2A.X KO erythroblasts (Additional file 2: Figure S3) [32,33,34,35,36].
Multiple erythroid-specific genes that are expressed at D0 and upregulated during maturation, such as β-globin (HBB), GYPA, ALAS2 and GATA1, are decreased at D0 in H2A.X KO cells, with GATA1 levels remaining lower at D6. GATA2 and STAT5b, which are typically downregulated during erythroid maturation, are significantly overexpressed in KO cells at D0 (Fig. 3E). We also observed that at D6 of maturation, the hematopoietic transcription factor TAL1 is overexpressed in KO cells. Additionally, STAT5b is overexpressed in KO cells at D6, while its expression is downregulated during normal erythroid maturation (Fig. 3E). These analyses indicate that H2A.X loss results in dysregulation of gene programs associated with late erythroid maturation, consistent with a loss of normal gene silencing programs and decreased activation of a number of late-expressed, erythroid-specific genes.
H2A.X signals for Caspase-Initiated Chromatin Condensation
Caspase-3, which is activated by protease cleavage during apoptosis, has been shown to be necessary for normal terminal erythroid maturation [3, 8, 9, 37]. Loss of Caspase-3 results in defective nuclear maturation and enucleation [9]. However, the late stages of erythroid maturation, during which Caspase-3 activation is observed, are associated with resistance to apoptosis [10, 11]. Notably, chromatin condensation prior to cell death is a hallmark of apoptosis [38]. Given our observation that both S139 and Y142 of H2A.X are phosphorylated during late erythroid maturation, we investigated the potential link between these modifications and activation of the caspase cascade, leading to CICC.
In addition to H2A.X pS139 and pS142, additional post-translational marks on histones are associated with apoptosis (Additional file 2: Table S2) [39,40,41,42,43]. However, it is unclear whether these other chromatin modifications are also present during terminal maturation. Histone H2B S14 phosphorylation (H2B pS14), is induced by multiple factors and is associated with apoptotic chromatin condensation, and has been shown to be downstream of Caspase 3-mediated cleavage of Acinus, which is also activated during erythropoiesis [9, 39, 40]. In addition, loss of H2B K15 acetylation (H2B K15Ac) is required for H2B S14 phosphorylation to occur [39, 42]. We observed that during HUDEP-2 and human CD34+ progenitor maturation, H2B pS14 is present (Fig. 4A, B), while levels of H2B K15Ac are low or undetectable (Fig. 4A, B). These data suggest that CICC is occurring during terminal erythroid maturation.
H4 K16 acetylation (H4 K16Ac) is a histone mark that has been associated with increased chromatin accessibility subsequent to DNA damage, which can contribute to apoptosis [39, 43]. We observed that H4 K16Ac levels are also very low during HUDEP2 and CD34+ maturation (Fig. 4A, B). This suggests that during maturation, erythroid cells are not undergoing DNA damage-related cell death despite high levels of γ-H2A.X.
To further validate these findings, we looked at the amount of cell death by Annexin V staining in maturing human CD34+ progenitor cell cultures (Fig. 4C). We observed a proportion of cells (~ 7%) that stain with Annexin V at D6 of maturation and are therefore likely apoptotic, with the proportion decreasing substantially (to 1–2%) at D8 and D10 (Fig. 4C). These dynamics do not match those of H4 K16Ac, which occurs at low levels that increase from D6-D10, or H2B pS14, where a substantial decrease is only seen between D8 and D10. These differing patterns suggest that the occurrence of these marks is unrelated to apoptosis in the CD34+ cultures.
H2A.X loss impairs the Caspase-Initiated Chromatin Condensation pathway
Next, we assessed if CICC requires normal H2A.X function. We observed that, as expected, H2A.X pS139 and H2A.X pY142 marks are eliminated in all 3 KO lines (Fig. 5A). Furthermore, we see that H2B pS14 levels are highly variable among KO subclones, in contrast to the consistent levels exhibited by WT subclones (Fig. 5B).
Caspase-3 has been shown to be necessary for normal terminal erythroid maturation [9, 44]. During HUDEP-2 maturation, full-length Caspase-3 expression decreases by D7, most likely due to cleavage and activation of the CICC response (Fig. 5B). However, full-length Caspase-3 expression did not decrease by D7 in H2A.X KO cells, and actually increased (Fig. 5B), suggesting a loss of Caspase-3 cleavage in KO cells. This indicates that in the H2A.X KO background, the CICC pathway is not properly activated.
