Distinct mechanisms direct SCL/tal-1 expression in erythroid cells and CD34 positive primitive myeloid cells.

The SCL/tal-1 gene (hereafter designated SCL) encodes a basic helix-loop-helix transcription factor which is pivotal for the normal development of all hematopoietic lineages and which is expressed in committed erythroid, mast, and megakaryocytic cells as well as in hematopoietic stem cells. The molecular basis for expression of SCL in stem cells and its subsequent modulation during lineage commitment is of fundamental importance for understanding how early “decisions” are made during hematopoiesis. We now compare the activity of SCL promoters 1a and 1b in erythroid cells and in CD34 positive primitive myeloid cells. SCL mRNA expression in CD34 positive myeloid cells did not require GATA-1. Promoter 1a activity was weak or absent in CD34 positive myeloid cells and appeared to correlate with the presence or absence of low levels of GATA-1. However, promoter 1b, which was silent in committed erythroid cells, was strongly active in transient assays using CD34 positive myeloid cells, and functioned in a GATA-independent manner. Interestingly, RNase protection assays demonstrated that endogenous promoter 1b was active in both erythroid and CD34 positive myeloid cells. These results demonstrate that fundamentally different mechanisms regulate the SCL promoter region in committed erythroid cells and in CD34 positive myeloid cells. Moreover these observations suggest that in erythroid, but not in CD34 positive myeloid cells, promoter 1b required integration in chromatin and/or additional sequences for its activity. Stable transfection experiments showed that both core promoters were silent following integration in erythroid or CD34 positive myeloid cells. Our data therefore indicate that additional regulatory elements were necessary for both SCL promoters to overcome chromatin-mediated repression.

One of the principal issues facing modern developmental biology concerns the molecular mechanisms whereby a multipotent stem cell gives rise to a variety of phenotypically distinct differentiated progeny. Hematopoiesis, the process of blood cell formation, provides a powerful experimental system for studying this process. Since lineage commitment and differentiation involve alterations in patterns of gene expression, the function and regulation of lineage-restricted transcription factors are of central importance for the behavior and subsequent fate of hematopoietic stem cells (1,2).
The SCL/tal-1 gene (hereafter termed SCL) encodes a lineage-restricted basic helix-loop-helix transcription factor with a pivotal role in the regulation of hematopoiesis. Mice lacking SCL protein die by embryonic day 10 and exhibit a complete absence of all hematopoietic cells (3)(4)(5)(6). Previous antisense experiments have suggested that SCL may perform different functions in distinct hematopoietic cell types. Introduction of antisense constructs into a multipotent cell line resulted in reduced proliferation and self-renewal (7), whereas similar experiments inhibited erythroid differentiation of a committed erythroid cell line (8). In addition, loss of SCL function increased apoptosis of a T cell line following serum starvation (9). However, despite its implicit involvement in several physiological processes, target genes for SCL have not yet been identified.
SCL is expressed predominantly in hematopoietic cells, although SCL mRNA and/or protein have also been detected in adult and developing brain together with endothelial cells (10 -13). Within the hematopoietic system SCL is expressed in committed erythroid, mast and megakaryocytic cells as well as in interleukin-3 dependent cell lines (10,(12)(13)(14)(15)(16). Erythroid differentiation of committed erythroid cell lines is accompanied by up-regulation of SCL mRNA although protein levels actually fall (10,17). SCL is also expressed in multipotent progenitors prior to lineage commitment. The use of growth factors to induce erythroid differentiation of a multipotent progenitor cell line was accompanied by up-regulation of SCL mRNA (but not protein), whereas induced granulocyte/monocyte differentiation resulted in extinction of SCL expression (18). Indeed, down-regulation of SCL expression not only accompanies, but may actually be required for normal myeloid differentiation, since over-expression of exogenous SCL impaired macrophage differentiation of M1 cells (19).
The mechanisms that regulate this complex pattern of expression are almost completely unknown. Studies of Drosophila embryogenesis have shown that complex patterns of gene expression are frequently specified by autonomous regulatory elements directing expression to distinct cell types (20,21). Similar mechanisms may operate in vertebrate stem cell systems. However, direct experimental data are lacking and several aspects of Drosophila embryogenesis are very different from post-zygotic vertebrate development. We have therefore chosen to analyze the molecular mechanisms responsible for SCL expression in stem cells and its subsequent modulation during lineage commitment. Such studies should also shed light on the fundamental question of lineage determination during hematopoiesis. Moreover, by studying in detail the regulation of this one gene, we will gain insight into the molecular mechanisms necessary to direct exogenous gene expression in stem cells as well as in specific differentiated hematopoietic cell types.
