Graded Levels of GATA-1 Expression Modulate Survival, Proliferation, and Differentiation of Erythroid Progenitors*

Transcription factor GATA-1 plays an important role in gene regulation during the development of erythroid cells. Several reports suggest that GATA-1 plays multiple roles in survival, proliferation, and differentiation of erythroid cells. However, little is known about the rela-tionship between the level of GATA-1 expression and its nature of multifunction to affect erythroid cell fate. To address this issue, we developed in vitro embryonic stem (ES) culture system by using OP9 stromal cells (OP9/ES cell co-culture system), and cultured the mutant ( GATA-1.05 and GATA-1-null ) and wild type ( WT ) ES cells, respectively. By using this OP9/ES cell co-culture system, primitive leading to decreased apoptotic incidences. This, together with altered cell cycle kinetics, accounts for the increased proliferation potential seen in definitive erythroblasts differentiated from GATA-1.05 ES cells.

Transcription factor GATA-1 recognizes conserved GATA motifs ((T/A)GATA(A/G)) in the regulatory regions of many genes encoding erythroid-restricted proteins, such as globins, heme biosynthetic enzymes, membrane proteins, and transcription factors (1,2). To analyze the in vivo function(s) of GATA-1, GATA-1-deficient mice were generated (3). Disruption of primitive erythropoiesis caused GATA-1 homozygous null mutant embryos to die by embryonic day (E) 1 11.5, demonstrating that GATA-1 is required for primitive erythropoiesis. This early demise precluded the possibility of analyzing the role of GATA-1 in definitive erythropoiesis. To experimentally circumvent this impediment, chimeric mice were derived using GATA-1 Ϫ/Ϫ ES cells, and this confirmed that the null mutant cells did not contribute to the mature definitive erythroid pool (4). Thus, GATA-1 is required for the terminal differentiation of both primitive and definitive erythroid progenitors.
Impaired primitive and definitive erythropoiesis in both GATA-1-null and hypomorphic mutant embryos resulted in the generation of extremely limited numbers of erythroid progenitors that could be used for further cytological and molecular analyses. Additionally, we suspected that defective erythropoiesis in the mutant embryos could cause secondary growth retardation, which would in turn affect later hematopoietic development. Under such circumstances where cell-autonomous as well as non-cell-autonomous deficiencies could contribute to the phenotype, it becomes difficult to determine conclusively in vivo how different quantitative levels of GATA-1 may affect the developmental decisions available to an erythroid progenitor cell.
The generation of homogenous erythroid populations from ES cells (7) is a useful experimental tool for analyzing the definitive erythroid population in instances where gene-targeted mutation leads to embryonic death prior to the onset of definitive erythropoiesis. In a two-step ES cell in vitro differentiation method, the ES cells are cultured in methylcellulose medium containing stem cell factor and erythropoietin (Epo) (7). Alternatively, if ES cells are co-cultured with OP9 stromal cells, they differentiate into a hematopoietic cell population that consists of erythrocytes, neutrophils, macrophages, mast cells, megakaryocytes, and lymphoid cells (8). Subsequently, Nakano et al. (9) developed the ES/OP9 cell co-culture system, in which two waves of erythroid (primitive and definitive) cell production were detected after either 6 or 14 days of induction, respectively. Consequently, the ES/OP9 cell co-culture system reflects not only primitive, but also definitive, erythropoiesis in vivo and is a useful tool for dissecting in vitro the functional role of any molecule of interest during erythroid development.
Here, we report that both differentiation and apoptosis are inhibited in GATA-1.05-definitive erythroid cells. Although GATA-1 Ϫ/Ϫ -definitive erythroid cells are similarly arrested in differentiation, they, unlike the GATA-1.05 cells, preferentially undergo apoptosis. Hence, we propose that although normal levels of GATA-1 promote terminal differentiation, the low level of intracellular GATA-1 is insufficient to block continuous progenitor proliferation but is sufficient to prevent apoptosis. In contrast, the complete absence of GATA-1 favors an apoptotic response. In this way, graded levels of transcription factor GATA-1 modulate multiple facets of erythroid cell physiology, including survival, proliferation, and differentiation.

