Transcription Factor SCL Is Required for c-kit Expression and c-Kit Function in Hemopoietic Cells

The TF-1 cell line (40) was a gift from Dr. T. Kitamura (DNAX, Palo Alto, CA). The cells were maintained in IMDM (GIBCO BRL, Gaithersburg, MD) supplemented with 10% FCS (GIBCO BRL) and 5 ng/ml human GM-CSF. The cells were passaged every second day at 1.5 × 10⁵/ml. Jurkat T cells were also maintained in IMDM supplemented with FCS (10%), and were passaged three times weekly at a concentration of 1.5 × 10⁵/ml.

Purified recombinant human GM-CSF was a gift from Dr. Steve Clark (Genetics Institute, Cambridge, MA) and purified recombinant SF was from Dr. K. Langley (Amgen, Thousand Oaks, CA). The monoclonal antibody 2TL136 specific for human SCL was provided by Dr. Danièle Mathieu-Mahul (INSERM, Marseille, France; references 10, 42). The monoclonal antibody ACK2 is specific for murine c-Kit (GIBCO BRL). The monoclonal mouse anti–human c-Kit was from Cedarlane Labs. (Hornby, Ontario, Canada). The monoclonal mouse anti–rabbit eIF-4E cross-reacts with the human protein (Transduction Labs., Lexington, KY). The monoclonal anti-CD45 (clone 4B2; ATCC number HB 196, American Type Culture Collection, Rockville, Maryland) was used at 1:10 dilution of hybridoma supernatant. Monoclonal anti–hGM-CSF receptor β chain was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY) and used at a 1:100 dilution. Annexin V was purchased from Biodesign (Kennebunk, ME), and used at a concentration of 400 mg/ml.

The human SCL cDNA, the dn-SCL devoid of DNA binding domain (12) and the as-SCL were all cloned in the EcoRI site of the murine stem cell virus (MSCV)-neo vector (43). High titer amphotropic viruses (10⁶ PFU/ml) were produced by transient transfection into BING cells (44). The cells were irradiated and used for coculture with TF-1 cells for 48 h. Murine c-kit cloned in the LXSN retroviral vector was a gift from Dr. A. Bernstein (Samuel Lunenfeld Research Institute, Toronto, Ontario, Canada). Ecotropic viruses were produced by transfection into the BOSC23 packaging cell line (44).

For retroviral infection, 10⁶ exponentially growing cells (TF-1 or A31) were presensitized with polybrene at 2 μg/ml for 24 h and cocultured with virus-producing clones for another 24 h. Nonadherent TF-1 cells were separated from the infected fibroblasts. A polyclonal population was analyzed 7 d after selection in G418 at 1 mg/ml. For long-term experiments, cloning was performed through limiting dilution (12).

The human SCL cDNA was cloned in the antisense orientation under the control of the metallothionein promoter by digesting the plasmid pMSCL (12) with EcoRI and religation. The EcoRI sites come from the plasmid polylinker.

Stable TF-1 transfectants harboring the as-SCL were obtained through Lipofectin-mediated DNA transfer (GIBCL BRL) as previously described (9, 12). Cells were cloned immediately after gene transfer by limiting dilution, and the selective pressure was applied the following day, at a concentration of 1 mg/ml G418 (GIBCO BRL). GM-CSF was present throughout the gene transfer procedure. After selection, G418 concentrations were lowered to 400 μg/ml in order to expand the cells and prepare a large stock of frozen cells immediately after characterization. Each cell line was kept for not more than 2 mo in culture.