In view of the powerful experimental tools available for studying hematopoiesis in the mouse, we have focused on the murine SCL gene. The structure of the murine gene is very similar to that of the human gene (22)(23)(24). Transcription of both genes is driven by two promoters in alternate 5Ј exons. SCL promoter 1a directs lineage-restricted expression in erythroid cells and is regulated by GATA-1 (8,25,26), but the molecular basis for SCL expression in hematopoietic stem cells or in other specific cell types is not known. Several lines of evidence suggest that GATA-1 and GATA-2 regulate both overlapping and unique sets of target genes (27). In addition, a number of observations have implied that SCL may be regulated by GATA-2 in primitive hematopoietic cells, a role that is subsequently assumed by GATA-1 following commitment to the erythroid lineage. SCL is almost invariably co-expressed with GATA-1 and/or GATA-2 in hematopoietic cells (10,15,18) and a critical GATA motif in the SCL promoter has been shown to bind GATA-2 (8). In addition, the ratio of GATA-2 to GATA-1 mRNA is high in multipotent cells but reversed during erythoid differentiation (18). Finally, proerythroblasts from GATA-1 null mice express SCL at normal levels but GATA-2 mRNA levels were markedly up-regulated (27).
Taken together these data suggest that GATA-1 and GATA-2 may act through the same regulatory elements to direct SCL expression in distinct human cell types. To address this issue, we have compared the mechanisms responsible for SCL expression in erythroid cells with those operating in primitive myeloid cells which express the stem cell antigen, CD34. Our results show that very different mechanisms are responsible for transcriptional activity of the SCL promoter in CD34 primitive myeloid cells compared with erythroid cells. Moreover, in the CD34 positive cells, transcriptional activity of the SCL promoter region was independent of GATA motifs. This suggests that GATA-2 does not direct SCL promoter activity in primitive hematopoietic cells, an observation with important implications for the hierarchy of transcription factors operating in hematopoietic stem cells.

MATERIALS AND METHODS
Cell Lines-The murine CD34 positive myeloid cell lines M1 and 416B, the murine T-cell line BW 5147, and the murine erythroid cell line J2E have been described previously (25, 28 -30 Plasmids-SCL promoter sequences in all constructs are from a genomic clone isolated from a Balb/C genomic library (24) and have been subcloned into the pGL-2 basic luciferase reporter vector (Promega) as described previously (25). The numeric nomenclature of the different constructs refers in all cases to the 5Ј end of exon 1a (GenBank accession number U05130; position 551). pEF-BOS lacZ contains the lacZ gene under the control of the pEF-promoter (31) and the pA3RSVluc positive control luciferase plasmid contains the enhancer and promoter of the 3Ј-long terminal repeat of Rous sarcoma virus and were kindly provided by Dr. K. Chatterjee (Department of Medicine, Cambridge University, United Kingdom). The construction of pT7 SCL 1b is described under "Ribonuclease Protection Assay." PEF-BOS GATA-1 has been described (25) and pCDNA 3 expression vector was purchased from Invitrogen.
Transient Transfection of Cell Lines-Transient transfection were performed as described (25) with the difference that during the electropulse voltages of 260 V (for the J2E cell line) and 240 V (for the M1 and 416B cell lines) were applied.
␤-Galactosidase and Luciferase Assays-␤-Galactosidase and luciferase assays were carried out as described (25). A linear relationship between light units and amount of luciferase reporter plasmid was confirmed in all cell lines. In the transfection experiments 10 g of pEF-BOS lacZ vector and 15 g of the appropriate luciferase reporter construct were used. In the co-transfection experiments, 15 g of the pEF-BOS GATA-1 expression vector (25) or 15 g of pCDNA 3 empty expression vector (Invitrogen) were used together with 10 g of pEF-BOS lacZ vector and 15 g of the Ϫ2000 1a luciferase reporter construct. For each pulse the relative light units were normalized for transfection efficiency against ␤-galactosidase values obtained from the same sample. The relative light units values presented are the mean of at least four independent experiments and the same pattern was also obtained using a different DNA preparation.
Stable Transfection of Cell Lines-3 ϫ 10 7 F4N or M1 cells were electroporated (960 microfarads, 250 V) with 3 g of linearized phosphoglycerate kinase gene promoter puromycin poly(A) vector (kindly provided by Dr. S. Cory, Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia) and a 10-fold molar excess of the linearized luciferase plasmid. After 24 h, electroporated cells were transferred into selective medium containing 2 g/ml (F4N) or 10 g/ml (M1) puromycin (Sigma). Cells electroporated with SCL reporter constructs were maintained as three independent pools.
Puromycin resistant pools were derived 2-4 weeks following electroporation and luciferase assays were performed using extracts derived from 3 ϫ 10 5 cells for each assay point as described above. For each experiment, assays were performed in duplicate on each pool and a positive control (a pool of cells transfected with pA3RSVluc) and a negative control (a pool of cells transfected with pGL-2 basic) were included in each experiment. This experiment was repeated on three separate occasions for each SCL construct. Results were expressed as fold elevation over the negative control. In addition, this relative luciferase activity was normalized for luciferase DNA content by Southern blot analysis of BamHI and HindIII digested DNA. Filters were probed with a nonrepetitive sequence from the mouse vav gene locus 1 to normalize DNA loading and with the HindIII/BamHI fragment of the luciferase gene from pGL-2 basic. Relative luciferase values were also corrected by normalizing for luciferase DNA content of test pools compared with the control pool carrying pGL-2 basic.