EXPERIMENTAL PROCEDURES
Cell Culture-E14 ES cells were maintained on embryonic fibroblast cells and kept undifferentiated in the presence of recombinant leukemia inhibitory factor (1000 units/ml, ESGRO, Chemicon International). OP9 cells were cultured as described previously (8). After the harvest of ES cells from OP9 feeder cells by trypsinization, 7 ϫ 10 3 cells were plated onto subconfluent OP9 cells grown in ␣-minimum essential medium supplemented with 10% FBS, mouse vascular endothelial growth factor (10 ng/ml, Peprotech), and human bone morphogenic protein-4 (5 ng/ml, R&D systems). After 4 days of co-culture, ES cells were trypsinized and replated onto fresh OP9 cells in ␣-minimal essential medium supplemented with 10% FBS, Epo (2 units/ml; generous gift from Chugai Pharmaceutical), and stem cell factor (50 ng/ml, generous gift from Kirin Brewer Co.).
Non-adherent cells observed on day 6 and day 11 were analyzed as primitive and definitive erythroid cells, respectively. To isolate erythroid cells, floating cells were incubated with biotinylated c-Kit (2B8), CD11b (Mac-1, M1/70), and Gr-1 (RB6-8C5) antibodies for 30 min on ice. After washing twice with washing buffer (2% FBS in phosphatebuffered saline), cells were incubated with streptavidin-conjugated Dynabeads (M-280, Dynal Biotech) for 20 min on ice. Subsequently, cells were washed once with washing buffer, and non-adherent cells attached to the magnet (VarioMACS, Miltenyi Biotec) were collected for further experiments.
On day 11 of differentiation, adherent cells were examined in CFU-OP9 colony formation assay. After washing the OP9/ES cell co-culture dish with phosphate-buffered saline, adherent cells were trypsinized and resuspended in ␣-minimal essential medium supplemented with 10% FBS and incubated for 1 h to eliminate stromal cells. Non-adherent cells were collected and cultured with OP9 cells for 4 -6 days in ␣-minimal essential medium, 10% FBS in the presence of Epo (2 units/ml) and stem cell factor (50 ng/ml). Cobblestone-like colonies, termed CFU-OP9, developed on OP9 cells and were scored.
Flow Cytometry-Cultured ES cells and mouse fetal liver cells (E12.5-E14.5) were harvested and incubated in washing buffer containing Fc block (CD16/CD32; 1:200, BD Pharmingen) for 15 min on ice. Subsequently, cells were washed twice with washing buffer and incubated for 30 min on ice with fluorescence-conjugated antibodies. Then, cells were washed twice and analyzed using FACSCalibur and Vantage (BD Biosciences). The following antibodies were purchased from BD Pharmingen and used for analyses: allophycocyanin-conjugated c-Kit antibody (2B8), phycoerythrin-conjugated TER119 antibody, fluorescein isothiocyanate-conjugated CD71 antibody (C2), and allophycocyanin-conjugated CD44 antibody (IM7). DNA content analysis was performed as described previously (10). Floating cells were gently harvested from ES/OP9 cell co-culture on day 11, and cells were fixed in 70% ethanol. Then, cells were treated with 50-g/ml propidium iodide and 100-units/ml RNase A. Cell cycle distribution was analyzed using ModFit LT software (Verity Software House).
Construction of Retroviral Vectors and Retroviral Infection-Plasmid for murine stem cell virus-internal ribosomal entry site-enhanced green fluorescent protein (MSCV-IRES-EGFP) was kindly provided by Dr. Akihiko Kume. Murine GATA-1 and p16 INK4A cDNAs were independently ligated into the BamHI and EcoRI restriction sites of MSCV-IRES-EGFP. Phoenix-Eco packaging cells were maintained in complete Dulbecco's modified Eagle's medium containing 10% FBS. Phoenix-Eco cells at 80% confluency on 6-cm dishes were transfected with 1 g of DNA using the FuGENE transfection kit (Roche Applied Science). Retroviral supernatant was collected after 72 h and added to NIH3T3 cells to titrate the virus in the presence of 8 g/ml Polybrene (Sigma). After 4 days, FACS analysis for enhanced green fluorescent protein fluorescence was performed to measure the virus titer. Titers of viruses used in this study were Ͼ1 ϫ 10 7 infectious particles/ml.
