Erythroblast Transformation by FLI-1 Depends upon Its Specific DNA Binding and Transcriptional Activation Properties*

FLI-1 is a transcriptional regulator of the ETS family of proteins. Insertional activation at the FLI-1 locus is an early event in F-murine leukemia virus-induced erythroleukemia. Consistent with its essential role in erythroid transformation, enforced expression of FLI-1 in primary erythroblasts strongly impairs the response of these cells to erythropoietin (Epo), a cytokine essential to erythropoiesis. We show here that point mutations in the ETS domain that abolished FLI-1 binding to specific DNA elements (ETS-binding sites) suppressed the ability of FLI-1 to transform erythroblasts. The exchange of the entire ETS domain (DNA binding domain) of FLI-1 for that of PU.1 changed the DNA binding specificity of FLI-1 for that of PU.1 and impaired FLI-1 transforming properties. In contrast, ETS domain swapping mutants that maintained the DNA binding specificity of FLI-1 did not affect the ability of FLI-1 to transform erythroblasts. Deletion and swapping mutants that failed to inhibit the DNA binding activity of FLI-1 but impaired its transcriptional activation properties were also transformation-defective. Taken together, these results show that both the ability of FLI-1 to inhibit Epo-induced differentiation of erythroblasts and to confer enhanced cell survival in the absence of Epo critically depend upon FLI-1 ETS-binding site-dependent transcriptional activation properties.

FLI-1 is a member of the ETS family of transcriptional regulators, which plays an essential role in development and oncogenesis (for review see Ref. 1). FLI-1 shares with other ETS proteins a conserved ETS domain that is responsible for its targeting to the nucleus and specific binding to 10-bp-long DNA sequences centered over a GGA core (ETS-binding sites (EBSs) 1 ) (2). When bound to specific EBSs in the context of cellular, viral, or model promoters/enhancers, FLI-1 most often activates transcription, a property that relies on two activation domains localized on the C-terminal and N-terminal sides of the ETS domain (3)(4)(5). However, in the context of specific promoters, FLI-1 binding results in transcriptional repression through ill defined mechanisms (6,7). In addition to its transcriptional regulatory properties resulting from its tethering to DNA, FLI-1 has been shown to modulate in trans the activity of other unrelated transcriptional regulators through protein-protein interactions (8,9).
The human FLI-1 and the highly related ERG genes are rearranged as the result of specific chromosomal translocations in Ͼ95% of the cases of Ewing sarcoma, a pediatric tumor of neuroectodermal origin (10,11). In this disease, the 3Ј-part of FLI-1 or ERG is translocated to the 5Ј-half of EWS, a member of the TET gene family of RNA-binding proteins (12). The resulting fusion gene encodes an EWS-FLI-1 or EWS-ERG fusion protein in which the N-terminal activation domain of FLI-1/ERG is replaced by the potent N-terminal activation domain of EWS, thereby generating an altered regulator of EBS-driven transcription (13)(14)(15)(16). EWS-FLI-1 transforms NIH3T3 cells to anchorage-independent growth and accelerates the tumorigenic potential of these cells following their transplantation in nude mice (14,17). The molecular events involved in EWS-FLI-1 transforming properties remain to be characterized but appear to rely on both DNA binding-dependent and DNA binding-independent properties (18,19).
FLI-1 was originally identified as a common proviral integration site in erythroleukemia induced in the newborn mouse by the F-murine leukemia virus component of the Friend virus complex (for review see Refs. 1 and 20). The other component of the Friend virus complex, SFFV, induces erythroleukemia in adult mouse. In this case, the hallmark of the disease is the proviral insertional activation of Spi-1/PU.1, another member of the ETS gene family (21).