3 × 10⁵ TF-1 and as-SCL transfectants were fixed in 500 μl of Bouin's fixative for 15 min at room temperature. The cells were then pelleted at 200 g for 5 min and resuspended in 200 μl of 0.2% Triton X-100, followed immediately with 800 μl of PBS. The cells were pelleted again, washed with 1 ml of PBS, and resuspended in 100 μl of 1:10 dilution of anti-SCL monoclonal antibody (clone BTL73, provided by Dr. Danièle Mathieu-Mahul). After 30 min of incubation on ice, cells were washed twice with 1 ml of PBS followed by a 10-min wash in 1 ml of PBS, resuspended in 100 μl of FITC-coupled goat anti–mouse antibody at the recommended dilution (Caltech, San Francisco, CA), and incubated for 30 min at 4°C. The cells were washed as above before flow cytometry analysis.

For surface marker staining, 3–5 × 10⁵ cells were washed once with PBS supplemented with 2% FBS and 0.05% sodium azide (staining buffer), labeled with the primary antibody for 30 min on ice in a total volume of 100 ml, washed three times with 1 ml of staining buffer, and labeled with a biotinylated goat anti–mouse antibody (Cedarlane, Labs.) in staining buffer supplemented with 1% normal goat serum (Sigma-Aldrich, Oakville, Ontario, Canada) for 30 min on ice. After washing, the cells were incubated with PE-conjugated streptavidin (Sigma) for an additional 30 min on ice, washed, and analyzed on the FACScan^® flow cytometer (Becton Dickinson, San Jose, CA).

Cells were homogenized in the presence of guanidium and RNA isolated by acid-phenol extraction as previously described (9, 45). Northern blotting and hybridization were performed as previously described (9). Probes used were the 1.2-kb Hind–XbaI human SCL cDNA fragment (nucleotides 109–1293; reference 12), the 1.4-kb SalI–XbaI fragment of human GATA-1 cDNA (26), a 1.5- kb DraI–KpnI human c-kit cDNA fragment from human c-kit–BS plasmid (gift from Dr. J. Simms, Immunex, Seattle, WA), and the 1.2-kb PstI– XbaI fragment of the rat glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA (46). Blots were exposed to a PhosphorImager screen for quantitation.

Cell lysates were prepared as described previously (42). Protein concentrations were determined with the Bio-Rad Protein Assay reagent (Bio-Rad, Hercules, CA). Equal amounts of proteins (20 μg) were loaded on a 12% SDS-polyacrylamide gel and transferred to nitrocellulose using the mini “Transblot” (Bio-Rad) for 1 h at 100 V.

Membranes were blocked in 5% nonfat dry milk and 1% BSA, incubated with 2TL13G hybridoma tissue culture medium (monoclonal anti-SCL) diluted 1:300 for 2.5 h at room temperature, washed, incubated with a goat anti–mouse alkaline phosphatase– linked antibody (Bio-Rad), diluted 1:1,000 for 1.5 h, and washed extensively. Western blots were then developed by incubating membranes in the dark with 10 ml of BCIP (165 μg/ml) and NBT (330 μg/ml) (GIBCO BRL) for 2 min, and the reaction was terminated by several washes with water as described previously (42). In parallel, Western blotting was performed with a monoclonal antibody against the eukaryotic translation initiation factor (eIF-4E) (Transduction Labs.) as a control for protein loading.

Nuclear extracts were prepared as previously described (47). Protein concentrations were assayed with the Bio-Rad Protein Assay reagent. Binding reactions for electrophoretic mobility shift assays were performed as previously described (42). In brief, the binding reactions were allowed to proceed at room temperature for 15 min in the presence of 0.5 μg of poly(dI-dC) as nonspecific competitor DNA in 20 mM Hepes (pH 7.5), 50 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, 10 μg BSA, 100,000 cpm double-stranded synthetic oligonucleotide, and 25 μg nuclear extract in a total volume of 10 μl. 50-fold molar excess of unlabeled oligonucleotides were used for self-competition and added before nuclear extracts. For antibody supershift assays, 200 ng of affinity-purified anti-E2A (Santa Cruz Biotechnology, Santa Cruz, CA), 1 μl of the monoclonal anti-SCL antibody (BTL73; reference 48), or 1 μl of an unrelated antibody were added to the binding reaction. Protein complexes were resolved by electrophoresis on 4% polyacrylamide gel buffered in 0.25× Tris-borate– EDTA, 195 mM glycine (pH 8.5) at 4°C. The sequences of the probes (coding strand) are: ACCTGAA CAGATG GTCGGCT TAL1 consensus (21); CTAGGGAG CACCTG CCAGGTG GCTGGCCC murine c-kit probe (49); and an E box probe derived from the rat POMC E box also referred to as DE2c (50).