Electrophoretic Mobility Shift Assays-Conditions for the band shift assays were as described (25). Oligonucleotides used in the band shift assays are depicted in Table II. The N6 anti-GATA-1 antibody (34) was kindly provided by Dr. G. Partington, Randall Institute, London, United Kingdom.
RNase Protection Assay-Total cellular RNA was prepared using the RNAzol method according to the manufacturer's protocol (Bio/Gene, Bolnhurst, United Kingdom). Genomic murine SCL fragments encompassing nucleotides 760 -1156 (24) were inserted in antisense orientation into the pT7 vector (38) to generate pT7 SCL 1b. Radiolabeled antisense transcripts were generated by transcription in vitro using T7 RNA polymerase (Promega). Ribonuclease protection assays were performed using the Ambion RPA II kit (Ambion, Austin, TX) and 40 g of total RNA for each protection. The amount and integrity of each RNA preparation was confirmed using the pTRI-␤-actin mouse plasmid (Ambion) which contains a 250-bp mouse ␤-actin gene fragment in antisense orientation and gives rise to a 304-nucleotide transcript. The size of the corresponding protected fragments was determined by running a sequencing reaction alongside the protected fragments.

SCL mRNA Expression in CD34 Positive Primitive Myeloid Cells Did Not
Require GATA-1-SCL mRNA is expressed in committed erythroid, mast, and megakaryocytic cells as well as in multipotent stem cells (10, 14 -16, 18). Since it is difficult to obtain sufficient numbers of normal stem cells for biochemical experiments, we have chosen to study two primitive myeloid cell lines. Both of these cell lines expressed mRNA for the hematopoietic "stem cell" antigen CD34 (Fig. 1). 416B and M1 cells also expressed SCL mRNA, as did two erythroid lines (F4N and J2E). The two erythroid lines expressed high levels of both GATA-1 and GATA-2 mRNA. By contrast, GATA-1 mRNA was undetectable in M1 cells and only very low levels were observed in 416B cells. These results are in accord with previous data (14,16,39) and suggest that SCL mRNA expression in primitive myeloid cells may not require detectable GATA-1 expression.
SCL Promoter 1a Activity in Erythroid and CD34 Positive Primitive Myeloid Cells-We have previously demonstrated that SCL promoter 1a is active in a murine erythroleukemia cell line but not in a murine T cell line (25). Since murine erythroleukemia cell lines cannot be equated with normal erythroid cells, it was important to confirm the generality of our results using an alternative erythroid line, which had been generated in a different manner. For this purpose it was decided to study the J2E cell line, which was derived in vitro by infection of fetal liver with a retrovirus containing myc and raf oncogenes (30). J2E cells are growth factor independent but undergo terminal differentiation in response to erythropoietin.
Luciferase reporter constructs containing SCL promoter 1a were introduced into J2E and a murine T cell line (BW 5147) in transient assays. A ␤-galactosidase control plasmid was also included to control for variation in DNA uptake, and luciferase values were normalized against the corresponding ␤-galactosidase values. A construct, containing approximately 2 kb of exon 1a upstream sequence, was active in J2E cells (34-fold stimulation over background) but not in BW 5147 ( Fig. 2A). Deletion of sequences between Ϫ2000 and Ϫ187 did not decrease promoter 1a activity but further deletion to Ϫ55 resulted in significant loss of promoter activity. The pattern obtained using the promoter 1a deletion constructs in J2E cells was therefore similar to that previously seen in murine erythroleukemia cells (25). We therefore conclude that sequences between Ϫ187 and ϩ26 were sufficient for lineage-restricted activity of promoter 1a in both murine erythroleukemia cells (F4N) and J2E cells.
The same constructs were then used to perform transient assays in 416B and M1 cells. Luciferase constructs containing approximately 2 kb of sequence upstream of promoter 1a were weakly active in both 416B and M1 cells ( Fig. 2A). Deletion of sequences between Ϫ2000 and Ϫ187 upstream of promoter 1a had no effect. By contrast, further removal of sequences between Ϫ187 and Ϫ55 significantly reduced promoter activity. The pattern of promoter activity obtained with the various constructs in M1 and 416B cells was similar to the pattern observed in erythroid cells. However, the overall activity of promoter 1a was much weaker in the primitive myeloid cell lines than in J2E cells.
