INTRODUCTION

Top Abstract Introduction Materials and methods Results Discussion References

The Friend viral complex is composed of two different entities, a replication-defective viral component (spleen focus-forming virus [SFFV]) and a replication competent virus (Friend murine leukemia virus [F-MuLV]), which cause erythroleukemia in susceptible mice (5). The initial phase of the disease induced by the Friend viral complex is a polyclonal expansion of erythroblasts which are still able to differentiate. It occurs due to constitutive activation of the erythropoietin (Epo) receptor mediated by its physical interaction with the gp55^env glycoprotein encoded by SFFV (6, 29). After several weeks of infection, erythroleukemic cells of clonal origin begin to emerge which have unlimited self-renewal capacities and do not differentiate. Most erythroleukemic cell lines established from this second step contain SFFV proviral integrations in the Spi-1 locus. This leads to the transcriptional activation of the adjacent gene encoding the ETS family transcription factor Spi-1/PU.1 (33-35, 37). On the other hand, the initial phase of the disease induced by F-MuLV alone is characterized by severe anemia and a massive proliferation of infected erythroid progenitor cells within the spleen and liver. These cells, unlike those derived from SFFV-induced erythroleukemias, are unable to grow directly in culture (22). However, erythroleukemic cell lines can be established following serial in vivo passages of primary tumor cells in syngenic animals. Molecular analyses established that proviral integration occurred in the Fli-1 locus in 75% of these erythroleukemic cell lines, leading to transcriptional activation of the adjacent gene encoding another ETS family transcription factor, Fli-1 (3-5). Insertional activation of the Fli-1 gene appears to be the first genetic event associated with F-MuLV-induced primary erythroleukemias. Rearrangement of the Epo gene resulting in constitutive Epo expression is also often detected in leukemic cells derived from BALB/c mice infected by F-MuLV (23). In addition, inactivation of the tumor suppressor gene p53 is also a very common genetic alteration observed in most erythroleukemic cell lines induced by either SFFV or F-MuLV (5, 28). Thus, erythroleukemias induced by both viruses are associated with similar genetic events including activation of the Epo receptor signaling pathway, inactivation of the p53 gene, and activation of ETS family transcription factors. However, they differ in two mains aspects: (i) the temporal order of these genetic events and (ii) the member of the ETS gene family activated, Spi-1/PU.1 or Fli-1.

Various strategies have been used to ascertain the role of Spi-1/PU.1 in erythroid cell transformation. Earlier studies demonstrated that infection of long-term bone marrow cultures with an Spi-1/PU.1-transducing retrovirus caused the proliferation of proerythroblast-like cells that differentiated at low frequency into hemoglobinized cells (42). Alternatively, antisense oligonucleotides were used to reduce Spi-1/PU.1 expression in SFFV-transformed cell lines. Treated cells exhibited a reduced proliferative capacity, again suggesting a role for Spi-1/PU.1 in the self-renewal of transformed erythroblastic cells (10). Transgenic mice overexpressing Spi-1/PU.1 were also established and shown to develop spontaneously multistep erythroleukemias mimicking the disease induced by SFFV in vivo (36). In addition, it has been shown that chemically induced erythroid differentiation of SFFV cell lines is associated with the downregulation of Spi-1/PU.1 expression (10, 15, 21, 43). Interestingly, when high levels of Spi-1/PU.1 are maintained by stable transfection of Spi-1/PU.1-expressing vectors, this chemically induced erythroid differentiation is blocked (39, 49), indicating that Spi-1/PU.1 deregulation is also involved in the inhibition of erythroid differentiation. Although the role of Spi-1/PU.1 in erythroid cell transformation is clearly established, the precise mechanism by which this occurs remains unclear.

To date, similar attempts to establish the role of Fli-1 in erythroid cell transformation have been unsuccessful. For example, transgenic mice overexpressing moderate levels of Fli-1 protein have been established. In contrast to Spi-1/PU.1 transgenic animals, Fli-1 transgenic mice do not develop erythroleukemia but instead develop renal disease due to an immune dysfunction (52). Mice carrying targeted inactivation of exon 2 of both alleles of Fli-1 have also been established. Interestingly, these Fli-1^/ mice remain susceptible to the development of erythroleukemias following infection with F-MuLV. However, the latency period is significantly increased, and erythroleukemic cells derived from the infected Fli-1^/ mice display proviral integration in the rearranged Fli-1 locus and produce high levels of a truncated Fli-1 protein. This truncated Fli-1 protein has been shown to arise from an internal translation initiation site and alternative splicing around the neo cassette used in gene targeting (32). These results obtained with Fli-1^/ mice strongly suggest that the deregulation of Fli-1 expression must contribute to erythroleukemia induced by F-MuLV. Furthermore, Fli-1 overexpression has also been found in other erythroleukemic cell lines, for example, those derived from transgenic mice expressing c-Myc under the control of the GATA-1 regulatory sequences (47). This latter observation may suggest that deregulation of Fli-1 gene expression is a more general and critical event involved in erythroleukemia than previously anticipated.

