Epigenetic Determinants of Erythropoiesis: Role of the Histone Methyltransferase SetD8 in Promoting Erythroid Cell Maturation and Survival

Abstract

Erythropoiesis, in which committed progenitor cells generate millions of erythrocytes daily, involves dramatic changes in the chromatin structure and transcriptome of erythroblasts, prior to their enucleation. While the involvement of the master-regulatory transcription factors GATA binding protein 1 (GATA-1) and GATA-2 in this process is established, the mechanistic contributions of many chromatin-modifying/remodeling enzymes in red cell biology remain enigmatic. We demonstrated that SetD8, a histone methyltransferase that catalyzes monomethylation of histone H4 at lysine 20 (H4K20me1), is a context-dependent GATA-1 corepressor in erythroid cells. To determine whether SetD8 controls erythroid maturation and/or function, we used a small hairpin RNA (shRNA)-based loss-of-function strategy in a primary murine erythroblast culture system. In this system, SetD8 promoted erythroblast maturation and survival, and this did not involve upregulation of the established regulator of erythroblast survival Bcl-x_L. SetD8 catalyzed H4K20me1 at a critical Gata2 cis element and restricted occupancy by an enhancer of Gata2 transcription, Scl/TAL1, thereby repressing Gata2 transcription. Elevating GATA-2 levels in erythroid precursors yielded a maturation block comparable to that induced by SetD8 downregulation. As lowering GATA-2 expression in the context of SetD8 knockdown did not rescue erythroid maturation, we propose that SetD8 regulation of erythroid maturation involves multiple target genes. These results establish SetD8 as a determinant of erythroid cell maturation and provide a framework for understanding how a broadly expressed histone-modifying enzyme mediates cell-type-specific GATA factor function.

INTRODUCTION

The capacity of stem and progenitor cells to generate multiple cell lineages is orchestrated by cell-type-specific transcription factors that instigate lineage-specific genetic networks. These factors function with a cadre of broadly expressed transcription factors and coregulators, including chromatin-remodeling and -modifying enzymes. Cell-type-specific factors endow broadly expressed factors with activities important for establishing and/or maintaining the specialized transcriptome. Despite this paradigm, the functions of many broadly expressed chromatin-remodeling and -modifying enzymes have not been investigated in cell type-specific contexts. Considering the feasibility of devising small-molecule strategies to target enzymes, it is instructive to identify enzymatic components mediating important biological processes. We have been addressing this problem by asking how GATA factors with specialized expression patterns and functions utilize broadly expressed coregulators to mediate cellular transitions required for development of hematopoietic stem cells (HSCs), progenitors, and differentiated progeny, including the erythrocyte.

The family of dual zinc finger GATA transcription factors (1) recognize DNA with a WGATAR consensus (2, 3). GATA-2 is expressed predominantly in hematopoietic stem/progenitor cells (HSPCs), mast cells, endothelial cells, and neurons (4–8). Through its actions to induce HSC generation (9, 10) and to regulate HSPC function (11–13), GATA-2 mediates multilineage hematopoiesis. Mutations that alter the GATA2 coding region (14–16) or an essential cis element 9.5 kb downstream of the Gata2 1S promoter (+9.5 site) (17, 18) cause a primary immunodeficiency syndrome (MonoMAC) commonly associated with myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML). The +9.5 site enhances Gata2 transcription and induces HSC generation from hemogenic endothelium in the aorta gonad mesonephros (AGM) region of the developing embryo (9). LIM domain binding protein 1 (LDB1) and the chromatin remodeler Brahma related gene 1 (BRG1) confer activation through the +9.5 site (19). GATA-2 occupancy at this site in the transcriptionally active human and murine Gata2 loci suggests positive autoregulation (20–22).

GATA-1 is expressed predominantly in erythroid cells, megakaryocytes, mast cells, and eosinophils (6, 23–25) and is essential for controlling the development of these cells (26–29). GATA-1 utilizes its cofactor Friend of GATA-1 (FOG-1) to activate and repress most target genes, including Gata2 (30, 31). Some GATA-1 target genes have little or no FOG-1 requirement for regulation (31, 32). Since GATA-2 is expressed in multipotent hematopoietic precursors, its chromatin occupancy commonly precedes that of GATA-1. As GATA-1 levels rise during erythropoiesis, GATA-1 displaces GATA-2 from chromatin sites (29). These “GATA switches” occur at numerous sites in the genome, including 5 sites at the Gata2 locus, and are often associated with altered transcriptional output (21, 33–36). GATA-1/FOG-1 recruit the histone acetyltransferase CBP/P300 (37) and the nucleosome-remodeling and deacetylase (NuRD) complex (38–40), and we demonstrated that the chromatin-modifying enzyme SetD8 (PR-Set7) is a context-dependent GATA-1 corepressor at select GATA-1 target genes (41).

SetD8 is the sole enzyme known to monomethylate histone H4 at lysine 20 (H4K20me1) (42). Targeted disruption of murine Setd8 is embryonic lethal between the 4- and 8-cell stages (43). SetD8 levels are regulated during the cell cycle, and its degradation is required for cell cycle progression (44, 45). While the precise biochemical consequences of H4K20me1 are not established, this histone mark has been reported to correlate with activation and repression. H4K20me1 localizes to inactive heterochromatic regions of Drosophila polytene chromosomes (42). H4K20me1 can promote chromatin compaction directly, as well as through subsequent di- and trimethylation (43, 46). Loss of H4K20me1 from H4K20me1-encriched genes increases transcription (47). In support of SetD8 and H4K20me1 involvement in transcriptional activation, the genomic H4K20me1 profile in human T lymphocytes and CD36⁺ erythroid precursor cells correlates with transcriptional activity (48–50). We analyzed endogenous SetD8 function in a genetic complementation assay in GATA-1-null erythroid precursor cells (G1E-ER-GATA-1) (41). In this system, ER-GATA-1 induces a physiologically relevant window of erythroid maturation over a 2-day time course (51, 52). The G1E-ER-GATA-1 studies provided evidence that SetD8 confers repression of a subset of GATA-1-repressed target genes, and SetD8 almost exclusively mediates repression (41). We also demonstrated that H4K20me1 levels increased across a broad region of repressed, but not activated, loci, which provided evidence for H4K20me1 as a repressive chromatin mark in erythroid cells (41). These results led to the hypothesis that SetD8 controls erythroid cell maturation or function physiologically. Using an ex vivo murine fetal liver erythroblast culture system, we establish SetD8 as a positive regulator of primary erythroid cell maturation and survival and developed mechanistic insights of broader relevance to SetD8 function in diverse contexts.

MATERIALS AND METHODS

Cell culture.