The early phase of F-murine leukemia virus-induced erythroleukemia is characterized by the expansion of erythroblasts in the spleen of infected animals and by severe anemia. The emergence of proliferating erythroblasts is concomitant with the rearrangement of the FLI-1 locus and the activation of FLI-1 expression (20,22). At that stage, proliferating cells are not immortalized, and additional genetic events are required, including the loss of the p53 tumor suppressor gene, to bypass senescence and induce immortalization (22,23). Consistent with its central role in F-murine leukemia virus-induced erythroleukemia, enforced expression of FLI-1 in a mouse erythroleukemic cell line (7) and primary avian erythroblasts (24) has been shown to strongly interfere with the normal response of these cells to erythropoietin (Epo), a cytokine essential to erythropoiesis (25,26). The mechanisms underlying the transforming properties of FLI-1 in erythroblasts are mostly not understood. They have been proposed to involve FLI-1 interference with molecular events important for erythroid differentiation (7,8) and FLI-1-directed activation of novel genes not normally expressed in erythroblasts (27).
We report here the analysis of the DNA binding activity, transcriptional regulatory properties, and erythroblast transforming ability of a series of FLI-1 mutants. Our results show that the transforming properties of FLI-1 in primary erythroblasts are critically dependent upon its binding activity to specific EBSs. In addition, both the survival-inducing and differentiation-inhibiting properties of FLI-1 require a transcriptionally active protein, indicating that they involve the transcriptional activation of specific genes.
Cell Culture, Retroviral-mediated Gene Transfer, and Differentiation Assays-HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Invitrogen), 1 mM glutamine, penicillin, and streptomycin. Chicken embryo fibroblasts were grown in the same medium supplemented with 2% chicken serum (Sigma).
Infectious avian retroviruses were generated by co-transfection of primary chicken embryo fibroblasts with the replication-competent pRCAS-mEpoR and the different pSFCV retroviral vector derivatives and infected cells selected by their resistance to G418 (Invitrogen). To generate primary erythroblasts, chicken bone marrow cells were doubly infected with the S13 virus, encoding the temperature-sensitive ts-v-Sea protein kinase (34) together with the different SFCV/R-CAS-EpoR viral stocks as described previously (28). Infected bone marrow cells were seeded in CFU-E-methocel supplemented with 100 ng/ml recombinant chicken SCF, 1.4 nM insulin (Novo-Nordik), and 2.7 mg/ml G418. Under these conditions, only G418-resistant and transduced erythroblasts form colonies within 5-6 days at 37°C. Erythroblast colonies were picked and expanded in CFU-E medium containing SCF and insulin. Differentiation analyses were performed as described previously (28). To check for expression of the expected exogenous proteins, Western blots were performed using anti-EpoR (Santa Cruz Biotechnology, sc-697), anti-FLI-1 (Santa Cruz Biotechnology, sc-356), anti-HA (Santa Cruz Biotechnology, sc-805), anti-ERK2 (Santa Cruz Biotechnology, sc-154), and an anti-pan-ETS monoclonal antibody (kindly provided by N. K. Bhat, Frederick, MD).
Transient Tranfections, Luciferase Assays, and Immunofluorescence Analyses-For transactivation experiments, 2.5 ϫ 10 5 HeLa cells were plated in 6-well plates and transfected 24 h later with the indicated amount of plasmid DNA using the Lipofectamine Plus reagent (Invitrogen) in the absence of serum, as recommended by the manufacturer. The DNA mixture included the indicated amounts of reporter gene constructs and expression plasmids. The total amount of expression plasmid was kept constant by the addition of empty ⌬EB vector, and the total amount of DNA was kept constant to 1 g by the addition of carrier plasmid DNA. Cell lysates were prepared 24 h after transfection and assayed for luciferase activity by using the luciferase assay system kit (Promega). The results shown are the mean of at least four independent transfections experiments.
Immunofluorescence analyses were performed using HeLa cells transfected with 5 g of the indicated expression plasmids. Two days after transfection, cells were fixed with 4% paraformaldehyde and permeabilized by incubation in 0.3% Triton X-100 10 min. After incubation for 1 h with a 1:200 dilution in PBS, 10% fetal calf serum of anti-HA antibody, cells were washed in PBS and incubated with a 1:100 dilution in PBS, 10% fetal calf serum of a fluorescein isothiocyanate-conjugated goat anti-rabbit immunoglobulin (Amersham Biosciences) and nuclei were stained by Hoechst. After several PBS washes and dehydration, coverslips were mounted in Mowiol. Fluorescence was visualized with an epifluorescence microscope.