The human c-kit promoter was cloned by PCR from genomic DNA to generate a fragment that extends from 1146 bp upstream of the transcription initiation site to 43 bp downstream. In brief, genomic DNA was amplified in two steps, the first PCR providing a fragment that extends from position −634 to +60 (PCR1), and the second PCR a fragment that covers −1146 to −379 (PCR2). PCR1 was cut with SalI and BamHI and PCR2 with HindIII (site added) and SalI, and the fragments were cloned into the HindIII and BglII sites of the promoterless luciferase expression vector pXPII to generate kit 1146. The chimeric construct was verified through sequencing. TF-1 and A31 cell lines were transfected by electroporation. Cells were passaged 24 h before gene transfer at a concentration of 3 × 10⁵ cells/ml. Exponentially growing cells were then concentrated at 2.5 × 10⁷ cells/ml and electroporated at 900 μF and 350 mV using a Bio-Rad electroporator with 12 μg of reporter DNA and 1.5 μg of CMV β-galactosidase used as an internal control for the experiment, with or without different molar ratios of MSCV-SCL as shown; the total amount of transfected DNA was then filled to 25 μg with pGEM4 as a carrier. Cell lysates were prepared 15 h after transfection, normalized for β-galactosidase content, and assayed for luciferase activity. Rous Sarcoma virus–driven luciferase (RSV-luc) was used as an external control for all the transfections and pXP2 as a negative control as described previously (47).

Analysis of DNA fragmentation by agarose gel electrophoresis was performed as previously described (28). In brief, equal cell numbers were lysed in Sarkosyl buffer and treated with Proteinase K and RNase A before loading on a 1.2% agarose gel (20 cm), followed by a 16-h electrophoresis at 40 V. Cell viability was independently confirmed through trypan blue exclusion and counting (28, 29), or the double fluorochrome staining assay, as described below. Apoptosis was also assessed by flow cytometry analysis of cells labeled with Annexin V-FITC (1 μg/ml), shown previously to be an early marker of apoptosis (51, 52). Immediately before acquisition, 1 ml of propidium iodide solution (1 mg/ml) was added in order to detect dead cells.

A stock solution of dye mix containing 100 μg/ml acridine orange and 100 μg/ml ethidium bromide was prepared in PBS. Cell were resuspended at 5 × 10⁵ to 5 × 10⁶ in IMDM. 1 μl of dye mix was added to 25 μl of cell suspension, and 10 μl of the mixture was placed on a microscope slide for examination by fluorescent microscopy (Leica, Wetzlar, Germany). Cells with bright green chromatin were considered viable cells while those with bright orange chromatin or collapsed chromatin were considered dead cells. A minimum of 200 cells were counted.

To address SCL function and define SCL target genes in primitive hemopoietic cells, we chose to disrupt SCL function in the CD34⁺/c-kit⁺ cell line TF-1. To this end, we took two complementary approaches. The first one consists of expressing an as-SCL that interferes with SCL protein levels, and the second one involves a dn-SCL devoid of its DNA binding domain (12), which prevents SCL function without affecting protein levels. We reasoned that loss of SCL function may be detrimental to the cells and therefore decided to analyze the cells as early as 1 wk after gene transfer, at a time when the selective pressure (G418) is already optimal. To minimize cell loss during the selective pressure and, consequently, the generation time required for in vitro expansion, we chose retroviral infection as a method for high efficiency gene delivery for which we typically obtain 30– 50% transfer efficiency.