Primary fetal liver erythroid progenitor cells were maintained at a density of 2.5 × 10⁵ cells/ml in StemPro-34 (Gibco) supplemented with 10% nutrient supplement (Gibco), 2 mM l-glutamine (Cellgro), 1% penicillin-streptomycin (Cellgro), 100 μM monothioglycerol (Sigma), 1 μM dexamethasone (Sigma), 0.5 U/ml of erythropoietin, and 1% conditioned medium from a Kit ligand-producing CHO cell line for expansion. For differentiation, cells were cultured at a density of 1 × 10⁶ cells/ml in embryonic stem cell Iscove's modified Dulbecco medium (ES IMDM; glutamine-free; HyClone) containing 10% fetal bovine serum (FBS; Gemini), 10% plasma-derived serum (PDS; Animal Technologies), 5% protein-free hybridoma medium II (PFHM II; Gibco), 2 mM l-glutamine (Cellgro), 1% penicillin-streptomycin (Cellgro), 100 μM monothioglycerol (Sigma), and 5 U/ml of erythropoietin. Cells were grown in a humidified incubator at 37°C with 5% carbon dioxide. All percentages are vol/vol unless otherwise noted. G1E-ER-GATA-1 cells were cultured in IMDM (Gibco) containing 15% FBS (Gemini), 1% penicillin-streptomycin (Gemini), 2 U/ml erythropoietin, 120 nM monothioglycerol (Sigma), 0.6% conditioned medium from a Kit ligand-producing CHO cell line, and 1 μg/ml puromycin (Gemini). Estrogen receptor (ER)–GATA-1 activity was induced by treating cells with 1 μM β-estradiol (Steraloids, Inc.).

Quantitative real-time RT-PCR.

Total RNA was purified from TRIzol per manufacturer instructions (Invitrogen). cDNA was prepared by annealing 300 ng of RNA with 250 ng of a 1:5 mixture of random hexamer and oligo(dT) primers heated at 68°C for 10 min. This was followed by incubation with murine Moloney leukemia virus reverse transcriptase (RT; Invitrogen) combined with 10 mM dithiothreitol (DTT), 20 U RNasin (Promega), and 0.5 mM deoxynucleoside triphosphates (dNTPs) at 42°C for 1 h. The mixture was heat inactivated at 95°C for 5 min and diluted to a final volume of 50 μl. RT-PCR mixtures (20 μl) contained 1 μl of cDNA, appropriate primers, and 10 μl of Power SYBR green master mix (Applied Biosystems). Product accumulation was monitored by SYBR green fluorescence using either a StepOnePlus or a Viia7 instrument (Applied Biosystems). A standard curve of serial dilutions of cDNA samples was used to determine relative expression. mRNA levels were normalized to 18S rRNA.

ChIP assay.

Quantitative chromatin immunoprecipitation (ChIP) was conducted as described previously (21, 53) using antibodies specific for monomethylated H4K20 (Millipore), GATA-1 (54), Scl/TAL1 (55), LDB1 (N18, sc11198; Santa Cruz Biotechnology), and acetylated H4 (Upstate). Samples were analyzed by quantitative real-time PCR using either a StepOnePlus or a Viia7 instrument (Applied Biosystems), and product was quantitated by SYBR green fluorescence. The amount of product was determined relative to a standard curve generated from a serial dilution of input chromatin. Dissociation curves revealed that primer pairs generated single products.

Primary fetal liver erythroid progenitor cell isolation.

Primary erythroid precursor cells were isolated from embryonic day 14.5 (E14.5) fetal livers using the EasySep negative-selection mouse hematopoietic progenitor cell enrichment kit (StemCell Technologies). Briefly, fetal liver cells were resuspended at a concentration of 5 × 10⁷ cells/ml in phosphate-buffered saline (PBS) containing 2% FBS, 2.5 mM EDTA, and 10 mM glucose. The EasySep mouse hematopoietic progenitor cell enrichment cocktail was added at 50 μl/ml, supplemented with 2.5 μg/ml biotin-conjugated CD71 antibody (eBioscience) for removal of proerythroblasts. After 15 min of incubation on ice, cells were washed once by centrifugation for 5 min at 1,200 rpm at 4°C. Cells were resuspended at a concentration of 5 × 10⁷ cells/ml in PBS containing 2% FBS, 2.5 mM EDTA, and 10 mM glucose, and EasySep biotin selection cocktail was added at 100 μl/ml. After 15 min at 4°C, EasySep mouse progenitor magnetic microparticles were added at 60 μl/ml. After 10 min at 4°C, cells were resuspended to a total volume of 2.5 ml and incubated with a magnet for 3 min. Unbound progenitor cells were carefully transferred into a 15-ml tube and used for subsequent experiments.

Retroviral infection.

The microRNA 30 (miR-30) context Setd8 and Gata2 shRNAs were cloned into MSCV-PIG vector, kindly provided by Mitchell Weiss, using BglII and XhoI restriction sites. Retroviruses expressing shRNA targeting luciferase (control), SetD8, and Gata2, containing a Gata2 cDNA (56) or an empty vector, were produced by transfecting 3 × 10⁶ 293T cells with 15 μg of both MSCV-PIG vector and pCL-ECO viral packaging vector. Retrovirus-containing supernatant was harvested 24 or 48 h posttransfection. Primary erythroid precursor cells were spinfected with 100 μl of retrovirus supernatant and 8 μg/ml Polybrene in 400 μl of fetal liver expansion medium at 1,200 × g for 90 min at 30°C. After centrifugation, 500 μl prewarmed fetal liver expansion medium was added, and cells were incubated at 37°C overnight.

The miR-30 context SetD8 shRNA sequence is TGCTGTTGACAGTGAGCGCAAGCACTGTTCTCCTGCTCAATAGTGAAGCCACAGATGTATTGAGCAGGAGAACAGTGCTTTTGCCTACTGCCTCGGA; the miR 30 context Gata2 shRNA sequence is TGCTGTTGACAGTGAGCGCAAGGAGTAGGCAAGAAGAAAATAGTGAAGCCACAGATGTATTTTCTTCTTGCCTACTCCTTTTGCCTACTGCCTCGGA.

Erythroid maturation assay and fluorescence-activated cell sorting (FACS).

Cells were washed with PBS once, and 1 × 10⁷ cells were stained with 0.8 μg of anti-mouse Ter119–APC and anti-mouse CD71–PE (eBioscience) at 4°C for 30 min in the dark. After staining, cells were washed three times with 2% bovine serum albumin (BSA) in PBS. Samples were analyzed using a FACSAria II cell sorter (BD Bioscience). Cells were gated on green fluorescent protein (GFP) to ensure retroviral expression. DAPI (4′,6-diamidino-2-phenylindole) exclusion was utilized for live/dead discrimination. Cells in the R1, R2, R3, and R4/5 populations were sorted into 5-ml round-bottom tubes and immediately processed for RNA isolation and cytospin or were cross-linked for subsequent ChIP experiments.

Apoptosis analysis.