Promoter 1b Was Independently Active in CD34 Positive Primitive Myeloid Cells but Not in Erythroid Cells-SCL promoter 1b with 128-bp upstream sequence has previously been shown to be inactive in both murine erythroleukemia cells (F4N) and T cells (25). Since murine erythroleukemia cells contain activated ETS family members and since a potential ETS binding motif is present upstream of promoter 1b, it was especially important to confirm the inactivity of promoter 1b in a different erythroid cell line. As shown in Fig. 2B, a construct containing both promoters together with 2 kb of DNA upstream of promoter 1a, was active in J2E cells but not in BW 5147 cells. The activity of this construct (Ϫ2000 1a1b, Fig. 2B) was similar to that observed with an analogous construct containing pro-moter 1a alone (Ϫ2000 1a, Fig. 2A). Moreover, the activity of Ϫ2000 1a1b in J2E cells was abolished by removal of part or all of promoter 1a, and a construct containing promoter 1b alone (with 128 bp of upstream sequence) was inactive. This pattern of activity mirrored that previously observed in F4N cells (25).
Very different results were obtained in M1 and 416B cells (Fig. 2B). In both cell lines, Ϫ2000 1a1b exhibited much higher luciferase levels than Ϫ2000 1a (40 -95-fold over background compared with 5-12-fold over background). Moreover, deletion of part or all of promoter 1a did not reduce the activity. Indeed the ϩ209 1b construct, which only contained 128 bp upstream of exon 1b, was still fully active. Deletion of promoter 1a removed the GATA sites that are situated upstream of exon 1a. Since neither ϩ26 1b nor ϩ209 1b constructs contain any additional GATA motifs, these data demonstrate that the activity of promoter 1b in myeloid cells was independent of GATA motifs.
Transcriptional Activity of Endogenous SCL Promoters in Erythroid and Primitive Myeloid Cells-The transient transfection assays described above suggest that promoter 1a is active in erythroid cells and promoter 1b in primitive myeloid cells. However, such experiments do not take into account the additional effects of distant regulatory elements, methylation status, or chromatin structure. It was therefore important to study the activity of the endogenous SCL promoter region in erythroid and primitive myeloid cells.
The SCL promoter locus is CpG-rich in both human and murine SCL genes (22)(23)(24). Furthermore, in cell lines (but not in primary cells) the methylation status of CpG-rich promoters correlates with their transcriptional activity (40,41). It was therefore important to see whether differential promoter usage in distinct cell types reflected cell-specific differences in pro- moter methylation. We therefore analyzed methylation of the SCL promoter region in J2E, BW 5147, 416B, and M1 cells. As shown in Fig. 3, both SCL promoters are contained within a 0.8-kb Sau3A fragment. Genomic DNA extracted from the four cell lines was digested with Sau3A alone or together with the methylation-sensitive enzyme HpaII and hybridized to the indicated probe. As anticipated, digestion with Sau3A alone generated a 0.8-kb fragment in all cell lines (Fig. 3). Addition of HpaII resulted in complete digestion of the 0.8-kb fragment in all SCL expressing cell lines (J2E, M1, and 416B), but not in BW 5147 cells which did not express SCL mRNA. These data demonstrate that the endogenous SCL promoter was predominantly unmethylated in both erythroid and primitive myeloid cell lines, an observation that fully accords with the transcriptional activity of the SCL promoter region in transient reporter assays. Moreover, within the resolution of this study, no difference was observed in the methylation pattern of the SCL promoter region in erythroid cells compared with primitive myeloid cells.
Reverse transcriptase-polymerase chain reaction has been used previously to assess endogenous transcription of exon 1a and 1b in different cell types (15,24). However, the generation of a polymerase chain reaction product using a primer specific for exon 1b could represent either activity of promoter 1b or read-through transcripts initiating from promoter 1a. RNase protection assays were therefore used to determine the endogenous activity of the two SCL promoters in erythroid, CD34 positive myeloid and T cells. A radiolabeled probe of 449 nucleotides was generated containing part of exon 1b as well as 117 nucleotides of intronic sequence between exon 1a and exon 1b (Fig. 4). This probe was annealed with total RNA extracted from F4N, J2E, 416B, M1, and BW 5147 cells. A murine ␤-actin antisense transcript which was predicted to protect a 250nucleotide fragment was also annealed with the same RNA preparations as a control for equal loading. SCL transcripts initiating from promoter 1b were predicted to give rise to a protected fragment of 280 nucleotides, whereas read-through transcripts that had initiated in exon 1a would protect a larger 397-nucleotide fragment. As shown in Fig. 4, a 280-nucleotide fragment was protected in all erythroid and CD34 positive primitive myeloid lines but not in the T cell line. In addition, a 397-nucleotide band was detected in both erythroid lines but not in the T cell or in either primitive myeloid cell line, although prolonged exposure of the autoradiograph gave rise to a faint band in the 416B lane (data not shown). It is probably no coincidence that 416B cells, which express low levels of GATA-1 mRNA, also exhibited low but detectable activity of promoter 1a in transient reporter assays together with a low level of exon 1a transcripts by RNase protection.