RESULTS
FLI-1 can regulate transcription either following its binding to specific DNA elements present in the enhancer/promoter regions of responsive genes or independently of DNA binding, through interactions in trans with other transcriptional regulators. To investigate the relative importance of these mecha-nisms to the transforming properties of FLI-1, a series of FLI-1 mutants were generated and characterized for their DNA binding, transcriptional regulatory properties, and transforming properties in erythroblasts.
NMR studies have shown that the DNA binding domain (ETS domain) of FLI-1 is composed of three ␣-helices (␣1-␣3) and a four-stranded ␤-sheet (strands ␤1-␤4) (36); see Fig. 1). Helices ␣2 and ␣3 together with intervening residues form a helix-turn-helix motif that lies over the surface formed by the four-stranded ␤-sheet. Crystallographic studies have shown that residues from both the helix-turn-helix and ␤3-␤4 motifs make direct contact with DNA and determine to a major extent the DNA binding specificity of ETS proteins (37,38). Three point mutants in the ETS domain were constructed (see Fig. 1 for a schematic description of the mutants). In FLI-1[R337,340L], the arginine residues in helix ␣3 that make direct base contacts with the guanine residues of the EBS GGA core were mutated to leucine. This mutation has been shown previously to abolish the specific DNA binding activity of the isolated ETS domain of FLI-1 (36) and of EWS-FLI-1 (13). In FLI-1[I347E], an isoleucine residue at the end of helix ␣3 was changed for a glutamic acid residue, a mutation that inhibits the specific binding to DNA of EWS-FLI-1 (18). We also constructed FLI-1[D344V] in which an aspartate residue immediately C-terminal to the ␣3 recognition helix was mutated into valine, a mutation that has been reported to only minimally affect the DNA binding activity of ETS proteins but broadens their DNA binding specificity (39). The DNA binding activity of these mutants was analyzed by electrophoretic mobility shift assay using in vitro translated proteins ( Fig. 2A, lanes 2 and 7-10) and a 32 P-labeled doublestranded DNA probe corresponding to an optimized FLI-1binding site (probe A, see "Experimental Procedures"). As expected, both FLI-1[R337,340L] and FLI-1[I347E] were defective for their specific binding to DNA, whereas FLI-1[D344V] bound DNA similarly as wtFLI-1 (Fig. 2B, compare  lanes 2 and 3; Fig. 2B, compare lanes 5 and 9 -10 1). The ETS domain of FLI-1 is closely related to that of ETS-1 (67% identity) but only distantly related to that of PU.1 (38% identity). Consistent with this, FLI-1 and ETS-1 display a similar DNA binding specificity in vitro (2,40,41), whereas PU.1 binds to EBSs (Pu boxes) with distinct base composition both at the 5Ј-and 3Ј-sides of the GGA core (42). Both FLI-1[EEE] and FLI-1 [PPP] were expressed at similar levels as wtFLI-1 following in vitro transcription/translation in reticulocyte lysates of the corresponding DNA templates ( Fig. 2A,  lanes 2-4). The DNA binding activity and specificity of these mutants were compared with that of wtFLI-1 by electrophoretic mobility shift analyses using two additional EBS probes, besides probe A. Probe T is identical to probe A except for the transversion of the last adenine of the GGAA core sequence for a thymine, whereas the Pu probe is a high affinity binding site for PU.1 (43). In vitro translated PU.1 was used as specificity control in these analyses. wtFLI-1 bound probes A and T but failed to bind the Pu probe (lane 2 in each panel of Fig. 2D) whereas PU.1 more efficiently bound the Pu probe than probe A and failed to bind the T probe (Fig. 2E). Similarly to wtFLI-1, FLI-1[EEE] bound probes A and T but failed to bind the Pu probe (lanes 4 in Fig. 2D). In contrast, FLI-1[PPP] displayed higher binding activity toward the Pu probe than to probe A and failed to bind probe T, thereby adopting the DNA binding specificity of PU.1 (lanes 3 in Fig. 2D). This was confirmed in competitive EMSA in which the binding of FLI-1[PPP] to probe A was competed in a dose-dependent manner by the addition of an increasing molar excess of unlabeled oligonucleotides Pu and A but was barely competed by the same excess of oligonucleotide T (Fig. 3A, compare lanes 2-11  in FLI-1(PPP) panel). The same competition pattern was observed for PU.1 (Fig. 3A, lanes 2-11 in PU.1 panel). In contrast, binding of both wtFLI-1 and FLI-1[EEE] to probe A was competed in a dose-dependent manner by unlabeled oligonucleotide A and T but not by the Pu oligonucleotide (Fig. 3A, lanes 2-11  in the wt-FLI-1 and FLI-1(EEE) panels). As expected (Fig. 3A,  (lanes 2-12). The binding reaction mixtures included either no competitor oligonucleotide (lanes 1 and 2) or a 10- fold (lanes 3, 6, and 9), 30-fold (lanes 4, 7, and 10), or 90-fold ( lanes 5, 8, 11, and 12) molar excess of unlabeled competitor oligonucleotides as indicated at the top. B, EMSA was carried out with 200 fmol of oligonucleotide T as a probe and either control reticulocyte lysate (lane 1), in vitro expressed wtFLI-1, or the indicated FLI-1 derivatives (lanes 2-9). The binding reaction mixtures included either no competitor oligonucleotide (lanes 1 and 2) or a 10-fold (lanes 3 and 6), 30-fold (lanes 4 and 7), or 90-fold (lanes 5, 8, and 9) molar excess of the indicated unlabeled competitor oligonucleotide. lane 12), none of the protein-DNA complexes were competed by a mutant Am competitor that is identical to oligonucleotide A except for GG-to-CC transversion in the GGA core, a mutation known to abolish specific binding of ETS proteins to DNA. These experiments show that substitution of the ETS domain of FLI-1 by that of ETS-1 generates a protein with a DNA binding specificity similar to that of FLI-1, whereas its substitution for the ETS domain of PU.1 changes the DNA binding specificity of FLI-1 for that of PU.1.
The three-dimensional fold of several ETS domain-DNA complexes, including those of FLI-1 and PU.1, are highly superposable (38). Additional swapping mutants were therefore generated in which specific structural elements of the ETS domain of FLI-1 were replaced by the corresponding sub-domain of PU.1. Specifically, FLI-1 mutants PFF and FFP were generated by swapping either FLI-1 ␣1␤1␤2 or ␤3␤4 motifs for the corresponding domains of PU.1 (Fig. 1). Both mutants were stably expressed following in vitro translation in reticulocyte lysate ( Fig. 2A, lanes 5 and 6). Substitution of the ␣1␤1␤2 motif of FLI-1 by the corresponding motifs of PU.1 resulted in a protein with similar DNA binding activity as wtFLI-1, as evidenced by the efficient binding of FLI-1[PFF] to A (Fig. 2B) and T probes (Fig. 2C). Of note, FLI-1[PFF] displayed broadened DNA binding specificity because, unlike wtFLI-1, it was able to bind the Pu probe (Fig. 2F). Competitive EMSA showed that FLI-1[PFF] bound the A and T probes with comparable efficiency but only bound the Pu probe with low affinity because the FLI-1[PFF]probe A complex was inefficiently displaced in the presence of a large molar excess of unlabeled Pu oligonucleotide used as competitor (Fig. 3A, lanes 2-11 in the FLI-1(PFF) panel). Bind-ing of FLI-1[PFF] to the Pu probe is therefore at least an order of magnitude lower as compared with its binding to conventional EBSs.