BAZ1B mediates H2A.X Y142 phosphorylation and loss results in dysregulated genomic expression
To further investigate the potential role of Y142 phosphorylation in terminal erythroid maturation and CICC, we targeted the Y142 kinase in HUDEP-2 cells. Multiple kinases phosphorylate H2A.X S139, but Y142 is only known to be phosphorylated by BAZ1B/WSTF [45]. Of note, BAZ1B mRNA is expressed at high levels in human (Fig. 6A) and mouse (Fig. 6B) erythroid cells compared to other cell types [46]. We observed robust BAZ1B protein expression at D0 of HUDEP-2 maturation, which disappears by D7 (Fig. 6C).
To test the role of the BAZ1B kinase in erythroid maturation, we utilized CRISPR/Cas9-based methods to generate BAZ1B knockout HUDEP-2 subclones (Fig. 6D, E). BAZ1B KO cells appeared phenotypically normal at D0 of maturation, and proliferated normally (Fig. 6F, G). By D7 of maturation, the BAZ1B KO cells exhibited increased cell size with large and less condensed nuclei compared to wild type (Fig. 6F). These results indicate that loss of BAZ1B leads to defective chromatin condensation during terminal erythroid maturation, similar to that observed upon loss of H2A.X.
Next, we looked for changes in gene expression in WT and BAZ1B KO HUDEP-2 cells by RNA-seq. At D0 of maturation, 27 genes were upregulated while 80 genes were downregulated in KO cells compared to WT, while at D6 of maturation, 29 genes were upregulated and 117 genes were downregulated (Additional file 2: Fig. S4a–d, Additional file 1: Table S1). Next, we performed GO analysis to determine if any pathways were specifically perturbed by the loss of BAZ1B in erythroid cells [32,33,34,35,36]. Pathways bearing an obvious relation to erythroid differentiation or biology did not emerge from this analysis. However, genes upregulated in the BAZ1B KO cells suggest dysregulation of nonerythroid pathways both at D0 and D6 of maturation, including cardiac terms and those related to the spinal cord. In addition, the upregulated DEGs appear to be involved in multiple metabolomic pathways (Additional file 2: Fig. S5).
In contrast to the phenotype of the H2A.X KO, we found that erythroid and hematopoietic genes are not significantly affected by the loss of BAZ1B, and exhibit nearly equivalent expression levels to their WT counterparts (Fig. 6h). In general, the H2A.X and BAZ1B KOs exhibit few differentially expressed genes in common, although this may be due to the small number of genes affected in the latter. Genes exhibiting significant changes in steady-state expression levels in BAZ1B KO cells include transcription factors involved in embryonic development and development of other tissues (SOX4 and TBX20), proteins involved in neural development, and factors involved in formation and function of the blood–brain barrier (RELN and MFSD2A) (Additional file 2: Fig. S3e). In addition, factors associated with protection against oxidative stress (SNCA and SEPN1) and genes with metabolism (CYP27B1, RYR3 and ARG2) are dysregulated (Additional file 2: Fig. S4e). These data suggest that BAZ1B-KO HUDEP-2 cells maintain the erythroid gene expression program, but exhibit changes in a small number of genes not obviously related to erythroid biology. Furthermore, the data suggest that BAZ1B is involved in nuclear condensation in maturing erythroid cells, but that this effect is largely post-transcriptional, and thus overlaps only partially with the function of H2A.X.
Loss of BAZ1B impairs Caspase-Initiated Chromatin Condensation in erythroid cells
Next, we assessed how the loss of BAZ1B affected the post-translational marks that are associated with the CICC pathway. We observed that, as expected, H2A.X pY142 was lost in BAZ1B KO cells, but pS139 still occurred (Fig. 7A). Levels of H2B pS14 exhibit high variability between KO cell lines, but were generally decreased compared to WT, which in turn were characterized by highly consistent levels of this mark (Fig. 7A). We also observed that full-length Caspase-3 persists throughout maturation in the BAZ1B KO cells (Fig. 7B), indicating that Caspase-3 is not undergoing complete cleavage in the KO cells. These data suggest that BAZ1B phosphorylation of H2A.X Y142 is a necessary signal for the CICC pathway, and thus that loss of Y142 phosphorylation reproduces the nuclear condensation defect observed with the H2A.X KO, despite minimal effects on the transcriptome.