Taken together, these data demonstrate that the endogenous promoter 1b was active in both erythroid and primitive myeloid cell lines but that endogenous promoter 1a activity was only readily detectable in the erythroid cell lines. These results accord with the transient transfection data for promoter 1a. However, promoter 1b was strongly active following transient transfection into primitive myeloid cells, but was silent in erythroid cells. These results have two significant implications. First, they suggest that promoter 1b required additional sequences and/or integration in chromatin for its activity in erythroid cells. Second, they demonstrate that the activity of promoter 1b was achieved by very different mechanisms in erythroid and CD34 positive primitive myeloid cells.
Both Core Promoters Were Silent Following Integration in Chromatin-In view of the possibility that promoter 1b may be chromatin dependent and/or require additional regulatory elements in erythroid cells, a series of stable transfection experiments were performed. The core promoter 1a and promoter 1b luciferase constructs were electroporated into M1 and F4N cells. Three independent pools were generated for each SCL reporter construct in each cell type. Luciferase activity of each pool was analyzed on three separate occasions along with luciferase activity of positive and negative control pools.
As shown in Table I neither promoter 1a nor promoter 1b were active after integration in chromatin. By contrast the positive control vector produced high levels of luciferase activity. These results demonstrate that integration in chromatin was not sufficient for promoter 1b activity in erythroid cells. Indeed, integration in chromatin suppressed activity of both promoter 1a in erythroid cells and also promoter 1b in primitive myeloid cells. Our data therefore suggest that additional regulatory elements were necessary for both SCL core promoters to overcome the repressive effect of chromatin.
SCL Promoter 1a Activity in J2E Cells Was Dependent on GATA and Sp1 Sites-The results presented above demonstrate that SCL promoter 1a and SCL promoter 1b exhibited endogenous activity in J2E cells. However, only promoter 1a was independently active in transient assays and it was therefore readily amenable to further analysis. Previous results from this laboratory (25) have shown that the Ϫ37 GATA site was essential for promoter 1a activity in murine erythroleuke-TABLE I Analysis of SCL core promoters following stable transfection into erythroid and CD34 positive primitive myeloid cells Stably transfected erythroid (F4N) and CD34 positive myeloid cell (M1) were obtained by co-transfection of PGKpuroPa plasmid in conjunction with either pGL-2 basic (negative control), pA3RSVluc (positive control), Ϫ187 SCL 1a, or ϩ209 SCL 1b as indicated. Luciferase activity was normalised against pGL-2 and also corrected for copy number as described under "Materials and Methods." For the SCL constructs three independent pools were assayed in duplicate in three separate experiments. Numbers represent the mean (ϮS.D.) for each experiment. For the control constructs (pGL-2 basic, pA3RSVluc) each number represents the mean of duplicate assays of the same pool which was included in all experiments. mia cells. By contrast, both the Ϫ63 Sp1 and Ϫ69 GATA sites were dispensable. In J2E cells, mutation of the Ϫ37 GATA site dramatically impaired activity of a Ϫ2000 1a promoter construct (Fig. 5A). This result is consistent with previous studies and underlines the importance of the Ϫ37 GATA site for SCL transcriptional activity (8,25,26). No significant reduction of promoter activity was observed when the Ϫ101 AP-1 site was mutated which accords with our previous findings using F4N cells (25). However, mutation of the Ϫ69 GATA and the Ϫ63 Sp1 consensus binding sites reduced promoter 1a activity in J2E cells from 37-fold stimulation over background to 19-and 9-fold stimulation, respectively (Fig. 5A). In F4N cells mutation of the Ϫ69 GATA site produced a modest fall in activity but mutation of the Sp1 motif had little effect. The contrasting results obtained with Sp1 mutation in F4N and J2E cells may reflect differences in the stage of erythroid differentiation or in the complement of genetic changes present in the two cell lines. Band shift assays were performed to characterize the proteins in J2E extracts that bound to the GATA and Sp1 sites. A single strong complex bound the Ϫ37 GATA site (Fig. 6, lane 2). This was competed by cold probe (lane 3) but not by an oligonucleotide with a mutated Ϫ37 GATA site (lane 4) or an unrelated oligonucleotide (lane 5). The complex was shown to contain GATA-1 since a monoclonal antibody to GATA-1 produced a clear supershift (lane 6).