To establish retroviral packaging cell lines, supernatant from transfected Phoenix-Eco cells was harvested and used to infect PT67 cells (Clontech) in the presence of 8 g/ml Polybrene. After 2-4 days, the brightest green fluorescent protein-expressing cells were sorted using a FACS Vantage and expanded in culture. The expression of GATA-1 and p16 INK4A in NIH3T3 cells infected with viral supernatants was confirmed by RT-PCR. Floating cells were removed from 11-day ES/OP9 co-culture, and adherent cells on OP9 stromal cells were trypsinized. After a 1-h incubation on the dish to remove stromal cells, non-adherent cells were replated onto fresh OP9 cells and co-cultured in the presence of the viral supernatant with 4 g/ml Polybrene for 2 days. The cultures were kept for an additional 4 days in fresh ␣-minimal essential medium, 10% FBS containing Epo and stem cell factor. The number of colonies that developed on OP9 stromal cells was scored. Floating cells were harvested and cytospin samples were stained with May-Grü nwald-Giemsa to verify the differentiation stage of erythroid cells.

RESULTS
Definitive Erythroid Differentiation of GATA-1.05 and GATA-1-null ES Cells-We previously reported that differentiation in GATA-1.05/Y primitive and definitive erythroid cells was blocked at different stages thereby implicating different requirements for GATA-1 levels during erythroid development in distinct hematopoietic organs, such as the yolk sac and fetal liver (6). However, there is no clear molecular explanation for how varied GATA-1 levels might differentially affect the development of the primitive and definitive erythroid lineages. To address this question, ES cells carrying WT, GATA-1.05, and GATA-1-null alleles were separately co-cultured with OP9 feeder cells, and the resultant differentiated primitive and definitive erythroid cells were harvested for further analyses.
Morphological and gene expression analyses confirmed that ES cells cultured on an OP9 feeder layer differentiated into primitive and definitive erythroid cells from days 6 -8 and 11-14, respectively (11). Therefore, WT, GATA-1.05, and GATA-1-null ES cells were cultured on OP9 cells for 11 days, and the non-adherent fractions were then subjected to flow cytometric analyses using anti-c-Kit and TER-119 antibodies (Fig. 1, A and B). Mature definitive erythroid cells (c-Kit Ϫ TER-119 ϩ ) were largely absent from GATA-1.05 and GATA-1-null, but not from WT, ES cell cultures (Fig. 1A). In contrast, immature erythroid progenitors (c-Kit ϩ TER-119 ϩ ) were more abundant in both GATA-1 mutant (GATA-1.05 and GATA-1-null) cultures than in WT cells. Interestingly, the anti-c-Kit signal was ϳ10-fold higher in TER-119 ϩ GATA-1.05 cells than in WT and GATA-1-null cell equivalents. In addition, GATA-1-null ES cell cultures generated the lowest number of non-adherent erythroid cells (Fig. 1C), hinting at the possibility of reduced proliferation and/or exceptionally high cell death in the null mutant cells (below).
We previously reported that the adherent cells in the OP9 co-culture displayed numerous immature hematopoietic cell characteristics based on RT-PCR expression analyses and on morphology of the recovered colonies in methylcellulose medium (11). Surprisingly, the number of adherent cells and CFU-OP9 colonies derived from GATA-1.05 ES cells was the highest among the three ES cell types (Fig. 1, D and E), perhaps as a consequence of differentiation arrest and/or increased proliferation in the c-Kit ϩ TER-119 ϩ immature erythroid cells in GATA-1.05 co-cultures.
Collectively, we concluded that under conditions that promote erythroid differentiation both varieties of GATA-1 mutant ES cells, which express either no or a small amount of GATA-1 protein, produce predominantly immature proerythroblasts and few mature cells. Furthermore, cells with abnormally high proliferative potential are recovered from GATA-1.05 ES/OP9 cell cultures.