We first determined the efficiency of as-SCL in disrupting SCL protein levels, which were assessed through immunostaining of permeabilized cells with the monoclonal anti-TAL1 antibody BTL73 (Fig. 1). As expected, TF-1 cells express high levels of endogenous SCL. In contrast, 1 wk after transduction of as-SCL the endogenous SCL protein level was significantly decreased, with little if any overlap between the two populations. The cells were further analyzed for expression of several cell surface markers. Data shown in Fig. 1 indicate that the fluorescence intensity for c-Kit (CD117) was at least three- to fourfold lower in dn-SCL and as-SCL transfectants when compared with parental TF-1 cells (data not shown) or control cells expressing the vector alone (MSCV ). In contrast, the level of CD45 was not affected by as-SCL or dn-SCL expression and that of CD116 (GM-CSF receptor α chain) was either similar or slightly higher. Together, the results indicate that decreased SCL protein levels (as-SCL) or decreased SCL function (dn-SCL) specifically lower c-Kit levels without affecting other surface markers.

Since the GM-CSF receptor is not affected by SCL levels, we proceeded to derive stable TF-1 cell lines expressing dn-SCL (TF1-dn) following long term selection in G418 in the presence of GM-CSF as a survival factor, in order to directly address SCL function at the molecular and cellular levels. We also established clones that express as-SCL driven by a weaker promoter, the mouse metallothionein promoter, instead of the strong promoter contained within the retroviral LTR. Despite this, most of the 50 clones screened displayed weak levels of transgene expression, consistent with the reported difficulty in obtaining stable as-SCL clones in another hemopoietic cell line, K562 (9, 53; and Kirsch, I.R., unpublished data). The two highest antisense expressor clones, A30 and A31, were selected.

We next confirmed that SCL DNA binding activity was indeed reduced in the antisense clone A31 and the dn-SCL–expressing transfectants (TF1-dn) through electrophoretic mobility shift assays using the TAL1 consensus probe (reference 21; Fig. 2). TF-1 nuclear extracts produced four specific complexes (C1–C4), as shown by self-competition (Fig. 2, lanes 11 and 12). In contrast, a probe with divergent sequences (mDE2C) did not compete for binding (data not shown), indicating that these complexes were specific. All four complexes were also found using nuclear extracts from Jurkat cells (Fig. 2, lanes 9 and 10) and TF-1 transfectants harboring the vector alone (TF1-neo, lanes 1 and 6). The identity of each complex (C1–C4) was verified by antibody supershifting. Preincubation of TF1-neo extracts with either anti-SCL or anti-E2A antibody supershifted the two most slowly migrating complexes, C1 and C2 (Fig. 2, lanes 1, 2, and 6–8), indicating that these contain SCL/E2A heterodimers as previously described (21, 24). These complexes were absent in HL-60 cells, which were included as a negative control (lane 16). In the antisense-expressing clone A31, the SCL-containing complex C1 decreased significantly, whereas the C2 complex was only marginally affected (Fig. 2, compare lanes 3 and 5), suggesting that under conditions where SCL is limiting, the smaller complex, C2, is preferentially formed over C1. Rescuing clone A31 wild-type SCL (A31-SCL) restored C1 and C2 binding to the TAL1 probe (Fig. 2, lanes 3 and 4). In all cell lines, the two fastest migrating complexes, C3 and C4, were not supershifted with anti-SCL or anti-E2A, suggesting that these did not contain the corresponding proteins. These complexes were not affected in TF1-neo but were decreased in A31, possibly as a consequence of clonal variation, since they were also decreased in A31-SCL, a subclone of A31.

In the transfectant expressing a dn-SCL, which is a polyclonal population, both SCL-containing complexes, C1 and C2, were significantly decreased relative to TF1-neo whereas the non-SCL-containing complexes C3 and C4 were not affected (Fig. 2, compare lanes 5 and 6). Together, our results indicate that ectopic expression of as-SCL or dn-SCL significantly interferes with the formation of the two SCL-containing complexes, C1 and C2, on the TAL1 probe.