One million cells were washed with PBS, followed by a wash with 1× annexin binding buffer (Life Technologies). Cells were resuspended in 100 μl annexin binding buffer and labeled with 5 μl annexin V-Alexa Fluor 350 conjugate (Life Technologies) and propidium iodide (PI). Samples were analyzed using an LSRFortessa or LSRII cytometer (BD Biosciences). As an additional method of analyzing apoptosis, fetal liver cells were stained for active caspase-3 and cleaved poly(ADP-ribose) polymerase (PARP). Two million cells were washed with PBS, fixed in 1% paraformaldehyde, and stored overnight in 70% ethanol at −20°C. Following permeabilization in 0.25% Triton X-100 (Sigma), cells were stained with anti-active caspase-3 phycoerythrin (PE)-conjugated antibody (BD Pharmingen number 51-68655X) and anti-cleaved PARP allophycocyanin (APC)-conjugated antibody (BD Pharmingen clone F21-852). Fluorescence was monitored using a FACSAria or LSRII cytometer (BD Biosciences).

Proliferation analysis.

Cell proliferation was measured using the CellTrace Violet cell proliferation kit per manufacturer instructions (Life Technologies). One million cells were labeled with 10 μM CellTrace Violet dye for 30 min, washed with prewarmed media, and cultured for 24 h. After 24 h, CellTrace Violet fluorescence intensity was measured using an LSRII cytometer (BD Bioscience). To determine the fluorescence intensity of the first cellular generation, a control sample was labeled with 10 μM CellTrace Violet 30 min prior to sample analysis. The percentage of cells in each daughter generation was determined using ModFit LT software (Verity Software House).

Protein analysis.

Protein samples were isolated by centrifugation of 1 × 10⁶ cells from each condition, washed with cold PBS, and lysed in 1× SDS sample buffer (25 mM Tris, pH 6.8, 2% β-mercaptoethanol, 3% SDS, 0.005% bromophenol blue, 5% glycerol). Samples were boiled for 15 min and stored at −80°C. Samples were resolved by SDS-PAGE, and proteins were detected by semiquantitative Western blotting with ECL 2 Western blotting substrate (Thermo Scientific). The antibodies used were anti-SetD8 (07-316; Millipore), anti-α-tubulin (clone DM1A, 05-829; Millipore), and anti-β-actin (3700S; Cell Signaling Technology). Secondary antibodies included goat anti-mouse IgG–horseradish peroxidase (HRP) and goat anti-rabbit IgG–HRP (sc-2005 and sc-2004; Santa Cruz Biotechnology).

Statistical analysis.

Student's t tests were conducted using GraphPad software or Microsoft Excel.

Primers.

The following primers were used for quantitative RT-PCR (5′ to 3′): 18S rRNA, CGCCGCTAGAGGTGAAATTCT and CGAACCTCCGACTTTCGTTCT; SetD8 mRNA, TAGCTGGAATCTACAGGAAGCGA and GTGTCTTTGCTCTTCTGTTCATCTG; Bcl2l1 mRNA, GACAAGGAGATGCAGGTATTGG and TCCCGTAGAGATCCACAAAAGT; Bax mRNA, TGAAGACAGGGGCCTTTTTG and AATTCGCCGGAGACACTCG; Tal1 mRNA, CGAGCGCTGCTCTATAGCCTT and TCACCCGGTTGTTGTTGGT; Vim mRNA, CGAAAGCACCCTGCAGTCAT and AAGGTCAAGACGTGCCAGAGA; Gata2 mRNA (1), GCAGAGAAGCAAGGCTCGC and CAGTTGACACACTCCCGGC; Gata2 mRNA (2), GGCTCTACCACAAGATGAATGGA and AGGTGGTGGTTGTCGTCTGAC; Gata1 mRNA, GGCCCAAGAAGCGAATGATT and GGTTCACCTGATGGAGCTTGA; Fog1 mRNA, CCTTGCTACCGCAGTCATCA and ACCAGATCCCGCAGTCTTTG; Klf1 mRNA, CACGCACACGGGAGAGAAG and CGTCAGTTCGTCTGAGCGAG; Pu.1 mRNA, GGCAGCGATGGAGAAAGC and GGACATGGTGTGCGGAGAA; Kit mRNA, AGCAATGGCCTCACGAGTTCTA and CCAGGAAAAGTTTGGCAGGAT; Lyl1 mRNA, AAGCGCAGACCAAGCCATAG and AGCGCTCACGGCTGTTG; MyoD promoter, GGGTAGAGGACAGCCGGTGT and GTACAATGACAAAGGTTCTGTGGGT; Eif3k promoter, GTGATTTCCTTCCAGCAGTTGTAA and CTCACGCTATTGGTCTCTTTTAAGTG; Gata2 +9.5 Site −933 bp, CTTGCTGCTGGCTCTGAGAAC and AGTCCAGGGTCTTTTAAGGATAAATTC; Gata2 +9.5 Site −480 bp, AACCTTCAAATGCAGACACTTCAC and GAATCCGCCAGAACGAAGAC; Gata2 +9.5 Site, GACATCTGCAGCCGGTAGATAAG and CATTATTTGCAGAGTGGAGGGTATTAG; Gata2 +9.5 Site +446 bp, GCCGAGGGAGTTCAGTGCTA and AGCGCTACTCCTGTGTGTTCTTC; Gata2 +9.5 Site + 880 bp, TCCTGGCGACTCCTAGATCCTA and GAAAGCCCTGAGGAAGTTGGA; Gata2 −77 Site, CTTTACCACATCAGGATACAGAGCA and CACCGCACAGCAGTGATAGATAGT; Gata2 −22 Site, GCTTTATCAGGCCACAGCTTG and GCACAGTCCTGGCAAAGTTCTC; Gata2 −3.9 Site, GAGATGAGCTAATCCCGCTGTA and AAGGCTGTATTTTTCCAGGCC; Gata2 −2.8 Site, GCCCTGTACAACCCCATTCTC and TTGTTCCCGGCGAAGATAAT; Gata2 −1.8 Site, GCATGGCCCTGGTAATAGCA and CAGCCGCACCTTCCCTAA; Gata2 1S Promoter, CCCCTCGAAGTGATGTCGAA and TCTGGCTGTCTCTCGGTTCC; Gata2 1G Promoter, AGATACCCAGAAGGTGCACGTC and GCAGACCCTGCACCCCT; β-Globin HS2, AGGGTGTGTGGCCAGATGTT and ACCCAGATAGCACTGATCACTCAC; Lyl1 Exon 1, TCAGCATTGCTTCTTATCAGCC and CGCAGAGGCCAGAGGATG; and Kit −114 kb, GCACACAGGACCTGACTCCA and GTTCTGAGATGCGGTTGCTG.

RESULTS

SetD8 promotes a developmental transition required for primary erythroid cell maturation.

