Transcriptional Regulation of the Stem Cell Leukemia Gene by PU.1 and Elf-1*

The SCL gene, also known astal-1, encodes a basic helix-loop-helix transcription factor that is pivotal for the normal development of all hematopoietic lineages. SCL is expressed in committed erythroid, mast, and megakaryocytic cells as well as in hematopoietic stem cells. Nothing is known about the regulation of SCL transcription in mast cells, and in other lineages GATA-1 is the only tissue-specific transcription factor recognized to regulate the SCL gene. We have therefore analyzed the molecular mechanisms underlyingSCL expression in mast cells. In this paper, we demonstrate that SCL promoter 1a was regulated by GATA-1 together with Sp1 and Sp3 in a manner similar to the situation in erythroid cells. However, SCL promoter 1b was strongly active in mast cells, in marked contrast to the situation in erythroid cells. Full activity of promoter 1b was dependent on ETS and Sp1/3 motifs. Transcription factors PU.1, Elf-1, Sp1, and Sp3 were all present in mast cell extracts, bound to promoter 1b and transactivated promoter 1b reporter constructs. These data provide the first evidence that theSCL gene is a direct target for PU.1, Elf-1, and Sp3.

Lineage commitment of eukaryotic cells involves a highly controlled developmental progression through successive stages of differentiation. Hematopoiesis, the process of blood cell formation, provides a powerful experimental system for studying this process. During blood cell formation, pluripotent hematopoietic stem cells give rise to a hierarchy of distinct progenitors that differentiate into at least eight phenotypically different mature blood cell lineages (1). Since lineage commitment and differentiation involve alterations in patterns of gene expression, the function and regulation of tissue-restricted transcription factors represents a major issue (2)(3)(4). For this reason, we have chosen to study the transcriptional regulation of the SCL gene (also known as tal-1), which encodes a lineagerestricted basic helix-loop-helix transcription factor pivotal for the regulation of hematopoiesis (5).
SCL was originally discovered by virtue of a chromosomal translocation associated with T cell acute lymphoblastic leukemia (6 -9). Rearrangements of the SCL locus are now realized to be the commonest molecular pathology associated with T cell acute lymphoblastic leukemia (10). SCL expression is restricted to hematopoietic tissues, endothelial cells, and distinct regions in the brain (11)(12)(13)(14)(15)(16)(17)(18). Within blood cells, SCL expression is exclusively found in committed erythroid, mast, and megakaryocytic cells as well as in interleukin-3-dependent progenitor and primitive CD34-positive cell lines (11, 14 -17, 19). In cell lines, induction of erythroid differentiation resulted in up-regulation of SCL mRNA, although protein levels decreased (12,20), whereas induced granulocyte/monocyte differentiation was accompanied by extinction of SCL expression (11). Indeed, down-regulation of SCL may be necessary for normal monocytic differentiation (21).
Gene targeting experiments have demonstrated that SCL is required for the development of all hematopoietic lineages in mice, thus suggesting a crucial role for SCL in early blood cell development (22,23). In addition, SCL null ES cells failed to give rise to any colony-forming cells in vitro and did not contribute to hematopoietic tissues in general when used in chimeric mice (24,25). Antisense and overexpression approaches have implied that SCL performs different functions in distinct hematopoietic cell types. The expression of antisense SCL constructs in a multipotent cell line resulted in reduced proliferation and self-renewal (26), whereas similar experiments inhibited erythroid differentiation of a committed erythroid cell line (27). Furthermore, experiments involving constitutive expression of normal, mutant, or truncated SCL polypeptides in myeloid progenitor cell lines suggested that SCL regulates proliferation, differentiation, and apoptosis (28). Inhibition of apoptosis may also underlie the leukemogenic effect of SCL expression in T cells (29). However, despite implicit involvement in several physiological processes, no target genes for SCL have yet been identified.
The complex pattern of SCL expression raises important questions about the molecular mechanisms responsible for the cell-specific expression of SCL in hematopoietic stem cells and in a subset of committed hematopoietic cell types. We have therefore chosen to study the regulation of the murine SCL genomic locus. Two proximal promoters have been identified in alternate exons (30), and several distant enhancers and silencers have also been characterized (31). Promoter 1a was regulated by GATA-1 in erythroid cells but was silent in myeloid cells lacking GATA-1. By contrast, promoter 1b was active in transient transfection assays only in primitive myeloid cells (32,33).
Mast cells play a central role in the response to allergens and parasites that induce the release of a variety of mast cellspecific mediators including cytokines and histamine (34,35). However, relatively little is known about the transcriptional program of mast cells, and only a few mast cell-specific regu-latory elements have been reported (36 -43). In this paper, we have characterized the molecular mechanisms regulating the SCL promoter region in mast cells. Our results show that promoter 1a was regulated by GATA-1, Sp1, and Sp3, similar to its regulation in erythroid cells. However, promoter 1b was strongly active in mast cells, in marked contrast to the situation in erythroid cells, and was regulated by PU.1 and Elf-1 in addition to Sp1 and Sp3. These results provide the first evidence that SCL is a direct target for PU.1, Elf-1, and Sp3.