Elevated Apoptosis in GATA-1-null Erythroid Cells-May-Grü nwald-Giemsa staining revealed that WT ES cells after 6 days in co-culture contained primitive erythroid cells with orthochromatic cytoplasm and large nuclei, whereas immature blast cells with large, prominent nuclei and polychromatic cytoplasm were present in cultures that developed from both kinds of GATA-1 mutant ES cells (Fig. 2, A-C). Uptake of trypan blue dye (indicating dead or dying cells) and nuclear fragmentation (apoptotic intermediates) were most frequently WT ES cells, after 11 days in co-culture, contained definitive erythroid cells at various differentiation stages, whereas immature blast cells with large, prominent nuclei were the dominant species recovered in both of the GATA-1 mutant ES cell cultures (Fig. 2, G-I). Using benzidine staining to distinguish terminally differentiated erythroid cells, we noted that the benzidine-positive population was significantly reduced in both GATA-1 mutant, but not in GATA-1 ϩ/ϩ ES cell cultures (WT, 48.7 Ϯ 14.9%; GATA-1.05, 8.1 Ϯ 1.8%; GATA-1-null, 0.4 Ϯ 0.1%). Furthermore, at day 11 of differentiation both trypan blue dye uptake and nuclear fragmentation were observed at the highest frequency in the GATA-1-null ES cell culture (Fig.  2, J-L), indicating unusually high apoptotic activity in GATA-1 Ϫ/Ϫ erythroid progenitors.
Taken together, these data indicate that although the low GATA-1 level (5% in the GATA-1.05 mutant cells) is sufficient to avert apoptosis, higher GATA-1 expression levels are necessary to induce terminal erythroid maturation. This observation is consistent with the previous report that GATA-1 could act as a survival factor in committed erythroid cells (12), although the GATA-1 expression level was not determined in that study. Given the intriguing differences observed in ES cells that express graded levels of GATA-1 when exposed to differentiation stimuli, we investigated the activity of candidate GATA-1 target genes during erythroid proliferation versus differentiation.
Bcl-xL Is Highly Expressed in GATA-1.05 ES Cell-derived Definitive Erythroid Progenitors-It has been reported that Epo cooperates with GATA-1 to stimulate Bcl-xL gene expression and to maintain erythroid survival, and that Bcl-xL is essential for normal erythroid differentiation (13). Thus, Bcl-xL seems to be a critical downstream effector of GATA-1-and/or Epo-mediated signals. We therefore investigated the expres-sion of several Bcl-2 family members, including Bcl-xL, in primitive and definitive erythroid populations recovered from ES/OP9 cultures (Fig. 3).
Although Bcl-2 mRNA was absent in GATA-1-null primitive erythroid cells, it was present at similar levels in definitive erythroid cells differentiated from all three types of ES cells. Bcl-xL expression was not detected in primitive erythroid cells (at day 6) from both GATA-1 mutant ES cell cultures. Bcl-xL expression was also not found in GATA-1-null definitive erythroid cells. Remarkably, however, a slightly higher than normal Bcl-xL mRNA level was observed in GATA-1.05 definitive erythroid cells. These findings strongly suggest that Bcl-xL, but not Bcl-2, is important for cell survival during definitive erythroid differentiation and that low levels of GATA-1 may be adequate for inducing ROM, expression to protect against apoptosis.
In contrast, the expression of apoptotic inducers, such as Bax, remained constant during primitive and definitive erythroid differentiation. The stabilization and accumulation of the tumor suppressor protein, p53, has also been shown to contribute to apoptosis. We therefore examined p53 accumulation in GATA-1-mutant erythroid cells. Lower p53 expression was detected in primitive erythroid cells from both GATA-1.05 and GATA-1-null cells compared with WT cells, underscoring the possibility that the apoptosis observed in primitive erythroid GATA-1-null cells is independent of p53, as reported previously (12).