Substitution of the ␤3␤4 motif of FLI-1 by the corresponding region of PU.1 in FLI-1[FFP] resulted in a protein with slightly higher DNA binding activity toward the A and T probes as compared with wtFLI-1 (Fig. 2, B and C) and broadened DNA binding specificity as evidenced by its ability to bind the Pu probe (Fig. 2F). However, FLI-1[FFP] only inefficiently bound to the Pu probe as analyzed by competitive EMSA (Fig. 3A,  lanes 2-11, in FLI-1(FFP)  FLI-1 mutants were next analyzed for their transcriptional regulatory properties in transient co-transfection assays as described previously (13). Expression plasmids for the protein under study and an EBS-driven luciferase gene reporter construct were co-transfected in HeLa cells, and transactivation was monitored by luciferase assays. Importantly, all FLI-1 mutants analyzed were found to be expressed at levels similar to wtFLI-1 in HeLa cells (Fig. 4A) and to localize to the nucleus of transfected cells (Fig. 4B). We first used the tkD2A-Luc reporter in which a duplicate copy of the ETS-responsive re-  ). B, HeLa cells transfected with either control expression plasmid or expression plasmid for HA-wtFLI-1 or the indicated FLI-1 mutants were seeded to collagen-treated coverslips, fixed in 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and incubated with either anti-HA or anti-FLI-1 antibodies as indicated, followed by a fluorescein isothiocyanate-conjugated rabbit immunoglobulin G. The inset corresponds to Hoechst staining of the same field. gion-1 of the human T-cell leukemia virus type I-long terminal repeat is inserted upstream of the Ϫ55 herpes simplex virus thymidine kinase (tk) promoter. Our previous analyses have shown that transcriptional activation of this promoter by FLI-1 depends upon FLI-1 binding to two specific EBSs in ERR1 (3). Consistent with these results, wtFLI-1 activated luciferase expression from tkD2A-Luc in a dose-dependent manner (Fig. 5A) whereas FLI-1 mutants defective for their specific binding to DNA (FLI-1[R337,340L]; FLI-1[I347E]) were completely inactive (Fig. 5A). Mutant FLI-1[D344V], which binds DNA with the same activity as wtFLI-1, efficiently transactivated tkD2A-Luc (Fig. 5A). The transcriptional activation properties of FLI-1 have been reported to depend upon two activation domains that localize N-terminally (NTAD) and C-terminally (CTAD) with respect to the ETS domain, respectively (5). Both FLI-1[EEE] and FLI-1[FFP] transactivated tkD2A-Luc in a dose-dependent fashion, albeit with reduced efficiency as compared with wtFLI-1 (Fig. 5C). The reduced activity of these mutants correlated with their reduced binding to ERR1 as compared with wtFLI-1. Indeed, competitive EMSA showed that specific DNA binding of wtFLI-1, FLI-1[EEE], and FLI-1 [FFP] was similarly competed by increasing amounts of unlabeled oligonucleotide T (Fig. 3B). In contrast, when ERR1 was used as competitor, it inhibited more efficiently the specific binding of wtFLI-1 than that of either FLI-1[EEE] or FLI-1[FFP] (Fig. 3B). In line with their very low binding activity to ERR1 (data not shown), PU.1 and FLI-1[PPP] failed to transactivate the tkD2A-Luc reporter (Fig. 5C). However, both proteins efficiently transactivated PU 3 -tk81-Luc, a reporter plasmid carrying three copies of a Pu box (see "Experimental Procedures") inserted 5Ј of the Ϫ81 herpes simplex virus tk promoter (Fig. 5E). As expected, none of these proteins were able to transactivate the control tk81-Luc reporter lacking Pu boxes (data not shown).
To analyze the transforming activity of FLI-1 mutants and compare them to wtFLI-1, the corresponding cDNAs were introduced by retrovirally mediated gene transfer into primary avian erythroblasts, and the effects of these proteins on Epodependent differentiation and survival were analyzed as described previously (28) (see "Experimental Procedures"). A dozen clones expressing both mEpoR and the respective FLI-1 proteins at similar levels (Fig. 6A) were grown for further study. The results obtained with one representative clone of each combination are reported.