Cell Lines
The murine mast cell line 6Ϫ is a factor-independent subclone of Fmpl.6 and has been previously described (19). The murine MST (96.42 B3) mast cell line and the murine J558 L plasma cell line have also been described previously (44,45

Alcian Blue/Safranin Histochemical Stain
After spinning the cells onto glass slides using a cytospin apparatus at 1000 rpm for 5 min, cells were air-dried and subsequently fixed for 3 h in Carmoy's solution (ethanol:chloroform:acetic acid, 6:3:1). After fixation, cells were washed in H 2 O, transferred for 5 min to 3% acetic acid, and stained with Alcian blue solution (1% Alcian blue (w/v) in 3% acetic acid). After the Alcian blue staining, cells were washed twice in 3% acetic acid and subsequently stained in safranin solution (1% safranin O (w/v) in 3% acetic acid) for exactly 1 min followed by three washes in H 2 O. All steps were performed at room temperature.

Plasmids
SCL promoter sequences in all constructs are from a genomic clone isolated from a Balb/C genomic library (30) and have been subcloned into the pGL-2 basic luciferase reporter vector (Promega) as described previously (32). The numeric nomenclature of the different constructs refers in all cases to the 5Ј-end of exon 1a (GenBank TM accession no. U05130; position 551). PEF-BOS lacZ contains the lacZ gene under the control of the pEF-1␣ promoter and was kindly provided by Dr. K. Chatterjee (Department of Medicine, Cambridge University, UK). Plasmids pPacSp1, pPacSp3 long, and pPacSp3 short and the p97b lacZ expression plasmid were provided by Dr. G. Suske (Institut fü r Molekularbiologie und Tumorforschung, Philipps Universitä t Marburg, Germany) and have been described previously (47,48). Expression vectors pPac-PU.1, pPac-ETS-1, and pPac-Elf-1 were a gift from Dr. N. Speck (Dartmouth Medical School Hannover, NH) and have been described (49).

Site-directed Mutagenesis
Site-directed mutants were created using the Kunkel method (50) or the protocol described by Wong and Komaromy (51). The following oligonucleotides were used as primers in the mutagenesis reaction (mutated residues are underlined, and the wild type binding consensus is shown in boldface type below the mutated residues).

Sequences 1-8
Mutants were generated on wild type SCL single-stranded templates, and all mutations were subsequently confirmed by nucleotide sequence analysis.

DNase I Footprinting
Nuclear protein was prepared exactly as described previously (52). Footprinting probes were prepared by 5Ј-end labeling of DNA restriction fragments with T4 polynucleotide kinase and [␥-32 P]ATP and isolated after secondary restriction (53). G ϩ A markers were prepared by chemical sequencing reactions (54). DNase I footprint protection assays were performed in a volume of 100 l of storage buffer in the presence of 6 g of poly(dI-dC), 2 ng of end-labeled restriction fragment, and up to 80 l (30 g) of nuclear protein extract. After partial DNase I digestion, the nucleic acid was purified and resolved by denaturing 6% polyacrylamide gel electrophoresis and exposed for autoradiography.

Transient Transfection of Cell Lines
Transfection of Mammalian Cells-Transient transfections of MST, 6Ϫ, and J558 L cells were performed as described (32) with the difference that during the electropulse, voltages of 240 V (for MST and 6Ϫ cell lines) and of 220 V (for J558 L cell line) were applied. In the transfection experiments, 10 g of pEF-BOS lacZ vector and 15 g of the appropriate luciferase reporter construct were used.
Transfection of Insect Cells-One day prior to transfection, Drosophila SL2 Schneider cells were seeded at 5 ϫ 10 6 cells/10-cm plastic Petri dish. The next day, the medium was changed, and cells were transformed using the calcium phosphate method. Briefly, 5 g of pPac expression plasmid or empty pPac vector were co-transfected with 7 g of the ϩ209 SCL 1b pGL-2 luciferase reporter plasmid or a ϩ209 SCL 1b mutant version. As an internal reference, 4 g of the Sp1-independent p79b lacZ plasmid was also co-transfected in each transformation. Cells were harvested 40 h post-transfection and assayed for luciferase and ␤-galactosidase activity.

␤-Galactosidase and Luciferase Assays
␤-Galactosidase and luciferase assays for mammalian and invertebrate cell lines were carried out as described (32). A linear relationship between light units and amount of luciferase reporter plasmid was confirmed in all cell lines. For each pulse the relative light units were normalized for transfection efficiency against ␤-galactosidase values obtained from the same sample. The relative light unit values presented are the mean of at least four independent experiments, and the same pattern was also obtained using a different DNA preparation.

