The mediator complex functions as a coactivator for GATA-1 in erythropoiesis via subunit Med1/TRAP220

Med1/TRAP220 Day 10.5 Embryos Are Pale but Have a Normal Number of c-Kit+ Progenitor Cells.

Previous studies have shown that ablation of Mediator subunit Med1 leads to embryonic death at day 11.5, which is caused by defects in placental and cardiac development (10, 16, 29). Here, we show that at day 10.5, Med1-deficient embryos are phenotypical pale compared with wild-type embryos, especially visible in the heart region (Fig. 1A). That prompted us to examine erythrocyte development in detail (Fig. 1B). To exclude that this is due to a lack of stem cells in Med1-deficient embryos, we performed FACS analysis with the stem/progenitor marker c-Kit using yolk sac, aorta-gonads mesonephros (AGM) and fetal blood cells of 10.5 days postcoitum (dpc) embryos. In all three locations, c-Kit+ progenitor cells were found in equal numbers, comparing both wild-type and Med1-deficient embryos (Fig. 2B and data not shown). In addition Med1 mRNA expression level remains constant during erythropoiesis as checked by RT-PCR (data not shown).

Analysis of Fetal Day 10.5 Precursor Cells by Flow Cytometry and Methylcellulose Assays.

To investigate erythroid development in Med1-deficient embryos, we first established a FACS scheme to analyze and sort erythroid progenitor cells for functional methylcellulose assays. For the FACS-scheme erythroid-specific marker TER119 and transferrin receptor (CD71) were used, as described for definitive erythropoiesis (30). To identify and sort progenitors from day 10.5 embryos, we used in addition to Ter119 and CD71 stem cell factor receptor c-Kit and panleukocyte marker CD45, which was found to be present also in the stem cell compartment (31). An expression profile of these markers during erythroid development is depicted in Fig. 1C. In the fetal blood of day 10.5-dpc embryos, there are ≈1% of c-Kit+ progenitor cells (Fig. 2A), which can be separated further in CD45+ and CD45- fractions. The sorted c-Kit+/CD45+ cells, staining negative for Ter119 and low for CD71, gave rise predominantly to BFU colonies and not CFU-Es [supporting information (SI) Fig. 6]. The c-Kit+/CD45- population, low for Ter119 and high for CD71, gave rise to CFU-Es but scarcely to BFU colonies (SI Fig. 6). Using this scheme, we performed single-cell sorts into 96-well plates of the different cell populations mentioned above and assayed them for colony formation. Although the precursor cells are present in both Med1-deficient and wild-type embryos (Fig. 2 B and D), we observed a 10× reduction of CFU-E formation in the Med1-deficient cells (Fig. 2C). For the BFU colonies, formation of myeloid and erythromyeloid colonies was normal, whereas formation of BFU-Es was 2- to 3-fold reduced (Fig. 2E). The absolute numbers of colonies formed from wild type with Med1-deficient progenitors can be seen in SI Table 1. As an aside, we note that the efficiency of colony formation using embryonic c-Kit+ cells is markedly higher than using cells from bone marrow (data not shown).

Med1 Physically Interacts with Erythroid Master Regulator GATA-1.

Next, we sought to investigate the molecular basis for the defective erythropoiesis in Med1-deficient embryos. Med1 is part of the Mediator complex that has been shown to interact with ligand-bound hormone receptor (reviewed in refs. 4 and 6). Recently, two groups described an interaction between GATA factors and Med1 (14, 32). Because GATA-1 is a master regulator in erythropoiesis, we investigated whether GATA-1 and Med1 can physically interact. To address this question, we used the yeast two-hybrid assay with four different regions of Med1 (amino acids 2–403, 386–728, 680-1085, and 1099–1560). Med1 fragments were fused to the Gal4 DNA-binding domain (bait), and GATA-1 was fused to the Gal4-activation domain (prey). Only amino acids 386–728 of Med1 were able to interact with GATA-1 (Fig. 3A). We verified these results using a GST-pulldown assays (SI Fig. 7). In vitro-transcribed and -translated GATA-1 was offered to interact with different regions of Med1 fused to GST and immobilized on glutathione beads. We could narrow down the interaction region of Med1 to amino acids 622–701 (SI Fig. 7A). In a modified GST-pulldown assay offering GST-GATA-1 and either HeLa nuclear extract (SI Fig. 7B) or affinity-purified Mediator only (SI Fig. 7C), endogenous Mediator binds to GATA-1. In further GST-pulldown assays using GST-GATA-1 and different in vitro synthesized Mediator subunits, not only Med1 but also Med17/TRAP80 and Med14/Rgr1/TRAP170 was found to interact with GATA-1 (Fig. 3B). In contrast, Med4, Med7, Med8, Med15, Med23, and Med27 did not interact with GST-GATA-1. These results suggest that GATA-1 contacts not only one but at least three different Mediator subunits. GATA-1 harbors two zinc-finger domains; the C-terminal zinc finger is required for the binding to DNA, and the N-terminal zinc finger known to interact with FOG-1 and GATA-1 itself. Using GATA-1 mutants (delta N-terminal zinc finger, delta C-terminal zinc finger, and the C-terminal truncation lacking both the N- and C-terminal zinc fingers), we could show that the N- but not the C-terminal zinc finger of GATA-1 is required for interaction with Med1 (Fig. 3C).