In accordance with previous studies (28), mEpoR control erythroblasts differentiated terminally in response to hEpo as evidenced by their exit from the cell cycle 2 days after differentiation induction (Fig. 6B, Co panel), their accumulation over time of high levels hemoglobin (Fig. 6C, Co panel), and their acquisition of the morphology of fully mature erythrocytes (Fig.  7, Co panel, ϩhEpo). In the absence of hEpo, control erythroblasts rapidly die by apoptosis (28) as evidenced by the presence of only cell debris in cultures maintained for 3 days in the absence of hEpo (Fig. 7, Co panel, ϪhEpo). As reported previously (24), expression of wtFLI-1 inhibited Epo-induced differentiation of mEpoR erythroblasts as evidenced by their lack of hemoglobin accumulation (Fig. 6C, wtFLI-1) and their maintenance of an immature morphology (Fig. 7, wtFLI-1). Rather, mEpoR/wtFLI-1 erythroblasts self-renewed for several generations as differentiation-arrested cells under these conditions (Fig. 6B). Consistent with our previous analyses (24,27), cell survival in mEpoR/wtFLI-1 erythroblasts was enhanced as compared with control mEpoR erythroblasts when these cultures were maintained in the absence of hEpo (Fig. 7, compare Co and wtFLI-1 panels, ϪhEpo).
FLI-1 [PPP] in which the DNA binding specificity of FLI-1 was switched for that of PU.1 was transformation-defective as mEpoR/FLI-1[PPP] erythroblasts terminally differentiated in response to hEpo and died in its absence (FLI-1(PPP) panels in Fig. 6, B and C, and Fig. 7). In contrast, FLI-1[EEE] and FLI-1[FFP], which displayed a DNA binding specificity similar to that of wtFLI-1, both inhibited hEpo-induced differentiation and induced enhanced survival in the absence of hEpo (Fig. 6,  B and C, and Fig. 7). FLI-1[EEE] and FLI-1[FFP] are therefore indistinguishable from wtFLI-1 for their ability to transform erythroblasts.
The transcriptionally defective FLI-1[PFF] mutant was unable to transform erythroblasts as mEpoR/FLI-1[PFF] erythroblasts were indistinguishable from control mEpoR erythroblasts in their ability to terminally differentiate in response to hEpo (Fig. 6, B and C, and Fig. 7) and to rapidly die by apoptosis in the absence of hEpo (Fig. 7). Of note, neither the NTAD nor the CTAD proved to be absolutely required in a non-redundant fashion to FLI-1 transforming properties because both mEpoR/FLI-1  and mEpoR/FLI-1[225-452] erythroblasts failed to differentiate in response to hEpo (Fig. 6, B and C, and Fig. 7) and showed prolonged survival in the absence of hEpo (Fig. 7). In contrast, deletion of both activation domains in FLI-1[276 -373] generated a transformation-defective protein (Figs. 6 and 7). Finally, we analyzed the transforming properties of VP16/FLI-1 [276 -373]. As shown in Fig. 7, mEpoR/VP16/FLI-1[276 -373] erythroblasts were only partially affected in their response to hEpo because these cells showed prolonged survival in the absence of hEpo when compared to mEpoR control erythroblasts and underwent only partial morphological differentiation in response to hEpo (Fig.  7). In line with this, VP16/FLI-1[276 -373] erythroblasts accumulated intermediate levels of hemoglobin (Fig. 6D). Taken together, these results show that erythroblast transformation by FLI-1 requires both its specific DNA binding activity and transcriptional activation properties.