Electrophoretic Mobility Shift Assays
Conditions for the band shift assays were as described (32). Oligonucleotides used in the band shift assays are depicted in Table I. The anti-GATA-1 antibody and the anti-Elf-1 antibody were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The PU.1 rabbit polyclonal antiserum (anti-PU.1 GST 1297) and the rabbit preimmune serum were generous gifts from Dr. R. Maki (La Jolla Cancer Research Foundation, San Diego, CA). Sp1-and Sp3-specific rabbit antisera were gifts from Dr. G. Suske (Institut fü r Molekularbiologie und Tumorforschung, Philipps Universitä t Marburg, Germany) and have been described (55).

Isolation of Murine Primary Mouse Cells
Primary murine mast cells were isolated as described previously (12,56). The final cell population was 89.3% positive for Fc⑀ receptor 1 ␣ by fluorescence-activated cell sorting analysis and over 90% Alcian bluepositive (data not shown). Fluorescence-activated cell sorting analysis was performed as described previously (56).

RNase Protection Assay
The RNase protection assays were performed as described previously (33) with the exception that the total cellular RNA was prepared using the TRIzol method according to the manufacturer's protocol (Life Technologies, Inc.), and 20 g of RNA was used for each protection.

Characterization of Mast Cell
Lines-Lineage-specific expression of the SCL gene has been studied in erythroid, early myeloid, and T cell lines (32,33,58). However, nothing is known about the molecular mechanisms underlying lineagerestricted expression of SCL in mast cells. Since it is difficult to obtain sufficient numbers of homogeneous normal mast cells, we have chosen to study transcriptional regulation of both SCL promoters in the MST and 6Ϫ murine mast cell lines. Northern blot analysis revealed the expression of SCL mRNA in MST and 6Ϫ cells and also the presence of GATA-1-and GATA-2specific transcripts (Fig. 1A). In addition, mRNA for the Fc⑀ receptor 1 ␣-chain receptor, which is known to be restricted to mast cells, basophils, and epidermal Langerhans cells (59), was detected in both cell lines. To further confirm the mast cell phenotype of MST and 6Ϫ cells, Alcian blue/safranin histochemical staining was performed (Fig. 1B). More than 95% of the MST and 6Ϫ cells contained Alcian blue positive granules, characteristic of mast cells (60). No Alcian blue staining was detected in the J558 L plasma cell line. These results therefore demonstrated that the MST and 6Ϫ cell lines available in our laboratory displayed several mast cell-restricted characteristics.
SCL Transcription in Mast Cells Was Initiated from Two Distinct Promoters-Luciferase reporter constructs were generated to determine SCL promoter usage in mast cells and to characterize mast cell-specific elements within SCL promoters 1a and 1b. Transient transfection of a luciferase reporter construct, which contained approximately 2 kilobase pairs upstream of exon 1a into either 6Ϫ or MST mast cell lines, resulted in 10 -30-fold stimulation when compared with values obtained using the empty luciferase vector ( Fig. 2A). Deletion of sequences upstream of position Ϫ187 (nucleotide numbering is relative to the transcriptional start site of exon 1a) did not decrease promoter 1a activity, but further deletion to Ϫ55 produced a marked fall in luciferase activity. By contrast, no transcriptional activation was detected when the same constructs were electroporated into the SCL-negative J558 L murine plasma cell line. These results demonstrate that promoter 1a could initiate transcription in mast cells and that a core promoter element spanning nucleotides Ϫ187 to ϩ26 was sufficient for this activity.
In order to study promoter 1b, a second series of deletion constructs were tested in 6Ϫ and MST cells. As shown in Fig.  2B, constructs containing both exon 1a and exon 1b with approximately 2 kilobase pairs upstream sequence were between 25-and 45-fold more active than the empty expression vector. Removal of 5Ј sequences including exon 1a and intronic sequence to position ϩ209 did not reduce luciferase activity. The same constructs produced only weak luciferase activity in J558 L plasma cells. These results show that a region from ϩ209 to ϩ353 was sufficient for activity of promoter 1b in mast cells.
Our data therefore demonstrated activity of both promoter 1a and promoter 1b in mast cell lines. The activity of both promoters was confirmed in splenic mast cells using an RNase protection assay to detect transcripts from both promoters (data not shown).