On the other hand, other studies have shown that Spi-1/PU.1 and Fli-1 proteins are functionally distinct in that they recognize and transactivate through distinct DNA binding sites except the core GGA consensus which is common to all members of the ETS family (27, 40, 48, 51). Based on these data, it has been suggested that the deregulated expression of Spi-1/PU.1 or Fli-1 protein might contribute to erythroleukemia induced by SFFV or F-MuLV, through the deregulation of different sets of genes (51).

In this study, we demonstrate that all SFFV cell lines that express Spi-1/PU.1 actually coexpress Fli-1 transcripts and Fli-1 protein at levels comparable to those observed in F-MuLV-transformed cell lines. We found that in SFFV-transformed cell lines, Fli-1 transcripts are initiated from a new promoter located in region 270/41 which is different from the one used in F-MuLV-transformed cell lines. More interestingly, the activity of the 270/41 Fli-1 promoter in SFFV-transformed cell lines is under the control of Spi-1/PU.1, which binds to two highly conserved consensus ETS binding sites (EBSs). In addition, we show that Fli-1 expression is upregulated following stable transfection of a Spi-1/PU.1 expression vector in SFFV-transformed cell lines. Finally, by stable transfection of constitutive or inducible Fli-1 expression vectors, we demonstrate that deregulated overexpression of Fli-1 proteins inhibits hexamethylenebisacetamide (HMBA)-induced erythroid differentiation of SFFV-transformed cells. Taken together, these data establish that Fli-1 is a transcriptional target of Spi-1/PU.1 involved in the inhibition of erythroid differentiation in SFFV-derived cell lines. The importance of this newly established Spi-1/PU.1Fli-1 regulatory cascade in erythroid cell transformation and its putative involvement in normal hematopoiesis are discussed.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and methods Results Discussion References

RNA extraction and Northern blot analysis. Total RNA from cell lines was prepared with RNAzol or RNA PLUS (Bioprobe-Systems, Montreuil-sous-bois, France) according to the manufacturer's procedure. For Northern blot analysis, RNA samples were run on a 1% agarose-formaldehyde gel and transferred to Hybond C Extra membranes (Amersham) according to standard protocols (41) with a vacuum blotting system (VacuGene XL; Pharmacia). The probes used were the purified cDNA corresponding to the 1.7-kb EcoRI fragment for Fli-1 and the 1.2-kb EcoRI fragment for Spi-1/PU.1. Probes were labeled with [³²P]dCTP by the random priming method (13). Hybridizations were performed at 42°C as previously described (41). After two successive rinses in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.1% sodium dodecyl sulfate (SDS), membranes were incubated twice at 55°C in 1× SSC-0.1% SDS for 20 min and then exposed on Kodak X-Omat AR or BIOMAX MS film (Eastman Kodak Co., Rochester, N.Y.) with an intensifying screen at 80°C. Quantitation of specific signals on autoradiographs was performed by densitometric analysis using a Bioprofil 4.6 densitometer (Vilber Lourmat) or by phosphorimager (Bio-Rad GS-525 Molecular Imager) analysis after standardization against 18S rRNA (13-kb EcoRI genomic probe).

5' RACE. Rapid amplification of 5' cDNA ends (5' RACE) was performed according to the manufacturer's protocol (Marathon kit; Clontech, Palo Alto, Calif.). Briefly, CB7 or 745-A poly(A)⁺ mRNA was purified by oligo(dT)-cellulose chromatography. A modified lock-docking oligo(dT) primer was used as a primer for the first-strand cDNA synthesis along with Moloney murine leukemia virus reverse transcriptase at 42°C. Second-strand synthesis was performed by the method of Gubler and Hoffmann (18) with a convenient cocktail of Escherichia coli DNA polymerase I, RNase H, and E. coli DNA ligase. Following creation of blunt ends with T4 DNA polymerase, the double-stranded cDNA was ligated to the Marathon cDNA adapter. The 5' RACE PCR was primed with the Marathon adaptor primer and the internal gene-specific primer NFli3' (5'-GGGACTGATCGTCACTCACCACA-3') from the mouse Fli-1 exon 2, complementary to the mRNA sequence from +30 to +52 (relative to nucleotide A [+1] of the translation initiation codon ATG described by Barbeau et al. [1]) (see Fig. 3). The Fli-1-specific products were analyzed on 2.5% NuSieve GTG agarose (FMC Bioproducts, Rockland, Maine), subsequently cloned in the pTAg vector (R&D Systems), and sequenced.