To investigate whether the transcriptional regulatory properties of FLI-1 and its derived mutants as assessed in reporter gene assays would translate into gene deregulation in vivo, we analyzed the expression of several genes involved in FLI-1-induced erythroblast transformation. Our previous studies (27) have shown that BCL-2 is a direct target of FLI-1 in primary erythroblasts and that up-regulation of BCL-2 expression is involved in the ability of FLI-1 to induce erythroblast survival in the absence of hEpo. Two other genes were identified by differential screening to be strongly up-regulated in FLI-1-transformed erythroblasts: the first encodes SLAP, a SH2-and SH3-domain adaptor, and the second encodes inositol polyphosphate 5Ј-phosphatase ((45) and data not shown). mEpoR control erythroblasts, mEpoR/wtFLI-1, mEpoR/FLI-1[PPP], and mEpoR/FLI-1[PFF], were maintained in the absence of hEpo for 24 h, and expression of BCL-2, SLAP, and IN5PP5P was analyzed by RT-PCR. The results of Fig. 8 show that expression all three genes was up-regulated in wtFLI-1transformed erythroblasts as compared with control cells. In contrast, none of these genes are detectably expressed in erythroblasts expressing the transactivation-defective FLI-1[PFF] and in erythroblasts expressing the transformation-defective FLI-1[PPP]. DISCUSSION The early and recurrent activation of FLI-1 by retroviral insertional mutagenesis in F-murine leukemia virus-induced erythroleukemia suggests that up-regulation of FLI-1 expression is the initiating event in this multistep leukemia model. Consistent with this notion, enforced expression of FLI-1 in mouse erythroleukemic cell lines (see Ref. 7 and this study) and in primary avian erythroblasts (24) was found to inhibit Epoinduced differentiation and to promote cell survival in the absence of Epo. The molecular mechanisms involved in the ability of FLI-1 to modify the normal response of erythroblasts to Epo stimulation or withdrawal are only partially understood. The survival-inducing properties of FLI-1 in erythroblasts result in part from the activation of BCL-2 gene expression and subsequent induction of the anti-apoptotic BCL-2 protein (27,46). The ability of FLI-1 to inhibit erythroblast differentiation has been proposed to involve FLI-1-mediated repression of retinoblastoma gene transcription (7). This may result in inhibition of an intrinsic but non-cell autonomous retinoblastoma function required for terminal erythroid differentiation (47). Alternatively, a direct interference of FLI-1 with other transcriptional regulators important for erythroid differentiation such as retinoid/thyroid nuclear hormone receptors and EKLF has been proposed (8,9).
The results of the present study show that the transforming properties of FLI-1 in erythroblasts critically depend upon both its specific binding to DNA and transcriptional activation prop-erties. First, point mutations in the ETS domain of FLI-1 that abolished specific DNA binding to FLI-1-responsive EBSs were found to suppress both the survival-inducing and differentiation inhibitory properties of FLI-1 in erythroblasts. Second, a mutant of FLI-1 retaining the DNA binding specificity of wt-FLI-1 but defective in its ability to activate EBS-driven transcription (FLI-1[PFF]) completely failed to transform erythroblasts. Third, substitution of the ETS domain of FLI-1 by that FIG. 6. Comparison of the transforming properties of wtFLI-1 and FLI-1 mutants in primary erythroblasts. Erythroblast clones expressing either EpoR (Co, control), EpoR ϩ wtFLI-1, or EpoR ϩ FLI-1 derivatives were generated by retrovirally mediated gene transfer as described under "Experimental Procedures." A, expression of exogenous proteins in representative erythroblast clones expressing EpoR (lanes 1 and 2), EpoR/HA-wtFLI-1 (lanes 3, 4, and 13 of the distantly related ETS protein PU.1 resulted in a FLI-1 protein (FLI-1[PPP]) with a DNA binding specificity skewed to that of PU.1 and to the loss of its ability to transform erythroblasts. In contrast, swapping mutants in the ETS domain that maintained the DNA binding specificity of FLI-1 (FLI-1[EEE]; FLI-1[FFP]) resulted in fully transforming proteins. Taken together, these results indicate that the survival-inducing properties of FLI-1 and its ability to inhibit erythroid differentiation both require the activation of specific target genes, the regulation of which is under the control of FLI-1-responsive EBS elements.