Erythroid Cells Derived from GATA-1 Mutant ES Cells Accumulate in S Phase-It has been reported that forced expression of GATA-1 alters the length of cell cycle segments (14) and that especially high levels of GATA-1 were found to lengthen S phase in NIH3T3 cells (15). In addition, Cullen et al. (16) showed that GATA-1 activity in mouse erythroleukemia cells was low in G 1 phase, peaked in mid-S phase, and then dimin- ished again in G 2 /M phase and showed that the accumulation patterns of GATA-1 protein and mRNA mirrored one another throughout the cell cycle. It is therefore tempting to speculate that GATA-1 may directly regulate the cell cycle and/or that its expression may be tightly regulated during erythroid differentiation.
To determine how the expression level of GATA-1 affects the cell cycle during erythroid maturation, flow cytometric analysis of ES cell-derived hematopoietic cells was initiated. Definitive erythroid cells were collected on day 11 postinduction (Fig. 4A). Although mature definitive erythroid cells from differentiating WT ES cells accumulated in G 0 /G 1 phase, immature definitive erythroid cells in both GATA-1 mutant ES cell cultures accumulated in S phase. These results are consistent with the previous data (Fig. 1, A and B), in which c-Kit ϩ TER-119 ϩ immature cells were found to be more abundant in GATA-1 mutant ES cell cultures, whereas mature definitive erythroid cells (c-Kit Ϫ TER-119 ϩ ) were more abundant in control cultures.
Cell cycle arrest is subject to a variety of regulators, including cyclin D-Cdk complex and INK family inhibitors. We analyzed cyclin D2 mRNA levels in definitive erythroid cells differentiated from GATA-1 mutant ES cells by RT-PCR (Fig. 4B). Cyclin D2 mRNA did not vary significantly among definitive erythroid cells recovered from all three ES cell types. Cyclin D-dependent kinases collaborate with cyclin E-Cdk2 to phosphorylate Rb and its family members, p107 and p130, inactivating their growth inhibitory functions and facilitating S phase entry (17). The expression of Rb was slightly up-regulated in cells carrying GATA-1 mutant alleles in comparison to WT cells (Fig. 4B).
The activities of cyclin D-and cyclin E-dependent kinases are linked to Cip/Kip family of Cdk inhibitors, including p27KIP1 and p21cip1 (18). Both p27KIP1 and p21cip1 are potent inhibitors of cyclin E-bound Cdk2, but are less effective in blocking the enzymatic activity of cyclin D-bound Cdk4 in live cells (17). However, we found that the levels of p27KIP1 and p21cip1 were quite similar among the three definitive erythroid cell populations with WT and mutant GATA-1 allele (Fig. 4B).
Because four INK4 proteins (i.e. p16 INK4A , p15 INK4B , p18 INK4C , and p19 INK4D ) are known to inhibit the activity of cyclin D-dependent kinases to prevent the phosphorylation of RB family proteins (18,19), we finally examined the level of p16 INK4A and p19 INK4D mRNAs by RT-PCR. We found that although the p19 INK4D mRNA level did not change much, p16 INK4A transcripts were significantly reduced in GATA-1.05 and GATA-1-null definitive erythroid cells compared with control cells (Fig. 4B).

GATA-1 but Not p16 INK4A Induces Differentiation of GATA-1.05 ES Cell-derived Erythroid Progenitors-
To determine the role of p16 INK4A and GATA-1 in definitive erythroid maturation, GATA-1.05 erythroid progenitors were infected with retroviruses carrying p16 INK4A or GATA-1 cDNAs (Fig. 5A). Adherent GATA-1.05 erythroid progenitor cells were harvested at day 11 and reseeded onto fresh OP9 cells together with supernatant containing no virus or containing viruses with wild type or modified genomes. We then scored the number of CFU-OP9, which represented immature cell populations. GATA-1 or p16 INK4A virus-infected cells were suppressed in colony formation (Fig. 5B), consistent with previous reports of cell cycle arrest (15,20).