SetD8 is a context-dependent GATA-1 corepressor at a subset of GATA-1 target genes (41). To evaluate SetD8 functions during erythroid maturation, we conducted an shRNA-based loss-of-function analysis in primary murine fetal liver hematopoietic precursors. Freshly isolated lineage-depleted hematopoietic precursors from E14.5 fetal livers were infected with retrovirus expressing shRNA targeting Setd8. After 72 h of culture in medium supporting erythroid precursor cell expansion, we quantitatively analyzed erythroid maturation using flow cytometry with the surface markers CD71 and Ter119 (Fig. 1A). Downregulating Setd8 mRNA by 70 to 80% (Fig. 1A), which strongly lowered SetD8 protein levels (Fig. 1B), significantly reduced cells in the R3 population (early and late basophilic erythroblasts) (2.5-fold decrease, P = 0.00001) and increased the R2 population (proerythroblasts) (1.6-fold increase, P = 0.0004) (Fig. 1C). In addition, downregulating SetD8 significantly increased cells in the R1 (CFU-E and BFU-E) and the very low abundance R4 (orthochromatic erythroblasts) and R5 (reticulocytes and erythrocytes) populations (Fig. 1C). In cells cultured under conditions promoting differentiation, SetD8 knockdown induced 4.3-fold (P = 0.0005) and 2.4-fold (P = 0.002) increases in R1 and R2 populations, respectively, while reducing R3 population cells 1.8-fold (P = 0.00005) (Fig. 1C). The accumulation of R2 cells, concomitant with reduced R3 cells, suggests that SetD8 promotes the developmental transition of the immature R2 erythroblast to a more mature R3 erythroblast.

FIG 1.

SetD8 promotes maturation of primary erythroid precursor cells. (A) (Left) Diagram depicting R1 through R5 erythroid precursor populations based on Ter119 and CD71 surface expression. (Right) Quantitative real-time RT-PCR analysis of Setd8 mRNA in FACS-sorted R1, R2, R3, and R4/5 cells (n = 6, values are means ± standard errors [SE]). *, P < 0.05; **, P < 0.01; ***, P < 0.001. (B) Analysis of SetD8 protein levels by Western blotting (image representative of 3 independent experiments). (C) Representative plots from flow cytometric analysis of erythroid maturation based on expression of the surface markers CD71 and Ter119 after 3 days of culture in medium promoting progenitor expansion (top) or differentiation (bottom). The average percentage of total cells in the R1, R2, R3, R4, and R5 populations after treatment with control or Setd8 shRNA is depicted on the right (n = 6 for expansion, n = 3 for differentiation; values are means ± SE). Asterisks are as defined for panel A. (D) Representative images from Wright-Giemsa-stained flow-sorted R2 cells treated with Setd8 or control shRNA.

Wright-Giemsa staining of flow-sorted, live R2 cells cultured under expansion conditions indicated that downregulating SetD8 induced profound membrane blebbing (16% of cells, a 160-fold increase over control), which is often an attribute of apoptosis (Fig. 1D). This result suggested that SetD8 might suppress apoptosis in maturing erythroid precursor cells. We tested whether SetD8 downregulation induces apoptosis by staining SetD8 knockdown and control cells with annexin V and propidium iodide (PI) and quantitating the percentage of apoptotic cells using flow cytometry. SetD8 knockdown increased the percentage of early apoptotic cells (PI negative, annexin V positive) 6.8-fold and increased late apoptotic cells 4-fold under expansion culture conditions (Fig. 2A). Flow cytometric quantitation of apoptosis using the apoptotic markers active caspase-3 and cleaved PARP indicated that SetD8 knockdown caused a 2.7-fold (P = 0.0005) increase in active caspase-3/cleaved PARP double-positive cells (Fig. 2B).

FIG 2.

SetD8 confers erythroblast survival. (A) Representative plots from flow cytometric analysis of apoptosis using annexin V and propidium iodide staining in cells treated with Setd8 or control shRNA under expansion culture conditions. The average percentage of cells in the live, early apoptotic, late apoptotic, and necrotic populations are displayed on the right (n = 3; values are means ± SE). (B) Representative plots from flow cytometric quantitation of apoptosis using anti-active caspase-3 and anti-cleaved PARP antibodies. Averages of active caspase-3/cleaved PARP double-negative and double-positive populations are displayed on the right (n = 3; values are means ± SE). (C) Quantitative RT-PCR analysis of Bcl2l1 and Bax mRNA in sorted R1, R2, R3, and R4/5 cells treated with control or Setd8 shRNA (n = 6; values are means ± SE). (D) Representative plots from cellular proliferation analysis of proerythroblasts (R2 population, top) and basophilic erythroblasts (R3 population, bottom) using CellTrace Violet dye. The average percentage of total cells in each daughter generation is depicted on the right (n = 3; values are means ± SE). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

An established erythroblast mechanism to counteract proapoptotic signals involves GATA-1-mediated induction of the antiapoptotic factor Bcl-x_L encoded by Bcl2l1 (57). In addition, our prior analysis revealed SetD8 repression of the gene encoding the proapoptotic factor Bax in G1E-ER-GATA-1 cells (41). We tested whether SetD8 downregulation deleteriously impacted erythroblast survival by decreasing Bcl2l1 and/or increasing Bax expression in any population. Real-time RT-PCR analysis indicated that SetD8 downregulation increased Bcl2l1 mRNA 2-fold (P = 0.005) and 5-fold (P = 0.006) in R3 and R4/R5 cells, respectively, without changing Bax expression (Fig. 2C). Thus, reducing SetD8 levels did not induce erythroblast apoptosis via downregulating Bcl2l1 or upregulating expression of the proapoptotic gene Bax.

SetD8 knockdown cells accumulated considerably slower in culture than in cells treated with control shRNA. We tested whether SetD8 inhibits cell proliferation by staining cells with a membrane-intercalating dye and monitoring cell divisions by flow cytometry. After 24 h, the vast majority of cells expressing control shRNA underwent 5 or 6 cell divisions. SetD8 knockdown had little to no effect on the percentage of cells that underwent 6 cell divisions in both the R2 and R3 populations (Fig. 2D). These results suggest that SetD8 confers fetal liver erythroblast survival by counteracting apoptosis, rather than exerting a major influence on proliferation of viable cells.

SetD8-mediated Gata2 repression in immature erythroblasts.