As observed with transient transfectants (Fig. 1), flow cytometry analysis of the A31 and TF-1 cells (Fig. 3,A) or TF1-dn and TF1-neo (data not shown) using a monoclonal antibody to human c-Kit indicated a quantitative decrease in c-Kit levels in A31 and TF1-dn as compared with control cells. There was a direct correlation with c-kit mRNA levels as determined through Northern blotting (Fig. 3,B). Thus, parental TF-1 cells or TF-1 cells harboring the vector alone (TF1-pac and TF1-neo) exhibited comparable levels of c-kit mRNA (Fig. 3,B). In contrast, c-kit expression was significantly lower in A31 and T-dn (Fig. 3, B and C). To exclude the possibility that c-kit mRNA expression was affected by the site(s) of integration in these antisense clones, we attempted to rescue c-kit expression by wild-type SCL through retrovirus-mediated gene transfer (MSCV; reference 43), using puromycin resistance as a second selective marker. Expression of the puromycin resistance gene alone did not affect c-kit expression in TF-1 cells (Fig. 3,B, TF1-pac) nor in A31 cells (data not shown). In contrast, elevating SCL expression in clone A31 restored c-kit expression (Fig. 3 B, A31-SCL) and further increased c-kit mRNA levels in parental TF-1 cells (data not shown). Together, the results indicate that the targeted attenuation of SCL protein levels and DNA binding activity directly leads to decreased c-kit expression.

Sequence analysis of the c-kit promoter revealed two adjacent E boxes at −374 and −381 that are conserved between human and mouse (49), and fit the core binding motif of MyoD/E2A. Electrophoretic mobility shift assays indicate that SCL–E2A complexes bind these two E boxes (our unpublished results), suggesting that these E boxes may contribute to c-kit promoter activity in response to SCL. We directly tested SCL transactivation properties on a chimeric construct in which the luciferase reporter gene was placed under the control of the human c-kit promoter (kit1146). Since the endogenous levels of SCL are elevated in TF-1 cells, we used TF1-dn for transient transfection. The reporter construct kit1146 was codelivered with varying amounts of the SCL expression vector MSCV-SCL. Results shown in Fig. 3 D indicate that SCL induced a dose-dependent increase in c-kit promoter activity, which was twofold at an optimum ratio of activator to reporter. Conversely, cotransfection of dn-SCL and kit1146 in parental TF-1 cells caused a twofold decrease in luciferase activity (data not shown). Together, these results are consistent with the hypothesis that SCL positively regulates c-kit transcription through binding to two adjacent E boxes in the c-kit promoter.

We have previously shown that both SF and GM-CSF suppress apoptosis in TF-1 cells. Parental cells and the different transfectants were, therefore, compared for their survival in response to optimal concentrations of SF, GM-CSF, or IL-3. Apoptotic death was revealed by the presence of a DNA ladder after agarose gel electrophoresis (data not shown). In the absence of growth factor, there was an intense DNA ladder, suggesting that the cells underwent apoptosis. At a concentration of 100 pM of SF, chosen to be near maximal suppression of apoptosis, parental TF-1 cells and the control TF1-neo line behaved similarly, with no detectable DNA degradation. Under identical conditions, apoptosis was evident in both antisense clones, A30 and A31 (data not shown), in which SCL protein levels (data not shown) and DNA binding activity (Fig. 2) were decreased. Cell viability was restored in clone A31-SCL, consistent with increased SCL DNA binding activity (Fig. 2, lane 4) and SCL protein levels (data not shown) in this clone, whereas control cells expressing the vector alone (A31-pac) behaved like the parental A31 cell line.