Morphological examination of the non-adherent fractions from GATA-1 and p16 INK4A cultures indicated the presence of more mature erythroid cells as compared with control cultures infected with no or vector-alone virus (Fig. 5C). Compared with the vector alone (Fig. 5C, striped column), the frequency of proerythroblast and basophilic erythroblasts diminished (gray column), whereas the polychromatophilic and orthochromatophilic erythroblast populations increased in the GATA-1-infected culture.
In contrast to GATA-1-infected cells, p16 INK4A -infected cells did not undergo extensive erythroid maturation (Fig. 5C), whereas erythroid-specific aminolevulinate synthase mRNA, a marker of terminal erythroid differentiation, was slightly upregulated in p16 INK4A -infected cells compared with the vector control (Fig. 5D). These results suggest that p16 INK4A induces, albeit marginally, erythroid differentiation. Interestingly, p16 INK4A expression was up-regulated in GATA-1-infected erythroid cells, but the reciprocal was not seen in that GATA-1 did not increase in p16 INK4A -infected cells suggesting that GATA-1 regulates, directly or indirectly, p16 INK4A expression.
To examine whether GATA-1 directly activates p16 INK4A gene transcription, we constructed a reporter plasmid bearing 3.9 kbp of the 5Ј-promoter proximal sequences from the murine p16 INK4A gene, and co-transfected it with a GATA-1 expression plasmid into QT6 cells. Despite the presence of three GATA consensus sequences within the 3.9-kbp DNA fragment, forced expression of GATA-1 failed to enhance reporter gene activity significantly (data not shown). However, co-transfection of FOG-1 cDNA along with GATA-1 cDNA increased the reporter activity to ϳ2.5-fold compared with the control vector transfection (data not shown). Thus, although the increase of reporter gene expression was not so significant as compared with the previously observed GATA-1-mediated increase of the reporter gene expression (for instance see Ref. 21), nonetheless these data further support the contention that the p16 INK4A gene is

FIG. 3. Expression profiles of genes involved in apoptosis in erythroid cells expressing varying GATA-1 levels.
Total RNA isolated from primitive and definitive erythroid cells harvested from day 6 and day 11, respectively, of co-culture was analyzed by RT-PCR (lower (A) and higher PCR cycles (B)). Compared with WT-definitive erythroid cells, Bcl-xL expression was undetected in GATA-1-null (G1-null)-definitive erythroid cells, but was higher in the GATA-1.05 (G1.05) cell equivalents. HPRT, hypoxanthine quanine phosphoribosyl transferase; ddw, deionized distilled water.
one of the targets of GATA-1-mediated transcriptional regulation. The relatively weak induction might be because of the presence of multiple regulatory influences within the 3.9-kbp genomic region. Alternatively, the 3.9-kbp fragment used in this experiment might be missing some important motifs cooperating with GATA-1 for the regulation of p16 INK4A gene.
Expression of the transferrin receptor (CD71) and TER-119 marker in the fetal livers were also analyzed. Most TER-119 ϩ erythroid progenitors expressed CD71 in WT fetal livers (92%); however in E13.5 and E14.5 GATA-1.05/X fetal livers, the expression of CD71 was gradually lost, coincident with loss of TER-119 ϩ cells (18%). We also examined the expression of CD44, which is an adherent cell surface marker on immature erythroid cells, and found that the frequency of TER-119 Ϫ CD44 ϩ cells increased in E14.5 fetal livers of GATA-1.05 embryos (81%), whereas the majority of WT fetal liver cells stained positive with TER-119 and CD44 (82%), indicating that immature TER-119-negative proerythroblasts were predominant in E14.5 GATA-1.05/X fetal livers. May-Grü nwald-Giemsa staining revealed that the GATA-1.05/X fetal livers were populated by immature erythroblasts, substantiating the conclusions from previous flow cytometric analyses (data not shown). The left red color peak represents diploid cells in G 0 -G 1 phase, and the right red color peak represents diploid cells in G 2 -M phase. The blue striped area represents diploid cells in S-phase. The left unshaded peak represents aneuploid cells in G 0 -G 1 phase, and the right unshaded peak represents aneuploid cells in G 2 -M phase. B, RT-PCR analysis of G 1 -S phase transition regulators in erythroid cells expressing different levels of GATA-1. Of all the G 1 -S phase transition regulators examined, only expression of p16 INK4A was impaired in GATA-1.05 and GATA-1-null erythroid cells. G3PDH, glyceraldehyde-3-phosphate dehydrogenase; DDW, deionized distilled water.