DNase I Footprinting Analysis of Promoter 1a and Promoter 1b-To identify transcription factor binding sites important for promoter 1a and promoter 1b function, we analyzed both core promoter regions by DNase I footprinting in vitro. Using MST nuclear extracts and a labeled sense strand, three different protected areas were identified in promoter 1a (Fig. 3A). Footprint A corresponds to a region between nucleotides Ϫ74 and Ϫ56 upstream of the transcription start site of exon 1a and contains a GATA and an Sp1 binding motif. Footprint B is located between positions Ϫ40 and Ϫ26 and contains a second GATA binding site. Footprint C overlaps the first exon and the transcription start site. The same protected areas were also observed with a labeled antisense strand (data not shown).
Within promoter 1b, two protected areas were detected using a labeled sense strand (Fig. 3B). Footprint D between position ϩ238 and ϩ249, contains a MYC-associated zinc finger protein binding site that also resembles an Sp1 binding consensus (55,61). The most striking feature within footprint E was a stretch of 20 purines containing two tandem GGAA ETS-binding motifs (ϩ265 dETS) and a single GGAA site (ϩ277 sETS) (62,63). The same footprinting pattern was also detected with a labeled antisense strand (data not shown). Fig. 3C summarizes the footprinting data for both promoters.
GATA-1 Regulated SCL 1a Promoter in Mast Cells-To further characterize specific transcription factors binding within footprints A and B, band shift assays were performed using either nuclear extracts from MST mast cells (SCL-positive) or from J558 L plasma cells (SCL-negative). Three complexes were detected following incubation of MST nuclear extracts with an oligonucleotide from footprint A (Fig. 4A, lane 2). Interestingly, incubation of the same oligonucleotide with J558 L nuclear extracts generated a different pattern of DNA-binding complexes (Fig. 4A, compare lanes 9 and 2). All shifts in both cell lines were specific, since an excess of the same oligonucleotide but not an unrelated oligonucleotide competed for binding (Fig. 4A, lanes 3, 4, 10, and 11). Incubation of Sp1 or Sp3 antiserum with MST and J558 L nuclear extracts produced a supershift in both cell lines (Fig. 4A, Sp1, lanes 5 and 12; Sp3, lanes 6 and 13; see also lanes 8 and 15). Furthermore, the addition of Sp3 antibody reduced the faster migrating complex in J558 L band shifts (Fig. 4A, compare lanes 12 and 13). The above findings demonstrated that both Sp1 and Sp3 could bind to the Sp1 consensus within footprint A and also identified a smaller isoform of Sp3 capable of forming a complex using J558 L extracts.
Since MST cells expressed GATA-1 and GATA-2 mRNA (Fig.  1A), it was possible that either factor could bind to the Ϫ69 GATA site within footprint A. The addition of a GATA-1-specific antibody identified GATA-1 as the factor binding to the Ϫ69 GATA motif (Fig. 4A, compare lanes 2 and 7). The more rapidly migrating complex, which supershifts with the GATA-1-specific antibody, is likely to reflect a proteolytic product of GATA-1 or the product of an alternatively AUG codon, as reported by Calligaris and colleagues (64). Incubation of an oligonucleotide corresponding to footprint B with nuclear extracts from MST cells generated a single shift (Fig. 4B, lane 2). The single shift obtained with MST extracts was specific and bound a GATA factor as shown by competition with excess cold probe (Fig. 4B, lane 3), a GATA consensus oligonucleotide (lane 4), a nonrelated oligonucleotide (lane 6), or an oligonucleotide containing a mutated GATA site (lane 5). Specific binding of GATA-1 to the oligonucleotide was demonstrated by incubation with an anti-GATA-1 antiserum, which abolished the complex and generated a supershift (Fig. 4B, lane  7, open triangle). No binding to this oligonucleotide was detected using J558 L nuclear extracts.
We then proceeded to test the functional significance of the GATA and Sp1/3 binding sites using transient reporter assays. Site-directed mutagenesis was used to individually abolish the Ϫ63 Sp1/3 motif, the Ϫ69 and Ϫ37 GATA binding sites, and also an Ap-1 binding consensus at position Ϫ101. As shown in Fig. 4C, mutation of the Ϫ101 Ap-1 and Ϫ69 GATA sites did not significantly reduce luciferase activity when compared with the wild type construct. Only a moderate reduction of activity was observed following mutation of the Ϫ63 Sp1/3 site. By contrast, mutation of the Ϫ37 GATA motif reduced promoter 1a activity significantly to about 50% of wild type activity in both mast cell lines. These data suggest that Sp1 and/or Sp3 and GATA-1 regulated promoter 1a activity in mast cells.