To investigate the association of endogenous Mediator with GATA-1 in vivo, we used the biotinylation tagging approach in erythroid cells (ref. 40; Fig. 3D). Input and pulled-down proteins were analyzed by Western blot for the presence of Med1, GATA-1 (positive control), and DEK (negative control). Only GATA-1 and Med1 but not DEK are present in the pulled-down fraction (Fig. 3D, lane 2, and data not shown).

Mediator Subunit Med1 Is Required for Transcriptional Activation by GATA-1.

We next tested whether the Mediator subunit is required for GATA-1 transcriptional activation at endogenous promoters and in a GATA-1 reporter assay (Fig. 4). A real-time PCR-based reporter assay was used with the reporter plasmid pG-OVEC containing four copies of an optimal GATA-1-binding site (34) transfected into mouse embryonic fibroblasts (MEFs) together with (or without) GATA-1. A mutated plasmid (pmt-OVEC), to which GATA-1 can no longer bind, was used as a negative control (35). The GATA-responsive promoter was activated only if GATA-1 was cotransfected in both Med1+/+ (10-fold) and but not in Med1−/− MEFs (Fig. 4). Importantly, GATA-1 transactivation function in Med1−/− MEFs was rescued by expression of either the full length human Med1 or the Med1 amino acids 1–670. The expression of the rabbit β-globin expression was increased up to 4-fold. The mutated pmtOVEC control plasmid was not activated in either of the different MEFs. Using the same reporter assay as in Fig. 4, different GATA-1 constructs were tested. The deletion of the N-terminal zinc finger, shown to interact with Med1, leads to a complete abrogation of rabbit β-globin expression (SI Fig. 8A). The same is true for the C-terminal zinc-finger deletion mutant or C-terminal truncation mutant (Delta 208-end) lacking both N- and C-terminal zinc fingers.

To look at endogenous expression of GATA-1 target genes in Med1-deficient embryos, we sorted c-Kit+ progenitor cells from the fetal blood of day 10.5 embryos (SI Fig. 8B). As expected, the Med1 transcript is not present in the Med1-deficient embryos, whereas hypoxanthine phosphoribosyltransferase and c-Kit transcript levels are comparable with wild-type level. GATA-1 is known to positively autoregulate its own expression during erythroid differentiation (33). Therefore, we observe not only a severe reduction of GATA-1 target genes βH1 globin and β-major globin but also a strong reduction in GATA-1 expression level.

Mediator Colocalizes with GATA-1 Regulatory Sequences in Vivo.

To investigate whether the (activating) Med1-GATA-1 interaction correlates with colocalization in vivo, ChIP experiments were performed (Fig. 5). First, we used bone marrow progenitor cells immortalized by transduction with HoxB4 (D.v.E., unpublished work) that were either noninduced (Fig. 5A and SI Fig. 9B) or induced with erythropoietin for 3 days (Fig. 5B and SI Fig. 9B). Preliminary FACS and RT-PCR experiments suggest that these cells correspond to early committed erythroid progenitor cells. GATA-1 is specifically found to bind at enhancer elements −26 and −0.75 kb with respect to the start of transcription in the GATA-1 locus. Both polII and Mediator (subunits Med1 and Med17) are found to be present to −0.75 and +0.08 kb but not at control sequences −14.4 and +13.5 kb. Furthermore, in SI Fig. 9A, similar ChIP experiments were performed by using erythroid cells isolated from the spleen of phenylhydrazine-treated mice (36). Erythroid (Ter119+) cells were first enriched by depleting all other cell types with a mix of lineage markers. Strikingly, the overall ChIP profile of GATA-1, Med components, and polII looks similar to the profile obtained in erythropoietin-induced progenitor cells (Fig. 5B). Similarly, ChIP experiments looking at the GATA-1 target gene β-globin show that Mediator subunits Med1 and Med17 as well as polII occupancy correlate with the presence of GATA-1 (SI Fig. 9B).

Animals and Tissue Preparation.