To dissect the molecular mechanism governing the SetD8-dependent R2-to-R3 transition, we tested whether lowering SetD8 levels altered the expression of key erythroid transcription factors and coregulators. To ensure that any potential alterations in gene expression do not reflect changes in cellularity in the hematopoietic precursor cell population, we isolated R1, R2, R3, and R4/5 cell populations by FACS and quantitated gene expression in these populations. While the SetD8 knockdown did not change Gata1, Fog1, Klf1, Tal1, and Pu.1 expression, Gata2 mRNA increased 2- to 3-fold (primer set 1; P = 0.0002) selectively in the R2 population (Fig. 3A). A second primer set targeting a distinct Gata2 exon-exon junction yielded an identical result; SetD8 downregulation induced Gata2 mRNA in the R2 population 2.4-fold (primer set 2; P = 0.0007). Western blot analysis of FACS-purified R2 cells revealed that reducing endogenous SetD8 protein upregulated GATA-2 protein levels (Fig. 3B). GATA-2 promotes HSC generation and function and regulates hematopoietic progenitor survival and proliferation (9–12). As GATA-2 overexpression in wild-type (WT) hematopoietic precursors inhibits hematopoiesis (58) and GATA switching is considered to be a driver of erythroid maturation (33, 59), SetD8-mediated repression of Gata2 transcription may be an important determinant of the transition from an immature erythroid precursor cell to an erythroblast destined to undergo enucleation to yield a reticulocyte and subsequently an erythrocyte.

FIG 3.

SetD8 represses Gata2 in the proerythroblast population. (A) Real-time RT-PCR analysis of Gata2 (2 primer sets targeting distinct exon-exon junctions), Gata1, Fog1, Klf1, Pu.1, Vim, Tal1, Kit, and Lyl1 mRNA in FACS-sorted R1, R2, R3, and R4/5 cell populations (n = 6; values are means ± SE). (B) Analysis of GATA-2 protein level in FACS-sorted R2 cells by Western blotting (image representative of 3 independent experiments). (C) Time course of Setd8 mRNA, Gata2 primary transcript, and Gata2 mRNA expression levels following estradiol treatment in G1E-ER-GATA-1 cells (n = 3; values are means ± SE). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

In the R2 population, SetD8 knockdown elevated Vim mRNA 2-fold (P = 0.03) (Fig. 3A). Vim encodes the intermediate filament vimentin, which is strongly repressed upon primary erythroid cell maturation. Previously, we demonstrated that GATA-1 utilizes SetD8 to repress Vim expression (41). SetD8-mediated Vim repression may have important functional consequences for this developmental transition, as it has been proposed that downregulating vimentin serves a permissive function for erythroid maturation (60, 61). SetD8 knockdown induced a small, but significant, increase in Kit mRNA levels in the R2 population, as well as a significant decrease in Lyl1 mRNA in the R5 population (Fig. 3A). As Kit is an established direct GATA-2 target gene (62), the increased Kit expression in R2 cells upon SetD8 knockdown may reflect increased GATA-2 levels/activity.

Previously, we reported that Gata2 mRNA expression was not regulated by SetD8 in G1E-ER-GATA-1 cells under conditions in which ER–GATA-1 was inactive and also under conditions in which ER–GATA-1 was conditionally activated for 24 h. However, our finding that SetD8 represses Gata2 mRNA in R2 cells but not R1 or R3 cells suggested that SetD8 may be required for initiation but not maintenance of Gata2 repression. To test this, we infected G1E-ER-GATA-1 cells with control or Setd8-specific shRNA and quantitated Gata2 primary transcripts and mRNA at several time points after β-estradiol-mediated ER–GATA-1 activation. SetD8 knockdown significantly elevated Gata2 primary transcripts at 0, 1, 2, 4, and 6 h of β-estradiol treatment. However, by 12 and 24 h, Gata2 primary transcripts were reduced to a level indistinguishable from that of the control (Fig. 3C). Whereas Gata2 mRNA levels were significantly elevated by SetD8 knockdown at 6 and 12 h after β-estradiol treatment, at 24 h, Gata2 mRNA was repressed to the same extent as in the control (Fig. 3C). These results support a mechanism whereby SetD8 is required for the initiation but not maintenance of GATA-1-mediated Gata2 repression and demonstrate that SetD8 regulates Gata2 in primary cells and in the G1E-ER-GATA-1 genetic complementation assay.

To test whether elevated Gata2 expression contributed to the maturation blockade resulting from SetD8 downregulation, we infected primary fetal liver erythroid precursor cells with a GATA-2-expressing retrovirus. After 72 h of culture under conditions that promote erythroid precursor expansion, GATA-2 expression induced a 1.7-fold increase in both R1 (P = 0.007) and R2 (P = 0.008) cells and a 1.5-fold (P = 0.003) decrease in R3 cells (Fig. 4A). GATA-2 overexpression was confirmed by Western blotting (Fig. 4B). We tested whether the GATA-2-mediated maturation blockade was associated with changes in the expression of other erythroid transcription factors. GATA-2 expression induced little to no change in Gata1, Klf1, Fog1, and Pu.1 expression in FACS-purified R1 to R4/5 populations (Fig. 4C). Thus, elevating GATA-2 expression in primary erythroid precursors induced a maturation barricade comparable to that resulting from endogenous SetD8 downregulation.

FIG 4.

GATA-2 expression inhibits the proerythroblast to basophilic erythroblast transition. (A) Representative plots from flow cytometric analysis of erythroid maturation, using the surface markers Ter119 and CD71, in cells infected with GATA-2-expressing retrovirus or empty vector control. Population averages are displayed on the right (n = 3; values are means ± SE). (B) Western blot analysis of GATA-2 protein levels (image representative of 3 independent experiments). (C) Quantitative RT-PCR analysis of Gata1, Klf1, Pu.1, and Fog1 mRNA levels in sorted erythroid precursor populations (n = 3; values are means ± SE). (D) (Top) Representative plots of flow cytometric quantification of erythroid maturation using cell surface markers CD71 and Ter119. Averages for R1 through R5 populations are depicted (bottom left; n = 3; values are means ± SE), and all results for experimental conditions in each population are significantly different from those for the control sample in the same population. The SE comparison displayed on the graph compares the SetD8 single-knockdown condition to the SetD8/GATA-2 double-knockdown condition. (Bottom right) Western blot analysis of GATA-2 protein levels (image representative of 3 independent experiments). *, P < 0.05.

To test whether reduced GATA-2 is sufficient to rescue the SetD8 knockdown phenotype, we performed a SetD8/GATA-2 double knockdown in fetal liver-derived erythroid precursor cells. Following culture under expansion conditions for 72 h, erythroid maturation was quantitated by flow cytometry. SetD8 knockdown significantly reduced the percentage of cells in R3 and increased the percentage of cells in R2 (Fig. 4D). GATA-2 knockdown induced a profound increase in late-stage erythroblasts; R4 and R5 cells increased 12- and 52-fold, respectively. Thus, endogenous GATA-2 expression strongly blocks maturation. However, other than a slight, but significant, decrease in R1 cells, the SetD8/GATA-2 double knockdown resulted in no significant differences in the percentages of cells in the R2, R3, R4, or R5 populations compared to those in the SetD8 single-knockdown condition (Fig. 4D). Western blot analysis confirmed an efficient knockdown of GATA-2 protein in both single- and double-knockdown conditions (Fig. 4D). Thus, while SetD8 represses Gata2 mRNA and GATA-2 protein in primary R2 cells, lowering GATA-2 is insufficient to rescue the maturation phenotype in SetD8 knockdown cells.