DISCUSSION
In the present study, we found that immature GATA-1.05 erythroid cells are highly proliferative and accumulated in S phase because of the down-regulation of p16 INK4A . GATA-1null definitive erythroid cells also accumulate in the S phase and p16 INK4A mRNA is down-regulated as well. However, unlike GATA-1.05 erythroblasts, GATA-1-null erythroid progenitors undergo apoptosis as a consequence of low Bcl-xL content. Thus, these results demonstrate that low versus no GATA-1 expression alters the developmental outcome of erythroid progenitors. This notion is summarized in Fig. 7.
Cell cycle exit is a prerequisite for terminal erythroid differentiation. It was previously reported that forced expression of GATA-1 accelerates erythroid differentiation along with the suppression of cell proliferation (22), whereas overexpression of the transcription factor during mouse erythroleukemia cell differentiation shows the opposite effect (14). Ectopic expression of human GATA-1 by retroviral infection into NIH3T3 fibroblasts results in reduced growth rates by prolonging S phase and altering the proliferative growth response to serum (15). These reports suggest that GATA-1 participates in the cell cycle control. Physical association of GATA-1 with Rb has been suggested to play a role in this process (14), but it has not been reported which cell cycle-associated molecules might be subjected to the actions of GATA-1 at the mRNA level, in partic-ular during erythroid development. The results shown here suggest that p16 INK4A is involved in suppressing cell proliferation.
Although p16 INK4A inhibits colony formation on OP9 cells, p16 INK4A is unlikely to affect profoundly the erythroid differentiation. INK4A gene targeting resulted in the increased proliferation of T cell lineage in the thymus and spleen (23). It is also noteworthy that INK4A heterozygote and homozygote mice are tumor-prone and develop a wide spectrum of cancers, particularly after exposure to chemical carcinogens (22). These results strongly suggest that GATA-1.05 or GATA-1-null erythroid progenitors may acquire genetic lesions at a higher frequency (see below).
Bcl-xL-deficient mice die at E13 because of severe anemia with apoptosis of definitive erythroid cells (24). However, primitive erythroid cells in Bcl-x Ϫ/Ϫ mice were not so drastically underrepresented (25), suggesting that a discriminatory antiapoptotic mechanism might be employed during primitive versus definitive erythroid differentiation in the mutant mice. In the ES/OP9 cell co-culture system, primitive erythroid cells appeared on day 6 and upon the analysis of anti-apoptotic gene expression in primitive erythroid cells, Bcl-xL mRNA was found to be expressed in WT primitive erythroid cells, but not in both GATA-1.05 and GATA-1-null cells. Because the number of primitive erythroid cells in GATA-1 mutants (both 1.05 and null) on day 6 represented one-third or one-fifth of the number of WT primitive erythroid cells, respectively (data not shown), these results suggest that distinct mechanisms for promoting progenitor cell survival may be operative in primitive versus definitive erythroid cells.
Bcl-2 has been shown to prevent apoptosis triggered by various stimuli including oxidative stress, chemotherapeutic drugs, viral infections, and growth factor deprivation (26). Therefore, Bcl-2 was one of the candidates mediating the antiapoptotic function of GATA-1. Importantly, however, erythro- poiesis is normal in Bcl-2-null mutant mice (27), and erythroid progenitors in which Bcl-2 is forcibly expressed maintain Epo dependence in vitro (28), implying that Bcl-2 is not essential for normal erythroid maturation. Because the expression of both Bcl-2 and Bcl-xL was severely diminished in the GATA-1-null primitive erythroid cells, this might be the explanation for the differential response to differentiation induction between GATA-1.05 and GATA-1-null ES cell-derived primitive erythroid cells (Figs. 1 and 2).