Sp1 and Sp3 Regulated SCL 1b Promoter-Gel shift analysis was performed to identify factors binding to the region of footprint D in promoter 1b (see Fig. 3C). Specific complexes were observed using both MST and J558 L nuclear extracts (Fig. 5A). Interestingly, MST nuclear extracts gave rise to additional complexes not detectable in J558 L extracts (Fig. 5A, compare  lanes 2 and 9). All complexes were competed by an oligonucleotide containing an Sp1 consensus (Fig. 5A, lanes 5 and 12), suggesting specific interaction with Sp1 family members. The addition of Sp1-specific antiserum to both nuclear extracts produced a supershift and resulted in the reduction of complex A (Fig. 5A, compare lane 2 with lane 6 and lane 9 with lane 13). In addition, incubation with Sp3-specific antiserum also generated a supershift, abolished complexes B and C, and reduced the intensity of complex A in MST cells (Fig. 5A, compare lanes  2 and 7). The same antiserum abolished complex B and reduced the intensity of complex A in J558 L cells (Fig. 5A, compare  lanes 9 and 14). Simultaneous incubation with both Sp1 and Sp3 antibodies abolished complexes A, B, and C in both cell lines (Fig. 5A, lanes 8 and 15).
These data demonstrate that both Sp1 and Sp3 can bind to the "Sp1 consensus" site in MST and J558 L cell lines. The faster migrating Sp3 form (complex C) was consistently ob-served using different preparations of MST nuclear extracts. It may represent the smaller Sp3 isoform already reported for stromal cells (65). However, we cannot exclude the possibility that it represents an MST-specific Sp3 breakdown product, although five different protease inhibitors were included during the preparation of nuclear extracts.
Site-directed mutagenesis was used to assess the functional significance of the Sp1/3 site in mast cells. Mutation of the ϩ239 CCCTCCC Sp1/3 motif drastically reduced luciferase reporter activity to near background levels, thus demonstrating that the Sp1/3 site was required for full SCL 1b promoter activity in mast cells (Fig. 5C).
We then proceeded to investigate whether Sp1 and Sp3 could directly transactivate promoter 1b in S2 Drosophila Schneider cells. As shown in Fig. 6A, both Sp1 and full-length Sp3 (97-kDa isoform) increased the activity of a promoter 1b luciferase construct about 5-6-fold, whereas the shorter Sp3 isoform (60 kDa) only generated a 2-fold increase in promoter 1b activity. Furthermore, mutation of the Sp1/3 motif resulted in a marked decrease in transactivation by Sp1 and both Sp3 isoforms (Fig.  6B). Previous experiments have demonstrated that the Sp1 and both Sp3 expression vectors produce similar levels of protein (55). Our data therefore demonstrate that both Sp1 and Sp3 are capable of regulating promoter 1b and that the shorter Sp3 polypeptide is a less powerful transactivator of this promoter.
PU.1 and Elf-1 Differentially Bound to Distinct ETS Motifs in Promoter 1b-Footprint E contained three GGAA ETS binding motifs (Fig. 3C). Furthermore, a good consensus binding site for the myeloid zinc finger transcription factor (66) was also detected at position ϩ251. To evaluate the functional importance of these binding sites for promoter 1b activity, a series of mutant luciferase reporter constructs were transiently transfected into MST and 6Ϫ mast cells. Abrogation of the ϩ251 myeloid zinc finger consensus did not alter the level of reporter activity (Fig. 5C). By contrast, mutation of the double ETS site at position ϩ265 reduced luciferase activity to about 25% in both mast cell lines, and mutation of the ϩ277 single ETS reduced activity by 50% or more (Fig. 5C). Thus, the results obtained in the transient transfection experiments demonstrated that both the single and the double ETS binding sites were required for normal SCL 1b promoter activity.
The ETS family of transcription factors contains at least 17 different members that are expressed in hematopoietic cells (67). Band shift assays were used to identify factors binding to the ETS motifs. Incubation of an oligonucleotide representing footprint E with nuclear extracts from MST and J558 L lines generated several shifts of different intensities (Fig. 5B, lanes  2 and 10). Incubation with excess unlabeled probe competed out all bands (Fig. 5B, lanes 3 and 11). Interestingly, competition with several nonrelated oligonucleotides enhanced com-