Dissecting the mechanism underlying SetD8-mediated erythroid maturation: H4K20me1 accumulation at a GATA switch site restricts Scl/TAL1 occupancy.

cis elements mediating endogenous Gata2 expression have been established (63). The +9.5 intronic E-box-spacer-GATA composite element strongly enhances endogenous Gata2 expression in hemogenic endothelial cells of the mouse AGM and in fetal liver HSPCs (9, 17). HSC generation from AGM hemogenic endothelium and the establishment of the fetal liver HSPC compartment are both +9.5 site dependent (9, 17). Furthermore, similar fetal liver results were obtained from a novel conditional Gata2 knockout strategy using a +9.5 site-containing sequence driving Cre recombinase expression (64). Accordingly, the +9.5 site is essential for embryogenesis, with +9.5^−/− embryo lethality occurring at ∼E13.5. In contrast, the −1.8 and −2.8 GATA switch sites modestly upregulate Gata2 expression in HSPCs (65, 66), while the −3.9 site has no apparent role (19); these sites are not critical determinants of hematopoiesis or embryogenesis, at least not in the steady state in the mouse. A leukemogenic chromosomal inversion extracts the −77 GATA switch site (22) from the GATA2 locus and relocalizes it ∼40 megabases away to the promoter of the EVI1 protooncogene, thereby upregulating EVI1 expression (67, 68).

We tested whether SetD8 catalyzes H4K20me1 at the essential +9.5 GATA switch site. Whereas the biochemical consequences of H4K20me1 are unresolved, given our demonstration that SetD8 functions predominantly as a repressor in erythroid cells, SetD8-catalyzed H4K20me1 might establish repressive chromatin at the GATA switch site, precluding regulatory complex assembly and/or destabilizing the complex. To determine whether SetD8 directly represses Gata2, we conducted quantitative chromatin immunoprecipitation (ChIP) analysis for H4K20me1 in primary fetal liver cells under expansion conditions (primarily R2 cells) and differentiation conditions (primarily R3 cells), in which Gata2 is active and repressed, respectively. Using primers that tile the +9.5 site, we detected a localized region of elevated H4K20me1 within 1 kb (upstream or downstream) from the +9.5 site coinciding with GATA-1 occupancy under expansion conditions (Fig. 5A). Attempts to measure SetD8 occupancy with anti-SetD8 antibodies did not yield convincing results, and therefore H4K20me1 was used as a surrogate for SetD8 function; SetD8 is the only enzyme known to catalyze this mark. In differentiating erythroid cells in which Gata2 is repressed, H4K20me1 was more broadly enriched, consistent with a spreading mechanism. H4K20me1 levels at the +9.5 site in differentiating cells were comparable to H4K20me1 levels at the repressed muscle-specific MyoD promoter. Consistent with active transcription, H4K20me1 levels at sites 1 kb upstream or downstream of the +9.5 site in expansion culture were comparable to that at the constitutively active Eif3k promoter. Importantly, acetylated histone H4 levels did not change under these conditions, suggesting that increased H4K20me1 did not reflect nucleosome repositioning or deposition of new nucleosomes (Fig. 5A). This result reiterates our prior data in the G1E-ER-GATA-1 cell system, in which repression of direct GATA-1/SetD8 target genes involved broad H4K20me1 spreading (41).

FIG 5.

H4K20me1 is induced at the +9.5 site upon repression of Gata2. (A) (Top) Schematic diagram depicting the Gata2 locus and highlighting the critical GATA switch site 9.5 kb downstream of the Gata2 1S transcription start site. (Bottom) H4K20me1, GATA-1, and acetylated H4 chromatin occupancy measured by quantitative ChIP at the +9.5 site in expanding and differentiating fetal liver progenitor cells. H4K20me1 levels at the repressed MyoD promoter and at the active Eif3k promoter serve as controls (n = 3; values are means ± SE). *, P < 0.05; **, P < 0.01. (B) ChIP-seq profiles of H3K4me3, H3K27me3, H3K4me2, H4K16ac, and H3K9ac in Ter119-negative and Ter119-positive cells at the +9.5 site.

To gain a deeper perspective of the chromatin landscape of the +9.5 site region of the endogenous Gata2 locus, we mined chromatin immunoprecipitation coupled with massively parallel sequencing (ChIP-seq) data in Ter119-negative (R1 and R2) and Ter119-positive (R3, R4, and R5) cells (69) to determine the distribution of histone modifications at the +9.5 site, which infer active transcription or repressive chromatin. Under these conditions, Gata2 expression was repressed 11.3-fold in R3 cells compared to R2 cells as measured by RNA sequencing (RNA-seq) (69). Compared to the +9.5 site in Ter119-negative cells, the +9.5 site in Ter119-positive cells exhibited increased H3K27me3, consistent with the transcriptional repression. The histone modifications commonly associated with transcriptional activity, i.e., H3K4me3, H3K4me2, and H4K16ac, decreased in Ter119-positive cells, also consistent with a repressive state (Fig. 5B).

To determine if SetD8 catalyzed H4K20me1 at Gata2, we knocked down SetD8 in primary fetal liver erythroid precursor cells and quantitated H4K20me1 levels. In the mixed population of erythroid precursor cells, downregulating SetD8 reduced H4K20me1 at sites 480 bp upstream (P = 0.027) and 446 bp downstream (P = 0.041) from the +9.5 GATA binding site. In addition, in R2 cells isolated by FACS, SetD8 downregulation significantly reduced H4K20me1 at sites 480 bp upstream (P = 0.001) and 446 bp downstream (P = 0.027) from the +9.5 composite element (Fig. 6A). These data support a direct mechanism of regulation in which SetD8 establishes and/or maintains H4K20me1 at the +9.5 site.

FIG 6.