Two major complexes in J2E extracts bound an oligonucleotide containing both Ϫ63 Sp1 and Ϫ69 GATA sites (Fig. 6, filled  arrows). Binding of both complexes was completed by cold probe (lane 9) but not by an unrelated oligonucleotide (lane 10). The upper complex bound to the Ϫ63 Sp1 concensus site and contained a member of the Sp1 family of transcription factors. Complex formation was not competed by an oligonucleotide in which the Sp1 site was mutated (lane 11) but was competed by an Sp1 consensus oligonucleotide (data not shown). Moreover, the complex co-migrated with recombinant Sp1 protein (data not shown). The lower complex contained GATA-1: binding was competed by two different oligonucleotides each containing consensus GATA motifs (lanes 12 and 13), but not by an unrelated oligonucleotide (lane 10) or by an oligonucleotide in which the Ϫ69 GATA site was mutated (lane 15). Furthermore, the N6 GATA-1 monoclonal antibody supershifted this complex (com- pare lanes 14 and 15). Since the GATA-1 specific supershift co-migrated with the upper complex, the supershift was performed in the presence of cold oligonucleotide, containing a mutated GATA-1 site but an intact Sp1 site, to compete the upper Sp1 complex.
A more rapidly migrating complex was evident in the presence of the N6 antibody (lanes 6 and 14). This complex was also found in the absence of antibody or competitor oligonucleotide in lane 8 and, after prolonged exposure of the autoradiograph, in lane 2. This complex has been observed previously (36,42). Its nature was not specifically addressed in our experiments, but it is likely to contain either a smaller proteolytic product of GATA-1 or the product of an alternative AUG start codon, as recently reported by Calligaris and colleagues (43). Both products will have lost the binding site for N6 and will therefore not supershift in our band shift assay.
The low/absent activity of promoter 1a in 416B and M1 cells correlated with the low/absent expression of GATA-1 mRNA. However, both 416B and M1 cells expressed GATA-2 ( Fig. 1) and it has been suggested that GATA-2 may be able to substitute for GATA-1 in directing SCL expression in primitive hematopoietic progenitors (18,27). Moreover, recombinant GATA-2 protein has been reported to bind to the SCL promoter (8). Inactivity of promoter 1a in 416B and M1 cells could therefore reflect either absence of GATA-2 binding to the SCL promoter or failure of bound GATA-2 to transactivate the SCL promoter. Band shift analysis using oligonucleotides containing the Ϫ37 or the Ϫ69 GATA site revealed no specific binding to either GATA site in M1 cells (data not shown), thus suggest- ing that endogenous GATA-2 does not bind to promoter 1a GATA sites in M1 cells. Furthermore, co-transfection of GATA-1 with SCL Ϫ2000 1a produced a 2-3-fold increase in luciferase activity (Fig. 7). These findings suggested that GATA-1 expression in M1 cells is sufficient to allow activity of SCL promoter 1a. It is therefore tempting to speculate that the low/absent SCL promoter 1a activity in 416B and M1 cells is, at least in part, due to low/absent GATA-1 expression in these cell lines.
SCL Promoter 1b Activity in CD34 Positive Primitive Myeloid Cells Required MAZ and ETS Motifs-The experiments described above have demonstrated that the activity of promoter 1b was achieved by very different mechanisms in erythroid and CD34 positive primitive myeloid cells. To dissect these mechanisms we have begun to characterize the transcription factors that regulate promoter 1b.
Previous sequence comparison of the human and murine promoter 1b has revealed a highly conserved region upstream of exon 1b (25). Furthermore, a MAZ-binding site at ϩ242 and overlapping tandem ETS motifs at ϩ264 were necessary for full transcriptional activity of constructs containing both promoter 1a and promoter 1b in F4N cells (25). Since promoter 1b was independently active in the primitive myeloid cell lines, mutations of the MAZ and ETS sites were introduced into the ϩ209 1b construct. As shown in Fig. 5B, both mutations markedly reduced activity of promoter 1b in 416B cells and M1 cells. This effect is unlikely to represent the introduction of novel proteinbinding sites since nuclear extracts from J2E, 416B, or M1 cells did not give rise to complexes after incubation with labeled mutant oligonucleotide (Fig. 8A, lanes 13, 14, and 22).
Band shift analysis was then performed to characterize the proteins binding to the MAZ and ETS motifs. Using an oligonucleotide containing the MAZ site as a probe, specific binding of three major complexes was observed (Fig. 8A). The same three complexes were seen in J2E, 416B, and M1 cells (lanes 2,   7, and 16). Three lines of evidence suggest that either MAZ or a MAZ related protein are present in these complexes. First, all complexes were competed with an excess of an oligonucleotide containing the known ME1a1 MAZ-binding site (44) from the myc promoter (lanes 6, 11, and 20). Second, all three complexes required an intact MAZ site since an oligonucleotide containing a mutated MAZ site could not compete for binding of any of the complexes (lanes 4, 9, and 18). Third, the three complexes co-migrated with complexes obtained using a known MAZbinding site (ME1a1) as a probe (data not shown).