5Ј-AGGCGGCTCCTTATCTCGGC-3Ј
GATA cons We focused our initial attention on PU.1 and Elf-1, since PU.1 has been implicated in the regulation of a mast cellspecific enhancer (37), and Elf-1 is known to selectively interact with ETS sites containing GGAA but not with GGAT core binding motifs (68 -71). Preincubation of MST and J558 L nuclear extracts with normal rabbit antiserum enhanced some of the specific complexes in both cell lines but did not shift any of the bands (Fig. 5B, compare lane 2 with lane 7 and lane 10  with lane 15). The addition of anti-PU.1 antiserum generated a supershift (Fig. 5B, lanes 8 and 16, lower open arrowhead) and prevented the formation of complex C in MST and J558 L band shifts (Fig. 5B, compare lane 7 with lane 8 and lane 15 with  lane 16). The PU.1-specific band shift in MST migrated slightly faster than the J558 L counterpart (Fig. 5B, compare lane 7 with lane 15). The reason for this difference in mobility was not further investigated, but it is likely to reflect differences in phosphorylation and/or other post-transcriptional modification (72)(73)(74). Incubation with an anti-Elf-1 antiserum resulted in a faint supershift (Fig. 5B, lanes 9 and 17, upper open arrowhead) and abolished most of complex B in both cell lines (Fig.  5B, compare lane 7 with lane 9 and lane 15 with lane 17). These results demonstrate that both PU.1 and Elf-1 bound to sites within footprint E in MST and J558 L cells.
To look for differential binding of PU.1 or Elf-1 to the ϩ265 double ETS site relative to the single ETS site at ϩ277, we generated oligonucleotides containing a mutation at either site. An oligonucleotide in which the double ETS site was mutated (ϩ265 m dETS) competed for the PU.1 but less well for the Elf-1 complex (Fig. 5B, compare lane 4 with lane 5 and lane 12  with lane 13). By contrast, an oligonucleotide in which the single ETS motif was mutated (ϩ277 m sETS) competed equally for both the PU.1 and Elf-1 complex (Fig. 5B, compare PU.1 and Elf-1 but Not ETS-1 Can Transactivate SCL Promoter 1b-To obtain direct evidence that PU.1 and Elf-1 can regulate promoter 1b, we analyzed the potential of both factors to transactivate a promoter 1b reporter construct in S2 Schneider cells. The promoter 1b luciferase plasmid was co-transfected with PU.1, Elf-1, or ETS-1 expression constructs. These constructs have previously been shown to express fully functional polypeptides when used in Schneider cells (49). As shown in Fig. 7A, the expression of PU.1 or Elf-1 resulted in about 5or 4-fold activation, respectively, of promoter 1b activity. Interestingly, ETS-1, a related ETS family protein, did not transactivate the promoter. These results demonstrate that promoter 1b does not respond to all ETS proteins but only to a specific subset.
We then asked which ETS binding sites within promoter 1b were necessary for transactivation by PU.1 and Elf-1. Reporter constructs containing a mutated ϩ277 single ETS site or a mutated double ETS site at position ϩ265 were tested in tran-sient co-transfection experiments. As shown in Fig. 7B, transactivation by PU.1 was reduced by approximately 30% by mutating the single ETS site and by approximately 50% by mutating the double ETS site at position ϩ265. Transactivation by Elf-1 was not affected by mutating the single ETS site and was slightly reduced by mutating the double ETS site (Fig.  7B). These results are consistent with our band shift data, which suggested that PU.1 interacts with both the single and the double ETS motifs, whereas Elf-1 preferentially binds to the double ETS site. The fact that inactivation of the double ETS site produced only a moderate reduction in transactivation by Elf-1 is likely to reflect the presence of additional ETS sites in the reporter construct.
In summary, our results suggest that SCL promoter 1b is a direct target for regulation by both PU.1 and Elf-1. PU.1 appears to act through both the single and the double ETS motifs, whereas Elf-1 preferentially operates through the double ETS site.