SetD8 catalyzes H4K20me1 and restricts Scl/TAL1 occupancy at the +9.5 site. (A) (Left) H4K20me1 chromatin occupancy measured by quantitative ChIP at the +9.5 GATA switch site in total cells treated with control or Setd8 shRNA (n = 3; values are means ± SE). (Right) H4K20me1 chromatin occupancy in FACS-sorted R2 cells treated with control or Setd8 shRNA (n = 3; values are means ± SE). (B) (Left) Quantitative ChIP analysis of GATA-1 at the +9.5 site in cells cultured under expansion conditions and treated with either control or Setd8 shRNA (n = 3; values are means ± SE). (Right) Quantitative ChIP analysis of GATA-1 occupancy at the −77, −3.9, −2.8, and −1.8 GATA switch sites at Gata2, the 1S and 1G Gata2 promoters, a negative-control site 22 kb upstream from the Gata2 1S promoter, and the positive-control sites 114 kb upstream from the c-Kit promoter and Lyl1 intron 1. (C) (Left) Quantitative ChIP analysis of Scl/TAL1 at the +9.5 site in cells cultured under expansion conditions and treated with either control or Setd8 shRNA (n = 3; values are means ± SE). (Right) Quantitative ChIP analysis of Scl/TAL1 occupancy at the −77, −3.9, −2.8, and −1.8 GATA switch sites at Gata2, the 1S and 1G Gata2 promoters, a negative-control site 22 kb upstream from the Gata2 1S promoter, and the positive-control sites 114 kb upstream from the c-Kit promoter and Lyl1 intron 1 (n = 3; values are means ± SE). (D) (Left) Quantitative ChIP analysis of LDB1 at the +9.5 site in cells cultured under expansion conditions and treated with either control or Setd8 shRNA (n = 3; values are means ± SE). (Right) Quantitative ChIP analysis of LDB1 occupancy at the −77, −3.9, −2.8, and −1.8 GATA switch sites at Gata2, the 1S and 1G Gata2 promoters, a negative-control site 22 kb upstream from the Gata2 1S promoter, and the positive-control sites DNase I hypersensitive site 2 (HS2) at the β-globin locus control region (LCR) and Lyl1 intron 1 (n = 3; values are means ± SE). *, P < 0.05; **, P < 0.01.

GATA-1 directly represses Gata2 transcription, and the hematopoietic-specific basic helix-loop-helix transcription factor Scl/TAL1 positively regulates Gata2 through the +9.5 site. Previously, we demonstrated that Scl/TAL1 chromatin occupancy is reduced at the +9.5 site during GATA-1-mediated repression of Gata2 (19). At the genomic level, loss of Scl/TAL1 chromatin occupancy correlates with transcriptional repression (70–72). In addition, we demonstrated that the LIM domain protein LDB1 occupies the +9.5 site and that reactivation of the repressed Gata2 locus in induced G1E-ER-GATA-1 cells upon removal of β-estradiol from the culture medium requires LDB1 (19). Since SetD8 induces H4K20me1 at the +9.5 site, we reasoned that this modification would reconfigure the chromatin environment to be inhospitable to factors interacting with cis elements within this region. To determine if the SetD8-mediated histone modification change alters GATA-1, Scl/TAL1, and LDB1 occupancy at Gata2, we quantitated GATA-1, Scl/TAL1, and LDB1 occupancy at the +9.5 site in expansion culture cells (primarily R2) infected with control shRNA- or Setd8 shRNA-expressing retrovirus. While knocking down SetD8 did not alter GATA-1 or LDB1 occupancy (Fig. 6B and D), the knockdown increased Scl/TAL1 occupancy at the +9.5 site 2-fold (P = 0.0039) (Fig. 6C). Scl/TAL1 occupancy was unaltered at the GATA-binding site in intron 1 of Lyl1, which is unresponsive to SetD8 knockdown (Fig. 6C). In addition, Scl/TAL1 and GATA-1 occupancy were unchanged at the −77, −3.9, −2.8, and −1.8 sites, the 1S and 1G promoters, a negative-control site 22 kb upstream from the Gata2 1S promoter, and the −114 kb GATA-binding element at Kit (Fig. 6B and C). LDB1 occupancy was unchanged at the −77, −3.9, −2.8, and −1.8 sites, the 1S and 1G promoters, a negative-control site 22 kb upstream from the Gata2 1S promoter, DNase I hypersensitive site 2 (HS2) at the β-globin locus control region, and intron 1 of Lyl1 (Fig. 6D). Thus, SetD8 limits the extent of Scl/TAL 1 occupancy at the +9.5 site.

To determine if SetD8-mediated Gata2 repression requires the +9.5 site, we infected freshly isolated HSPCs from an E14.5 mouse heterozygous for the +9.5 element with retrovirus expressing control or Setd8-specific shRNA and cultured for 72 h under expansion conditions. After isolating RNA from FACS-purified R2 cells, we utilized allele-specific primers to quantitate transcription from the WT and mutant (Mut) +9.5 alleles (Fig. 7A). Primary transcripts from the wild-type Gata2 allele were significantly upregulated by SetD8 knockdown. However, the mutant allele lacking the +9.5 element was insensitive to the SetD8 reduction (Fig. 7B). A second set of allele-specific primers yielded an identical result; SetD8 knockdown upregulated primary transcripts from the wild-type but not the +9.5 mutant allele. These results establish a requirement of the +9.5 site for SetD8 to repress Gata2. Since Scl/TAL1 occupies the Gata2 locus (55, 73) and upregulates Gata2 expression (21, 73), the SetD8 restriction of Scl/TAL1 occupancy at the +9.5 site constitutes an attractive mechanism for how this histone methyltransferase represses Gata2 transcription. Thus, SetD8 controls erythroid maturation by functioning as a corepressor for GATA-1, repressing Gata2 transcription, and conferring erythroblast survival (Fig. 8).

FIG 7.

SetD8 requires the +9.5 site to repress Gata2. (A) Schematic of the experimental paradigm in which fetal liver erythroid progenitors were isolated from E14.5 mouse embryos heterozygous for the +9.5 site at Gata2. Cells were infected with control or Setd8-specific shRNA and cultured under expansion conditions for 72 h. GFP⁺ R2 cells were isolated by FACS, and Gata2 primary transcripts were measured using allele-specific PCR. (B) Quantitative RT-PCR analysis of Gata2 primary transcripts using primers specific for the WT or +9.5 mutant allele. Two different sets of primers were used to confirm results (n = 3; values are means ± SE). *, P < 0.05.

FIG 8.

Model depicting SetD8 activity to promote erythroid maturation. SetD8 represses Gata2 specifically in the immature proerythroblast (R2) population. SetD8 catalyzes H4K20me1 and restricts Scl/TAL1 occupancy at the +9.5 site at Gata2. This mechanism limits the capacity of Scl/TAL1 to confer activation through the +9.5 site, which is required for SetD8 to repress Gata2 transcription. The model assumes that Gata2 is one of multiple SetD8 targets that collectively contribute to the control of erythroid cell survival and maturation.