Cell viability was therefore quantitated using the double fluorochrome staining technique (Fig. 3,B). Full dose– response curves for SF and GM-CSF were performed in the two antisense clones, A30 and A31, as well as in control cells. Data are shown for growth factor concentrations that provide 80% survival in TF-1 cells (Fig. 3 B). As observed with agarose electrophoresis, TF-1 and TF1-neo lines remained viable with SF, whereas A30 and A31 underwent apoptosis. As above, ectopic SCL expression in clone A31 (A31-SCL) restored cell viability in SF-containing cultures, indicating that apoptotic death was due to as-SCL. In contrast, cell survival was the same in control and antisense clones in GM-CSF or IL-3 stimulated cultures (data not shown).

Apoptotic cells were also detected through Annexin V staining of membrane phosphatidyl serine, which is exteriorized during apoptotic death, both in stable clones (data not shown) and in transient transfectants 1 wk after G418 selection. Results shown in Fig. 4 indicate that parental cells and the different transfectants survive well in GM-CSF–containing cultures, consistent with their staining pattern for GM-CSF receptor (GM-R) observed in Fig. 1. In contrast, when the cells were maintained with SF, 80% of the population expressing dn-SCL or as-SCL was apoptotic, whereas both controls (TF-1 and MSCV transfectants) survived readily. Impaired survival in response to SF stimulation in these transfectants is directly correlated with decreased surface c-Kit expression as shown in Fig. 1. Together, the results suggest that SCL specifically regulates c-Kit expression and cell survival in response to SF.

As discussed above, the A31 subclone in which endogenous c-kit levels were restored after ectopic SCL expression, A31-SCL, survived well in cultures stimulated with SF. We, therefore, addressed the possibility whether c-kit expression by itself was sufficient to rescue SF-dependent cell viability. Murine c-kit was delivered to A31 cells through retrovirus-mediated gene transfer, using the LXSN retroviral vector. Despite a high homology between murine and human c-kit, ligand binding is species specific such that cells expressing human c-Kit require five- to eightfold higher concentrations of murine SF as compared with human SF (Fig. 5,C). After a 2-d selection in low concentrations of murine SF, c-kit–infected A31 and TF-1 cells (A31-kit and TF1-kit, respectively) were expanded with human GM-CSF. Flow cytometry analysis indicated that >75% of TF1-kit or A31-kit cells express murine Kit (Fig. 5,A). In GM-CSF–containing cultures, all clones behaved similarly (data not shown). As expected, TF1-kit responded better than parental TF-1 cells to murine SF (Fig. 5 C). As with human SF, the response of A31 cells to murine SF was significantly lower than that of TF-1 cells, even at saturating concentrations of murine SF. In contrast, the response of A31-kit to murine SF was similar to that of TF1-kit, suggesting that c-kit overexpression was sufficient to overcome the effect of reduced SCL expression. These observations are compatible with the view that the major survival function of SCL in TF-1 cells is mediated through c-kit expression.

Ba/F3 is a pro–B cell line that expresses the IL-3 receptor but not c-Kit (54, 55) or SCL (data not shown). Therefore, we transduced the human SCL cDNA into Ba/F3 cells through retroviral infection in order to address the possibility that SCL can induce c-kit expression in a negative cell line. Control cells were transduced with the MSCV vector alone. 2 wk after selection in G418, polyclonal populations were analyzed for surface expression of murine c-Kit by flow cytometry (Fig. 6 A). As expected, Ba/F3-MSCV cells were c-Kit negative. In contrast, Ba/F3-SCL cells were clearly c-Kit positive.

We next determined whether the cells have acquired responsiveness to murine SF (Fig. 6 B). In the absence of growth factor, Ba/F3-SCL transfectants continuously die over a 3-d period, with the same kinetics as parental Ba/F3 or MSCV transfectants maintained with or without SF. In contrast, when Ba/F3-SCL transfectants were maintained with SF, cell numbers remained at input levels for 2 d and declined slightly on day 3, consistent with the presence of surface c-Kit on these cells. Together, our observations indicate that enforced SCL expression induces c-Kit and confers SF responsiveness in Ba/F3 cells.