Band shift assays were also performed to study the proteins binding to the ETS motif. The oligonucleotide used as a probe for this experiment contained three overlapping TTCC motifs. Using J2E, 416B, and M1 nuclear extracts, three main complexes were detected in all three cell lines (Fig. 8B, lanes 2, 6,  and 10). Addition of excess cold probe abolished binding of all complexes (lanes 3, 7, and 11), whereas an unrelated oligonucleotide did not compete for binding (lanes 5, 9, and 13). An oligonucleotide in which two of the three ETS binding motifs were mutated still competed for binding of the upper complex (lanes 4, 8, and 12) but not for binding of the lower two complexes. During these experiments it was noted that the addition of an excess of unlabeled oligonucleotide seemed to enhance the binding of the middle complex (asterisk) in all cell lines. We did not further investigate this finding, but one possible explanation would be the presence of an inhibitor which reduces binding of the complex but which is sequestered by the addition of excess cold oligonucleotide. These results suggest that the lower two complexes bound to the TTCCTT sequence that was mutated in the mETS oligonucleotide (Table  II), but that the upper complex bound elsewhere within the oligonucleotide.
Taken together these data demonstrate that the MAZ and ETS sites were important for full activity of promoter 1b in primitive myeloid cells. However, a similar pattern of complexes was observed in J2E cells in which transient reporter assays showed promoter 1b to be silent. These observations suggest that the MAZ and ETS sites are necessary but not sufficient for promoter 1b activity, and that additional lineagerestricted transcription factors may be needed for promoter 1b activity in primitive myeloid cells. Nevertheless, our data do not exclude the possibility that transcription factors binding to the MAZ and ETS sites may undergo lineage-restricted posttranslational modifications which are not detectable by band shift analysis, or may represent comigrating but functionally distinct transcription factor family members. DISCUSSION The SCL gene is essential for the development of all hematopoietic lineages and is expressed in hematopoietic stem cells  Table II as well as committed erythroid, mast, and megakaryocytic cells. Although GATA-1 has been shown to be important for transcriptional regulation of SCL in erythroid cells, the molecular basis for SCL expression in other hematopoietic cell types is unknown. In this paper we have shown that the molecular basis for SCL expression is very different in CD34 positive primitive myeloid cells compared with committed erythroid cells.
Transient reporter assays have previously been used to show that SCL promoter 1a was active in murine erythroleukemia cells but not in murine T cells and that core promoter 1b was inactive in both cell types (25). These studies, together with work from other laboratories (8,26), have also demonstrated a critical role for GATA-1 in regulating the activity of SCL promoter 1a. Murine erythroleukemia cells cannot be equated with normal erythroid progenitors, not least because the former express ETS family members as a result of retroviral insertions. It was therefore important to confirm the pattern of SCL promoter activity in a different class of erythroid cells. Whereas murine erythroleukemia cell lines are generated by infection of mice with Friend virus, the J2E cell line was generated in vitro by infection of fetal liver cells with a replication incompetent virus carrying raf and myc. Analysis of SCL promoter contructs in J2E cells gave rise to the same pattern of activity (active promoter 1a and silent promoter 1b) as previously found in murine erythroleukemia cells. In addition, GATA sites within promoter 1a bound GATA-1 and were critical for promoter 1a function.
Very different results were obtained when SCL promoter activity was studied in CD34 positive primitive myeloid cell lines. In both 416B and M1 cells promoter 1a activity was weak and no specific transcription factor binding could be detected at the Ϫ37 and Ϫ69 GATA sites upstream of promoter 1a. However, the most striking observation was the presence of strong core promoter 1b activity in both 416B and M1 cells. A series of 5Ј deletion constructs was used to show that removal of promoter 1a and the upstream GATA sites did not reduce activity of promoter 1b. Since the constructs with promoter 1b alone did not contain additional GATA sites, these data demonstrate that promoter 1b was functioning in a GATA-independent manner.
GATA-1 plays a central role in the regulation of SCL promoter 1a in erythroid cells. GATA-1 binds to the functionally important Ϫ37 and Ϫ69 GATA motifs and also transactivates promoter 1a in heterologous cells (8,25,26). However, levels of GATA-1 are low in primitive progenitors (18,45,46) and as, shown here, GATA-1 mRNA may be undetectable in primitive myeloid cells expressing SCL. Furthermore, in mice lacking GATA-1 protein, erythroid differentiation was blocked at the  Table II. ME1A1 is an oligonucleotide containing a MAZ site from the myc promoter (44). GATA cons is an oligonucleotide containing a consensus GATA site from the ␣1-globin promoter (42). Arrows indicate the three major complexes binding to the MAZ oligonucleotide. B, ETS site. An oligonucleotide containing the ϩ264 ETS site (ETS) was used as a probe and incubated with nuclear extracts from J2E, 416B, and M1 cell lines as shown. Unlabeled competitor oligonucleotides were included as indicated. The three major complexes are marked by arrows. An asterisk marks the complex with increasing intensity upon addition of cold competitor.  (44); GATA cons, an oligonucleotide containing a consensus GATA-binding site from the murine ␣1-globin promoter (42). Mutated binding sites are underlined.