DISCUSSION
Expression of the SCL gene during hematopoiesis is restricted to early progenitor cells as well as to committed eryth- FIG. 3. DNase I footprinting analysis of promoter 1a and promoter 1b. A, SCL promoter 1a footprint. SCL promoter 1a-specific fragment encompassing nucleotides Ϫ187 to ϩ22 was 32 P-labeled, incubated with MST nuclear extracts, treated with DNase I, and subsequently separated by denaturing gel electrophoresis (lanes on the right marked Bound). The G ϩ A track is a Maxam-Gilbert depurination of the same fragment. The lane marked Free shows a DNase I digestion of the same fragment in the absence of nuclear extracts. Protected regions are marked by black boxes. B, SCL promoter 1b footprint. SCL promoter 1b-specific fragment encompassing nucleotides Ϫ205 to ϩ351 was treated in the same way as described above. Protected areas are marked by black boxes. C, partial nucleotide sequence of SCL promoter 1a (upper panel) and SCL promoter 1b (lower panel). Footprinted areas are marked by a bar above the sequence, and potential transcription factor binding sites are shaded. Numbers refer to the 5Ј-end of exon 1a. Exons 1a and 1b are boxed. roid, mast, and megakaryocytic lineages. However, very little is known about the factors regulating SCL expression in different cell lineages. GATA-1 has been shown to regulate promoter 1a in erythroid cells (27,32,33,58), but no other lineage-restricted transcription factor has yet been implicated in the tissue-specific regulation of SCL. In this paper, we have characterized the mechanisms regulating SCL promoter activity in mast cells and have demonstrated for the first time regulation of SCL by two ETS family transcription factors, PU.1 and Elf-1.
Within the hematopoietic compartment, PU.1 is expressed in B lymphocytes, neutrophils, macrophages, early erythroblasts, mast cells, purified CD34-positive progenitors, and osteoclasts (37,(75)(76)(77)(78). SCL and PU.1 are therefore co-expressed in several hematopoietic cell types. PU.1 has been implicated in the regulation of various lymphoid-and myeloid-restricted promoters and/or enhancers including the mast cell-restricted interleukin-4 enhancer (37,79,80). However, to our knowledge, PU.1 has not been shown to directly regulate other cell-specific transcription factors. Our data strongly suggest direct regulation of SCL by PU.1 and therefore raise the possibility that PU.1 may also regulate SCL in other hematopoietic cell types in which the two transcription factors are co-expressed. Insight  Table I. GATA-1, Sp1, and Sp3 antiserum were included as shown. The filled arrows mark the four major complexes binding to the Ϫ69 Sp1/GATA oligonucleotide, and the open arrow shows specific supershifts. B, Ϫ37 GATA site. An oligonucleotide containing the Ϫ37 GATA site was used as a probe and incubated with nuclear extracts from MST and J558 L as shown. Unlabeled competitor oligonucleotides were included as indicated and are listed in Table I. The GATA-1-specific complex is marked by a filled arrow, and an open arrow indicates the GATA-1-specific supershift. GATA-1-specific antibody was included as indicated. C, activity of SCL promoter 1a mutants in 6Ϫ and MST cells. Transient transfections were performed using wild type SCL promoter constructs or the same constructs in which individual transcription factor binding sites were abolished by site-directed mutagenesis. Left, a schematic representation of the constructs indicating the position of each mutation. Luciferase values were calculated as in Fig. 2. into the relationship between SCL and PU.1 can be gleaned from knockout studies. Two independently derived mouse strains lacking PU.1 have been reported (81,82). Both PU.1 knockout strains exhibited yolk sac and fetal liver hematopoiesis but lacked macrophages, neutrophils, and B and T lymphocytes at birth. However, one group also reported the subsequent appearance of T cells and neutrophils after birth (81). By contrast, SCL null allelic mice completely lacked yolk sac hematopoiesis and died around embryonic day 9.5 (22,23). Moreover, in vitro differentiation assays and blastocyst reconstitution experiments demonstrated that SCL null ES cells failed to contribute to any hematopoietic lineage (24,25). Similar exper-iments using PU.1-negative ES cells showed a severe defect in the generation of lymphoid, myeloid macrophage, and osteoclast lineages but a normal contribution of PU.1-deficient ES cells to megakaryocytic and erythroid lineages (78,(83)(84)(85). These observations suggest that, although PU.1 may regulate SCL in cells committed to some lineages, PU.1 is not essential for SCL expression in multipotent hematopoietic stem cells during embryonic development.
We have also identified a second ETS family member, Elf-1, which regulates SCL transcription in mast cells. Elf-1 is expressed in several hematopoietic lineages (71). So far, Elf-1 has only been shown to regulate lymphoid genes (68, 70, 86 -90) FIG. 5. Band shift analysis of proteins binding to SCL promoter 1b and activity of promoter 1b mutants. A, nuclear extracts from MST mast cells and J558 L plasma cells were incubated with an oligonucleotide probe containing the ϩ239 Sp motif. Unlabeled competitor oligonucleotides (see Table I) and antiserum specific for Sp1 and Sp3 transcription factors were included as shown. Specific complexes A, B, and C containing Sp1 and Sp3 are marked by closed arrows. Sp1-and Sp3-specific supershifts are indicated by open arrows. B, an oligonucleotide containing the ϩ265 ETS site was used as a probe and incubated with nuclear extracts from MST and J558 L as shown. Unlabeled competitor oligonucleotides were included as indicated and are listed in Table I. Rabbit preimmune serum, PU.1, and Elf-1 antiserum were added as shown. Specific complexes A, B, and C are marked by filled arrows, and open arrows indicate specific supershifts. C, activity of SCL promoter 1b mutants in 6Ϫ and MST cells. Transient transfections were performed using wild type SCL promoter constructs or the same constructs in which individual transcription factor binding sites were abolished by site-directed mutagenesis. Left, a schematic representation of the constructs indicating the position of each mutation. Luciferase values were calculated as in Fig. 2. and human retroviruses (91,92). It is not yet known whether Elf-1 is expressed in hematopoietic stem cells, but our data raise the possibility that Elf-1 or other ETS factors may regulate SCL expression in hematopoietic stem cells.