This study provides evidence for a mechanism whereby SCL secures the survival of primitive hemopoietic cells through upregulation of the tyrosine kinase receptor c-Kit, which is itself essential for the generation of hemopoietic cells both in vivo and in vitro (17–20). Several lines of evidence indicate a direct link between SCL and cell survival in response to SF. First, the apoptotic phenotype is induced by both as-SCL and dn-SCL. Second, apoptosis in the antisense clone A31 is rescued by ectopic SCL expression in the sense orientation. Third, apoptosis is specific to the c-Kit pathway triggered by SF, and is not observed with GM-CSF or IL-3, which indicates exquisite biological specificity. Indeed, SF is essential for the normal development of hemopoietic cells, whereas IL-3 and GM-CSF are produced by activated T cells as part of an inflammatory/stress response, suggesting that SCL function is more important in steady-state hemopoiesis. Finally, enforced SCL expression in the pro–B cell line Ba/F3, which is c-Kit⁻, clearly induces c-Kit expression in these cells. In all, our study provides a direct link between SCL and the survival response of hemopoietic cells to SF, mostly through upregulation of c-kit transcription.

The conservation of two everted E boxes at positions −381 and −374 between the human and mouse promoter sequences (49) suggests a functional role in transcription regulation, which was confirmed through binding studies (data not shown) and promoter deletion analysis. Our gel shift assays indicate that SCL-containing complexes that are formed on the c-kit probe also include E2A, consistent with the view that SCL does not bind DNA on its own and requires interaction with E12, E47, or HEB, another ubiquitous basic helix-loop-helix (21, 24). Furthermore, SCL is also found in association with the LIM-only protein Rbtn2/Lmo2 (22, 23) in a multiprotein complex with GATA-1 and a LIM-binding protein, Ldb1 (24). An E box–GATA motif was identified through in vitro binding site selection with nuclear extracts instead of isolated proteins (24), which corresponds strikingly to the c-kit everted E box sequences at −381. Sequence analysis of the human c-kit promoter also reveals the presence of several potential GATA sites at −900. Although these are found upstream instead of downstream of the E box as in the CASTing experiment, it is possible that a looping mechanism favored by protein–protein association allows for optimal spatial arrangements such as those defined by CASTing.

Transactivation assays indicate that SCL represses the E2A-dependent activity of a promoter construct containing the TAL1 motif. It is only in the presence of Id that SCL was shown to relieve the inhibition conferred by Id1 on E2A (7). In addition, despite its high affinity binding to the Ebox-GATA motif, the SCL–E2A heterodimer has no activity on a reporter construct with a minimal promoter linked to two such motives, either in the absence or presence of Lmo2. This multiprotein complex, however, provides a twofold enhancement of luciferase activity driven by GATA-1 on the same construct (24). Thus, SCL appears to be a weak transactivator, consistent with our results indicating a twofold contribution to the activity of the proximal c-kit promoter in TF-1 cells. This twofold contribution was confirmed by three distinct approaches, Northern blotting for steady state c-kit mRNA levels, flow cytometry analysis of surface c-Kit protein, and analysis of the c-kit promoter driving the luciferase reporter gene. A recent report using semiquantitative reverse transcriptase PCR did not detect significant differences in c-kit mRNA levels in embryoid bodies grown from scl^−/− embryonic stem and wild-type embryonic stem cells (56). However, it should be noted that flow cytometry analysis and luciferase activity can quantitatively reveal a two- to fourfold difference, whereas reverse transcriptase PCR remains semiquantitative and requires much larger differences for detection.

It was previously shown that loss of SCL function is associated with premature apoptosis upon nutrient deprivation in the lymphoid cell line Jurkat (42) and decreased colony formation in the erythroid cell line K562 (53), suggesting a role for SCL in mediating cell survival. As observed with the antiapoptotic gene bcl-2, which infrequently induces lymphomas in transgenic mice after a long latency period of 18 mo (57), elevating SCL expression in thymocytes is not sufficient to cause tumors in transgenic mice (58). Directing SCL expression in the thymus can nonetheless shorten the time of appearance of T cell tumors in LMO-2 (59), LMO-1 (58), or casein kinase II transgenic mice (48). Although these observations suggest a role for SCL in leukemogenesis, its target genes remain unknown.