Oligonucleotide Sequence
Ϫ37 GATA level of the proerythroblast, but GATA-1 target genes including SCL were still expressed at relatively normal levels along with markedly raised levels of GATA-2 (27). It has therefore been suggested that GATA-2 and GATA-1 regulate both overlapping and unique sets of genes (47). According to this hypothesis, SCL would be regulated by GATA-2 in primitive hematopoietic progenitors, a role that is taken over by GATA-1 during erythroid differentiation. Several additional observations are consistent with this scenario: (i) the ratio of GATA-2 to GATA-1 is high in multipotent progenitors and reversed during erythroid differentiation (18,45,46); (ii) SCL and GATA-2 both regulate self-renewal in multipotent progenitor cell lines (7,48); (iii) both SCL and GATA-2 are required for the survival and/or differentiation of pluripotent stem cells during development (3)(4)(5)(6); (iv) GATA-2 binds to a functional GATA motif in the SCL promoter (8); (v) GATA responsive promoters can frequently be transactivated by different GATA proteins (49 -52). Taken together these previous data suggested that GATA-2 regulates SCL in primitive hematopoietic progenitors, and that this is likely to be achieved by GATA-2 interacting with the same regulatory elements that are bound by GATA-1 in erythroid cells. Our results argue against this model and demonstrate that the SCL promoter is active in a GATA-independent manner in CD34 positive primitive myeloid cells. Moreover, these data raise the possibility that initial activation of SCL transcription in pluripotent hematopoietic stem cells may be GATA independent, a speculation that accords with the recent finding that lack of SCL abolishes hematopoiesis even more effectively than lack of GATA-2 (5, 6, 53). However, our results do not address the possibility that GATA-2 may regulate SCL distal regulatory elements in primitive hematopoietic cells.
Transient reporter assays provide valuable information about the transcription factor environment within a cell. However, they do not reflect constraints imposed by chromatin structure or by distant regulatory elements, absent from the reporter constructs. An RNase protection assay was therefore established to investigate the activity of the endogenous SCL promoter in erythroid and primitive myeloid cells. Both promoter 1a and promoter 1b were active in murine erythroleukemia cells and J2E cells. This is consistent with previous reverse transcriptase-polymerase chain reaction data which revealed the presence of transcripts including exon 1a in erythroid cells (15,24). However, a very different pattern was observed in the two CD34 positive primitive myeloid cell lines, both of which exhibited clear activity of endogenous promoter 1b but weak/ absent promoter 1a activity. These results demonstrate differential usage of promoter 1a in different SCL-expressing cell types. Moreover, they suggest that promoter 1b may be the critical site of transcription initiation in stem cells that exhibit low or absent GATA-1 expression.
Our observations also suggest that the mechanisms responsible for promoter 1b activity are fundamentally different in committed erythroid cells compared with CD34 positive progenitor cells. RNase protection assays demonstrated that the endogenous SCL promoter 1b was active in both erythroid and primitive myeloid cells. However, in transient reporter assays the core promoter 1b was strongly active in primitive myeloid cell lines, but silent in erythroid cells. This observation implies that in erythroid, but not in primitive myeloid cells, the core promoter 1b required integration in chromatin and/or additional sequences for its activity.
To distinguish these possibilities stable transfections were performed. These demonstrated that integration in chromatin was not sufficient to allow activity of promoter 1b in erythroid cells. In fact both core promoters were silent following integration in erythroid or myeloid cells. Thus, although some hema-topoietic promoters are active after integration into chromatin (54,55), this is clearly not the case for either SCL promoter. Instead our results strongly suggest that additional regulatory elements are necessary for both SCL promoters to overcome chromatin-mediated suppression.
It will now be important to dissect the different mechanisms responsible for endogenous promoter 1b activity in the two cell types, not least because it may be possible to construct promoter 1b variants that target expression to hematopoietic stem cells. Mutagenesis of promoter 1b has previously demonstrated that the ϩ242 MAZ and ϩ264 ETS motifs were important for full activity of a reporter construct containing both promoter 1a and 1b in erythroid cells. However, it was not clear whether these effects were mediated by promoter 1a or promoter 1b (25).
We have now shown that the MAZ and ETS motifs were critical for activity of the core promoter 1b in primitive myeloid cells. Nevertheless, band shift analysis revealed the same pattern of complexes binding to the MAZ and ETS motifs in erythroid and primitive myeloid cells. The simplest interpretation of these results would be that MAZ and ETS sites are necessary but not sufficient for promoter 1b activity in primitive myeloid cells. Alternatively, subtle qualitative differences may exist between the comigrating complexes in erythroid and primitive myeloid cells. Thus, closely related transcription factors or post-translational modification may give rise to lineage-restricted functional differences that are not readily detectable by band shift analysis. Distinguishing between these possibilities may allow the construction of promoter 1b variants that, together with distal regulatory elements, will direct expression specifically to hematopoietic stem cells.