We have shown recently that the BCL-2 gene is a direct FLI-1 target in erythroblasts and that BCL-2 up-regulation contributes to the enhanced survival of these cells in the absence of Epo (27). However, overexpression of BCL-2 (27), although favoring cell survival, does not interfere with terminal differentiation, suggesting that the block imposed by FLI-1 upon Epo-induced differentiation must result from the activation of additional genes. The search for genes differentially activated in FLI-1-transformed erythroblasts is underway. Evidence gathered so far indicates that FLI-1 induces the de novo up-regulation of several genes in erythroblasts that encode adaptors and effectors of signaling pathways, 2 suggesting that FLI-1 may inhibit terminal differentiation through its ability to specifically interfere with EpoR signaling output. Besides its role in nuclear targeting and DNA binding, the ETS domain also contributes to the specific interaction of ETS proteins with other transcriptional regulators. These interactions can take place independently of DNA binding when ETS proteins interact in trans with specific factors, often through the ␤3/␤4 motif of the ETS domain (48 -52). Other interactions result in the cooperative binding of ETS proteins with unrelated factors to composite cis-acting DNA-response elements. In this case, structural studies have shown that cooperativity results from limited interactions between amino acid residues of each partner and often involves the winged helix-turn-helix region of the ETS domain. These interactions induce a local modification in the conformation of one or both factors that results in a change in specific DNA contacts to provide additional energy for the complex to bind composite DNA elements (53)(54)(55)(56). It cannot be excluded that the difference in activity between wtFLI-1 and transformation-defective mutants like FLI- [PPP] involves, in addition to a change in DNA binding specificity, the disruption of a protein-protein interaction critical to the transforming properties of wtFLI-1. In this scenario, this interaction would be conserved in other mutants like FLI-1[EEE] and would not involve the ␤3/␤4 fold of FLI-1 because substitution of the ␤3/␤4 region of FLI-1 by that of PU.1 in FLI-1[FFP] generated a fully transforming protein.
The requirement for specific DNA binding activity and deregulated EBS-driven transcription in order for FLI-1 to transform erythroblasts contrasts with results obtained for the EWS-FLI-1 fusion protein specific for Ewing sarcoma. EWS-FLI-1 is a potent activator of EBS-driven transcription (13), and part of its transforming properties is believed to derive from its ability to deregulate the expression of critical genes either directly (16,57) or indirectly (13, 17, 58 -60). As for wtFLI-1, introduction of the equivalent of the I347E and R337L,R340L mutations in the ETS domain of EWS-FLI-1 results in proteins that are also defective for their specific binding to DNA. However, in the case of EWS-FLI-1, these mutants retain substantial colony forming activity and essentially intact tumorigenic inducing potential (18,19). These results suggest that the transforming properties of EWS-FLI-1 also rely on DNA binding-independent mechanisms that could be related to the ability of EWS-FLI-1 to interfere with the activity of EWS as an adaptor to couple gene transcription to RNA splicing (31,61,62). The notion that the transforming properties of FLI-1 and EWS-FLI-1 involve distinct mechanisms is further substantiated by the fact that although deletion of the C-terminal domain of FLI-1 in FLI-1  does not impair its ability to transform erythroblasts, the same deletion in EWS-FLI-1 severely reduces its anchorage-inducing and tumorigenic properties (17). It should be stressed, however, that these notions mostly derive from the analysis of EWS-FLI-1 transforming properties in NIH3T3 fibroblasts, a cellular background distinct from the neuroectodermal origin of the cells presumably targeted by EWS-FLI-1. It remains to be seen FIG. 7. Comparison of the transforming properties of wtFLI-1 and FLI-1 mutants in erythroblasts, morphological analyses. A representative EpoR erythroblast clone (Co, control) and representative EpoR clones expressing either wtFLI-1 or the different FLI-1 mutants were maintained at 42°C in differentiation medium either in the presence (ϩhEpo) or absence of hEpo (ϪhEpo). Aliquots of cells were cytocentrifuged 3 days after differentiation induction and stained with neutral benzidine and Giemsa. Immature cells stain blue; hemoglobinized cells stain brown.

FIG. 8. RT-PCR analysis of FLI-1 target genes in erythroblasts.
Two clones of each mEpoR, mEpoR/wtFLI-1, mEpoR/FLI-1[PPP], and mEpoR/FLI-1[PFF] erythroblasts were maintained for 24 h at 42°C in the absence of hEpo; RNA were extracted and reverse-transcribed. Primers specific for the genes encoding BCL-2, SLAP, inositol polyphosphate 5Ј-phosphatase (IN5PP5P), and ribosomal protein S17 (Co, control) were used to amplify the corresponding sequences. Aliquots of 0.5 and 2 l of each RT products were used. PCR-amplified products were analyzed by agarose gel electrophoresis in the presence of ethidium bromide. whether the differences between the determinants involved in the ability of FLI-1 to transform erythroblasts and those involved in EWS-FLI-1 transforming properties will hold when a more appropriate cellular system for the latter becomes available.