The cells that repopulate the thymus and a subpopulation of triple negative thymocytes are c-Kit⁺ (37). Interestingly, SF is expressed by fetal thymic stromal cells and epithelial cells (38, 39, 60). More importantly, T cell differentiation in thymic lobe reconstitution with fetal liver precursors is inhibited by anti-Kit (37) and is impaired in white spotting mice, indicating the functional importance of c-Kit (61). In parallel, the presence of SCL mRNA in a subset of thymocytes (11) and the absolute requirement in SCL for the generation of T cells in vivo (19, 20) are indicative of the importance of SCL for thymocyte development. Our observations suggest that SCL may also specify c-Kit expression in primitive thymocytes to sustain cell survival in response to SF. Consistent with this, elevating SCL levels in thymocytes in double CD2-SCL and CD2-Lmo2 transgenics results in a twofold expansion of the double negative CD4⁻/CD8⁻ thymocyte population relative to CD2-Lmo2 transgenics (48), which, in light of our results, may be due to an increase in c-Kit⁺ subpopulations. We propose that constitutive SCL expression in the T cell lineage caused by chromosomal rearrangements results in constitutive c-kit expression in primitive thymocytes and prolongs their survival, which may represent an initiating event in T cell ALL. Secondary events are probably involved in the full development of T cell leukemia (58, 62), a mechanism commonly observed in tumor biology (63).

The regulation of cell survival is central to normal developmental processes and stress response. It is widely accepted that survival cues are crucial for the differentiation process (64), and emerging evidence suggests a novel role for tissue-specific transcription factors, otherwise important for driving a defined pattern of gene expression, in regulating cell survival. Thus, the zinc finger transcription factor GATA-1 is required for terminal maturation in the erythroid lineage (65). GATA-1 gene ablation in embryonic stem cells blocks the differentiation at the proerythroblast stage (65, 66) and causes apoptotic death (66). Similarly, GATA-4 is required for cardiomyocyte differentiation from embryonal carcinoma cells (67) and cell survival in the cardiomyocyte pathway but not the neuronal lineage. Hence, apoptosis may be linked to insufficient survival signals, but also to inefficient or abortive cell differentiation. Similarly, SCL is required early in hemopoiesis, possibly in specifying ventral mesoderm to the hemopoietic fate (19), and is shown here to be crucial for SF-mediated cell survival. Therefore, it is proposed that lineage-specific transcription factors orchestrate both lineage-specific gene expression and the survival of the specified cell types.

The authors wish to thank Dr. Robert Hawley (Oncology Division, Toronto General Hospital, Toronto, Ontario, Canada) for providing the MSCV retroviral vectors; Drs. Rafick P. Sékaly and Guy Sauvageau (IRCM, Montreal, Quebec, Canada) for critical reading of the manuscript; Dr. André Haman for help with Northern blotting; Dr. Richard Martin for the MSCV-dn-SCL construct; Dr. Don Lombardi for the pMTH as-SCL construct; and Magali Domin and Vivianne Jodoin for expert secretarial assistance.

This work was supported by grants from the National Cancer Institute of Canada with funds from the Canadian Cancer Society and a research traineeship from the Heart and Stroke Foundation of Canada to F. Charron. M. Nemer is a Senior Scientist of the Medical Research Council of Canada and T. Hoang is Senior Scientist from the Fonds de la Recherche en Santé du Québec and a Visiting Scientist of the National Cancer Institute (USA).

as-SCL

antisense SCL

dn-SCL

dominant negative SCL

GAPDH

rat glyceraldehyde-3-phosphate dehydrogenase

MSCV

murine stem cell virus

SF

Steel factor