GATA-1.05/X mice develop normally but begin to die after about 5 months of age (5). Our extensive cohort analysis has proved that these mice develop two types of overt leukemia; one type is c-Kit-positive non-lymphocytic leukemia, whereas the other is CD19-positive B cell leukemia (29). Both type of mice display anemia and thrombocytopenia with marked splenomegaly. Histological analyses clearly demonstrated that in the former type of leukemic mice proerythroblasts and megakaryocytes accumulated in the spleen, indicating that both erythroid and megakaryocytic lineages were severely affected by the lowered expression of GATA-1.
Interestingly, it was reported that the majority of hemizygous GATA-1 ϩ/Ϫ female mice were anemic at birth but recovered postnatally (3). It has been suggested that clonal selection for erythroid cells expressing GATA-1 occurs in vivo. We also examined the number of GATA-1.05/X neonates that survived postnatally despite their usually anemic appearance. As a result, we found that 14 of 54 neonates from a GATA-1.05/X versus WT crossing were GATA-1.05/X genotype, which was close to the expected Mendelian ratio. The anomalous c-Kit ϩ cells in GATA-1.05/X embryos probably disappeared after birth and/or a small population of normal erythroid progenitor cells proliferate vigorously to compensate for the precipitously diminishing erythroid progenitor pool (Fig. 6), as reported in the GATA-1 ϩ/Ϫ female mice (3). However, it is still unclear how erythropoiesis proceeds normally in postnatal GATA-1.05/X mice or how leukemic changes arise in the adult GATA-1.05/X mouse spleen (5,29). Analyses of GATA-1.05/X mice during embryogenesis may provide insights into the origin of the expanded population of proerythroblasts in the GATA-1.05/X enlarged adult spleen. In addition, a more precise study of the time course leading to the appearance of splenomegaly or severe anemia may shed light on the mechanism(s) underlying the unusual expansion of proerythroblasts in GATA-1.05/X mice.
GATA-1 mutations are frequently found in acute megakaryoblastic leukemia patients with Down's syndrome (30 -35). A premature stop codon is introduced into the coding region encoding the GATA-1 N-terminal activation domain, which results in a truncated GATA-1 protein. Although it is still unclear how expression of the truncated form of GATA-1 pathologically leads to Down's syndrome-acute megakaryoblastic leukemia, one hypothesis that must be entertained in the light of our findings is that the weakened activity of aberrant GATA-1 protein has the same effect on erythroid proliferation and/or differentiation as reduced levels of functionally intact GATA-1 in GATA-1.05 mice and propels the cells toward leukemogenesis. Possibly, reduced expression of p16 INK4A plays a part in this process.
Erythroid progenitors express high levels of both GATA-1 and GATA-2, and GATA-2 expression is markedly down-regulated during erythroid differentiation (7,36). In the presence of WT GATA-1, ectopic expression of GATA-2 in primary chicken erythroblasts leads to a maturation block (37), and forced expression of GATA-2 inhibits erythroid differentiation and promotes megakaryocytic differentiation in K562 cells (38). These data strongly suggest that the intracellular GATA-1/GATA-2 ratio may be crucial in regulating the balance favoring terminal differentiation or proliferation of erythroid progenitors (39).
We demonstrated previously that GATA-1.05 mutant ES cells differentiate into both primitive and definitive erythroid progenitors; however both progenitor populations were unable to mature further in the ES/OP9 cell co-culture system (11). This finding is consistent with the GATA-1.05 mutant embryo studies in vivo, in which maturation arrest was observed in the primitive erythroid population (6). Interestingly, GATA-2 expression was induced 100-fold in GATA-1.05 definitive erythroid cells compared with WT cells (11). Also, abundant GATA-2 expression, which was ϳ50-fold higher than that in WT, was observed in GATA-1-deficient proerythroblasts (40).
Thus, it will be of interest to decipher the cellular mechanisms that govern the coordinated regulation of GATA-1 and GATA-2 during the various stages of erythroid development in vivo. This in turn would help us understand the ontogeny of the leukemogenic changes that occur in GATA-1.05/X mice.