Med1/TRAP220 heterozygous deficient mice were generated by Roeder and colleagues (10). Embryos at 10.5 dpc, staged by somite counting, were dissected and genotyped by PCR. Embryonic yolk sac and AGM were isolated and digested with 0,1% collagenase D and DNase I in PBS/5% FCS at room temperature over 2 h. The cell suspension was filtered and stored on ice until further use. Embryonic blood was collected over 1 h by bleeding the headless embryo in PBS/5% FCS on ice.

Flow Cytometry Analysis and Cell Sorting.

Isolated YS, AGM, and embryonic blood cells were blocked with 10 μg of rat IgG antibody in PBS/5% FCS (Jackson ImmunoResearch, West Grove, PA) over 15 min on ice. Cells were washed once with PBS/FCS and stained with antibodies against c-Kit, CD45, CD71, and Ter119 (PharMingen, San Diego, CA) in PBS/FCS over 1 h on ice. Flow cytometry analysis and cell sorting were performed by using BD FACS Aria (Beckton Dickinson, Franklin Lakes, NJ). A single-cell sort of c-Kit+/CD45+/CD71−/Ter119− and c-Kit+/CD45−/CD71+/Ter119− cells into 96-well plates containing methylcellulose and growth factors was performed to investigate colony-forming ability.

Methylcellulose Assays.

Cells obtained as described above were cultured in methylcellulose (M3334; Stem Cell Technologies, Vancouver, BC, Canada) to investigate CFU-E and BFU colony-forming capacity. CFU-E assays were supplemented with 2 units erythropoietin (Epo)/ml and incubated over 2 days at 37°C with 5% CO₂. BFU assays were performed by adding 2 units of Epo/ml and IL-3 (supernatant) over 8 days at 37°C with 5% CO₂. The cultures were stained with benzidine (12% acetic acid containing 0.4% benzidine and 0.3% H₂O₂) to indicate the amount of erythroid colonies.

GST-Pulldown and Biotinylation-Tagging Approaches.

GST pulldown assays were performed as described in ref. 41 (see also SI Text). Mouse erythroid leukemia cells expressing bacterial BirA biotin ligase as well as tagged GATA-1 were used for the purification as described (ref. 27; see also SI Text).

Transcription Assays.

MEFs (see SI Text) were transiently transfected with 10 μg of plasmid DNA in total by using the Amaxa (Cologne, Germany) MEF transfection kit. A CMV-β-Gal plasmid was included as transfection efficiency control. Plasmids used were: WT (pGOVEC) or mutated (pmtOVEC) plasmids containing a rabbit globin reporter (kindly provided by J. Strouboulis, Erasmus Medical Center, Rotterdam, The Netherlands); GATA-1 cDNA cloned into pcDNA 3.1; pcDNA-FLAG-human Med1 full length; pMSCV human Med1 amino acids 1–670 (kindly provided by K. Ge, The Rockefeller University); GATA-1 D amino acids 308-C term; GATA-1 D N-finger amino acids 200–248; and GATA-1 D C-finger amino acids 249–290 (kindly provided by G. Blobel, Pennsylvania School of Medicine, Philadelphia, PA). Transcription was assayed by real-time PCR with primers for exon 2 of the reporter gene.

ChIP Assays.

In the ChIP assays, we used either uninduced or HoxB4 cells (see below) induced with 3 units/ml EPO over 3 days. Chromatin preparation by using 1 × 10⁷ of the indicated cells was done with 0.4% formaldehyde for 10 min at room temperature. Sonication to ≈500-bp fragments and immunoprecipitations was performed as described in the Upstate Biotechnology (Lake Placid, NY) protocol. For the GATA-1 ChIP, preclearing of the chromatin was done with Sepharose-G beads overnight at 4°C, and the GATA-1 protein-DNA complexes were immunoprecipitated for 3 h in an additional step using an AffiniPure rabbit anti-rat antibody (Jackson ImmunoResearch, West Grove, PA). Antibodies were used as follows: GATA-1, N6; TRAP220, M-255; TRAP80 (CRSP77), G17; and Pol II, H-224 (all from Santa Cruz Biotechnology, Santa Cruz, CA).

Cell Culture.

HoxB4 cells were cultured in SF-Medium [IMDM with l-glutamin and 25 mM Hepes supplemented with 2% FCS/300 μg/ml Primatone RL (Sigma, St. Louis, MO)/5 mg/ml insulin (Sigma)/1× Pen-Strep (Gibco, Carlsbad, CA)/1× MEM nonessential amino acids (Gibco)] containing 20 μg of stem cell factor/2% IL-6 and -3 final (supernatant). Induction of erythroid differentiation was done over 3 days with 3 units/ml EPO in IMDM/15% FCS/1× Pen-Strep/1% BSA.