DISCUSSION

Given the plethora of chromatin-modifying and -remodeling enzymes in the proteome (74, 75), one would assume that these factors are intrinsic regulators of essentially all critical biological processes. However, many of these enzymes have not yet been studied in cell-type-specific and/or physiological contexts. Furthermore, targeted deletions of genes encoding these presumptively critical factors can yield early embryonic lethality that precludes tractable functional analysis. This scenario applies to SetD8, which has been rigorously studied, especially in model systems such as Saccharomyces cerevisiae (76), and its mouse knockout yields embryonic lethality before the 4- and 8-cell stages (43). Unlike histone marks in the epigenome catalyzed by multiple enzymes, intriguingly, H4K20me1 is known to be catalyzed by SetD8 only (76). Based on these attributes, our prior work establishing SetD8 as a context-dependent GATA-1 corepressor (41), and many unanswered questions regarding how committed progenitors progressively mature into erythrocytes (77), it is compelling to ask whether SetD8 has important physiological functions in red cell biology. Given the massive chromatin and nuclear reconfiguration associated with erythroid maturation (77, 78), the erythroid system offers unique potential to elucidate how a histone-modifying enzyme orchestrates and/or negotiates complex nuclear transactions in a specialized, but critical, cell type from a multicellular organism. See the companion article by Malik et al. (79).

Although SetD8 is a context-dependent GATA-1 corepressor, whether SetD8 controls the maturation and/or function of primary erythroid precursor cells was unclear. Here, we demonstrated that SetD8 promotes the maturation and survival of definitive erythroblasts. In our prior G1E-ER-GATA-1 cell analysis, small interfering RNA (siRNA)-mediated SetD8 downregulation did not induce apoptosis (41). While the G1E-ER-GATA-1 system recapitulates a normal window of erythropoiesis, i.e., maturing from proerythroblasts to basophilic erythroblasts (51), these cells do not mature efficiently beyond this stage. As maturation in this system is driven by ER–GATA-1, which is constitutively expressed from an integrated retroviral vector, dynamic control of GATA-1 expression from the endogenous Gata1 locus in this cell line differs from that in primary erythroid cells. Though SetD8 downregulation did not alter GATA-1 expression, the unique sensitivity of the primary erythroid cells illustrates that the two systems do not share all parameters. Furthermore, G1E-ER-GATA-1 cells were derived from murine ES cells (52), and the cell derivation process involved expression of the antiapoptotic factor Bcl-2, which might impact cell physiology and contribute to differential susceptibility to apoptosis in the two systems.

GATA-1 directly activates Bcl2l1, which encodes the antiapoptotic factor Bcl-x_L during erythroid maturation (57). SetD8 repressed Bcl2l1 (Fig. 2C), suggesting that SetD8-mediated survival does not involve the transcriptional induction of Bcl2l1. Loss of SetD8 in murine ES cells causes massive spontaneous DNA damage. SetD8 degradation in early G₁ is required for cell cycle progression through S phase, and SetD8 Ser 29 phosphorylation mediates progression through anaphase (45). SetD8 depletion induces DNA damage as a result of new DNA synthesis (43). As erythroblasts undergo several extremely rapid cell divisions during maturation, apoptosis resulting from SetD8 downregulation in the maturing erythroblast may be linked to DNA damage (80).

A common feature of the G1E-ER-GATA-1 and primary fetal liver erythroblast system is the capacity of SetD8 to repress Vim expression (Fig. 3A) (41). Since Gata2 expression is highly upregulated in uninduced G1E-ER-GATA-1 cells due to the loss of GATA-1 that directly represses Gata2 transcription (29), not surprisingly, SetD8 knockdown in these cells did not further enhance the already high-level Gata2 expression. In the primary cells, we established that SetD8 represses Gata2 and elevating GATA-2 expression induces a maturation blockade similar to lowering SetD8 expression. However, reducing the level of endogenous GATA-2 in the context of SetD8 knockdown did not rescue the maturation blockade. These results suggest that SetD8 activity to promote the proerythroblast to basophilic erythroblast transition involves a multicomponent mechanism, including promoting cell survival, repressing Gata2, and repressing other target genes, including Vim. SetD8 catalyzed H4K20me1 at the essential +9.5 GATA switch site, and this chromatin modification was associated with suppression of Scl/TAL1 occupancy at this site but at not other genomic sites bound by GATA factors but not functionally impacted by SetD8 downregulation (Fig. 6).

Although our loss-of-function studies utilized a high-efficiency shRNA-based knockdown strategy, residual SetD8 might limit the magnitude of phenotypic alterations. Nevertheless, modest changes in GATA-2 levels translate into important functional consequences. The magnitude of Gata2 upregulation upon SetD8 knockdown was 2- to 3-fold (Fig. 3A), and studies with Gata2^+/− mice have demonstrated significant deficits in HSPC function with only a 2-fold drop in Gata2 expression (13, 81). GATA2 overexpression correlates with poor prognosis of pediatric (82) and adult (83) AML.

Composite cis elements consisting of an E-box, an 8-bp spacer, and a GATA motif were originally identified as sequences at which GATA-1 assembles a multimeric complex containing Scl/TAL1 and the non-DNA binding components LDB1 and LMO2 (84). GATA-1 and GATA-2 occupy a small percentage of these composite elements in the genome (20, 21). The +9.5 site represents such a composite element, and its targeted deletion in the mouse revealed its requirement for embryogenesis, HSC generation in the AGM, and establishment of the HSPC compartment (9, 17, 22, 55, 85). In certain cases, Scl/TAL1 occupies GATA motif-containing GATA factor-bound chromatin sites lacking the E-box component of the composite element (70).

While it is assumed that the various components of the +9.5 complex all are important mediators of transcriptional activation, whether +9.5 components that limit +9.5 activity exist is unclear. Scl/TAL1 loss from GATA-1-occupied chromatin sites genome-wide correlates with transcriptional repression (70–72). GATA-1, Scl/TAL1, and LDB1 cooccupancy correlates with active transcription. Our analysis demonstrating that SetD8 restricts Scl/TAL1 occupancy at the +9.5 site provides an example of a negative regulatory component that dictates assembly and therefore functionality of an endogenous GATA factor-Scl/TAL1 multimeric complex. As SetD8 catalyzes H4K20me1 at the +9.5 GATA switch site, under conditions in which Scl/TAL1 occupancy is restricted, SetD8-catalyzed H4K20me1 may contribute to the establishment of repressive chromatin, which counteracts mechanisms that ensure +9.5 site accessibility.

SetD8 represses Gata2 in proerythroblasts, in which Gata2 is not completely transcriptionally inactive (Fig. 3A). In the more mature basophilic erythroblast, Gata2 is silenced and insensitive to SetD8 downregulation (Fig. 3A). Since Gata2 is initially repressed in the proerythroblast population, our findings are consistent with a mechanism in which SetD8 is required for initiation but not maintenance of Gata2 repression. This is also relevant to Vim, as SetD8 represses Vim transcription selectively in the proerythroblast population. Previously, we demonstrated that establishment versus maintenance of GATA-1 target genes can be differentially regulated (86). It will be instructive to probe deeply into mechanisms underlying establishment and maintenance phases of GATA factor-mediated transcriptional control, to establish the extent to which SetD8 interfaces with other cellular proteins that establish the repressive state, and to elucidate why certain loci are SetD8 responsive while others are SetD8 insensitive.