Our finding that SCL promoter 1b is regulated by PU.1 and Elf-1 but not by ETS-1 demonstrated that this promoter does not respond to all ETS proteins but only to a specific subset. This is consistent with structure/function and binding site selection studies, showing that PU.1 and Elf-1 are distinct from other members of the ETS family in their DNA binding specificity (68 -71). It may also be relevant to note that both Elf-1 and PU.1 can physically interact with the retinoblastoma protein and that their function may be modulated during the cell cycle (93,94). Previous studies using cell lines have shown that SCL protein induces proliferation and/or blocks differentiation, raising the possibility that SCL function may also be linked to the cell cycle (26,28). It is therefore tempting to speculate that PU.1 and/or Elf-1 might regulate SCL expression in a cell cycle-dependent fashion. Further studies will be required to investigate this possibility.
We have previously shown that the ϩ239 CCCTCCC site in promoter 1b is functionally important (32,33). This site is identical to a MYC-associated zinc finger protein binding motif in the MYC promoter. However, our results now demonstrate that both Sp1 and Sp3 bind to this site and thereby transacti-vate promoter 1b. In addition, our data provide the first evidence that Sp3 as well as Sp1 is important for regulation of promoter 1a. These findings together with the results from promoter 1a highlight the importance of Sp1 and Sp3 for the regulation of both promoter 1a and promoter 1b.
Our data suggest that GATA-1 regulates SCL promoter 1a in mast cells and previous studies have reported a similar situation in erythroid cells (27,32,33,58). However, several lines of evidence suggest that GATA-1 is not essential for SCL expression in all hematopoietic lineages. First, GATA-1-deficient ES cells were able to differentiate into proerythroblasts, which expressed normal levels of SCL mRNA (95). Second, SCL was expressed in a CD34-positive cell line, which did not express GATA-1 mRNA (33). Third, in situ hybridization experiments showed that SCL expression in the yolk sac preceded the appearance of both GATA-1 and GATA-2 (18). We therefore postulate that GATA-1 acts to ensure maintenance or up-regulation of SCL expression in committed erythroid and mast cells but that during earlier stages of hematopoietic development, SCL expression may depend on different, as yet unidentified transcription factors.
In this paper, we have also studied the regulation of both SCL promoters in the J558 L SCL-negative plasma cell line. Comparison of factors binding to regulatory elements of promoter 1a and 1b in J558 L and MST cells reveals that Sp1/3, PU.1, and Elf-1 are equally expressed in both cell lines. Since we have also demonstrated that these transcription factors could directly transactivate SCL promoter 1b in heterologous FIG. 6. SCL promoter 1b is transactivated by Sp1 and Sp3 isoforms. A, expression vectors encoding Sp1, the full-length 97-kDa Sp3 (Sp3 long) or the smaller 60-kDa Sp3 isoform (Sp3 short) were co-transfected with a ϩ209 SCL 1b promoter luciferase construct into S2 Drosophila Schneider cells and assayed for luciferase reporter gene activity. Relative light units (RLU) represent luciferase activity (normalized against ␤-galactosidase values obtained by co-transfection of the p97b lacZ expression vector) relative to background luciferase activity obtained using the empty pPac expression plasmid. B, comparison of reporter gene activity of wild type ϩ209 SCL promoter 1b luciferase constructs and reporter constructs containing a mutated ϩ239 Sp motif after transactivation with Sp1/3 factors. Expression vectors encoding Sp1, the full-length 97-kDa Sp3, or the smaller 60-kDa Sp3 isoform were co-transfected with the wild type ϩ209 SCL 1b promoter luciferase construct or the ϩ239 mutant version into S2 Drosophila Schneider cells and assayed for luciferase reporter gene activity as in A. Reporter gene activity obtained with the wild type is represented as 100%. Expression vectors and luciferase reporter constructs are indicated. FIG. 7. PU.1 and Elf-1 can transactivate SCL promoter 1b. A, ETS-1-, PU.1-, and Elf-1-specific expression vectors were co-transfected with a ϩ209 SCL 1b core promoter luciferase plasmid into S2 Drosophila Schneider cells and assayed for luciferase reporter gene activity as in A. B, comparison of reporter gene activity of wild type ϩ209 SCL promoter 1b luciferase constructs with reporter constructs containing a mutated single ETS site (sETS mut) or a mutated double ETS site (dETS mut) after co-transfection with PU.1 (left panel) or Elf-1 (right panel). Relative activity obtained with the wild type is represented as 100%. Expression vectors and luciferase reporter constructs are indicated. cell lines, this finding raises an important question: why do these factors not transactivate SCL expression in J558 L plasma cells? Several explanations are possible. First, the observation that transcription factors Sp1/3, PU.1, and Elf-1 fail to transactivate SCL promoters in J558 L cells may reflect the absence of tissue-specific factors and/or co-factors that are missing in J558 L cells but are present in SCL-expressing lineages. Second, post-translational modifications may account for the inability of PU.1, Elf-1, and Sp1/3 to either directly stimulate transcription and/or to recruit other factors, necessary for transcriptional activation of SCL promoters. In this respect, it is tempting to speculate that the different electrophoretic mobility of PU.1, observed in J558 L and MST nuclear extracts, is due to post-translational modifications that are able to modify PU.1 function. However, additional studies will need to be performed to investigate these possibilities.
Finally, our data raise the issue of how the SCL promoters interact with SCL enhancers to generate the cell-specific pattern of SCL expression. The two SCL promoters clearly exhibit lineage-restricted activity. However, they are insufficient to drive reporter gene expression following integration into chromatin (33), and for this they require the presence of one or more enhancers (31). It will now be important to elucidate the molecular mechanisms governing this lineage-specific cross-talk between SCL promoters and enhancers and also to dissect the relative contribution of these elements to SCL expression in vivo.