RNase protection analyses. RNA samples were prepared as described for Northern blot analysis. Two antisense Fli-1 riboprobes were used. The first, a T7 antisense riboprobe corresponding to the Fli-1 genomic sequence from 1298 to 144 (probe 2), was prepared from CB7 genomic DNA after PCR with two oligonucleotides (upper primer R [5'-GGGTACCCGGGCGACTCA-3'] and lower primer 3 Fli-3' [5'-CTCGCTCCCCTGTGCACG-3']) and subcloning in the pTAg vector (R&D Systems). The second, an SP6 antisense riboprobe corresponding to the Fli-1 cDNA sequence from 262 to +52 (probe 3), was constructed by amplifying Fli-1 cDNA with two oligonucleotides (upper primer 4 Fli-5' [5'-GACTTCCTCCCCGATCGCAAAGT-3'] and lower primer NFli3' [5'-GGGACTGATCGTCACTCACCACA-3']) and subcloning in the pGEM-T vector (Promega). The hypoxanthine ribosyltransferase (HPRT) antisense riboprobe used for standardization was obtained by amplification of mouse genomic DNA with a set of oligonucleotides described previously (26). The amplified fragment was subcloned in pGEM-T and was used to synthesize in vitro the antisense riboprobe with the T7 RNA polymerase after ScaI linearization. T7 and SP6 antisense riboprobes were synthesized as described by the supplier (Boehringer Mannheim) and hybridized overnight (500,000 cpm/sample) to 20 µg of total RNA at 30°C in 25 µl of 80% formamide-400 mM NaCl-40 mM piperazine-N,N'-bis(2-ethanesulfonic acid) (PIPES; pH 6.5)-1 mM EDTA. The samples were then digested with RNases A and T₁ as instructed by the manufacturer (Clontech). Protected fragments were separated on a 6% polyacrylamide-8 M urea sequencing gel and analyzed by autoradiography.

S1 mapping analyses. Radioactive probe 1, corresponding to the CB7 Fli-1 genomic sequence from 1298 to 156, was 5' end labeled at the 156 ApaLI site. The probe (200,000 cpm/sample) was hybridized overnight at 53°C with 30 µg of total RNA in 20 µl of 80% formamide-400 mM NaCl-40 mM PIPES (pH 6.5)-1 mM EDTA. Hybrids were digested for 2 h at 25°C with 80 U of S1 nuclease (Appligene) in 250 µl of buffer (1 mM ZnSO₄, 30 mM sodium acetate (pH 4.5)-300 mM NaCl), ethanol precipitated and analyzed on a 6% polyacrylamide-8 M urea sequencing gel.

Primer extension. The oligonucleotide primer NFli3', complementary to Fli-1 mRNA sequence from +30 to +52 (see Fig. 3), was 5' end labeled with T4 polynucleotide kinase (Boehringer Mannheim) and [-³²P]ATP and purified on a denaturing 10% polyacrylamide gel; 500,000 cpm of the primer was hybridized overnight at 30°C with 20 µg of total RNA in 30 µl of 50% formamide-400 mM NaCl-40 mM PIPES (pH 6.5)-1 mM EDTA. After ethanol precipitation, primer extension was performed for 1 h at 42°C with Moloney murine leukemia virus reverse transcriptase as described by the manufacturer (Gibco BRL) except that actinomycin D (50 µg/ml) was added. The extension products were then digested for 30 min at 37°C with DNase free RNase (Promega) before ethanol precipitation and loading on a sequencing gel.

Western blot analysis. Cellular protein extracts (15 to 20 µg) were boiled for 10 min in buffer containing 1% SDS, 0.04 M Tris HCl (pH 6.8), 6% glycerol, 150 mM -mercaptoethanol, and 0.5% bromophenol blue before being loaded onto an SDS-10% acrylamide-1.34% bisacrylamide gel. After electrophoresis, proteins were transferred to reinforced cellulose nitrate membranes (BA-S 85; Schleicher & Schuell). Membranes were blocked for 60 min in NaCl (300 mM)-Tris (10 mM) with 0.5% Tween 20 (TBS-T) containing 10% (wt/vol) milk powder as a blocking agent. Specific primary antibodies directed against either Fli-1 protein (1/200 dilution of rabbit polyclonal antibody; catalog no. sc-356; Santa Cruz Biotechnology), Spi-1/PU.1 protein (1/400 dilution of rabbit antiserum; a gift from F. Moreau-Gachelin), or Grb2 protein (1/10,000 dilution of mouse monoclonal antibody; catalog no. G16720; Transduction Laboratories) were incubated for 2 h at room temperature in TBS-T containing 5% milk powder. After six 10-min washes in TBS-T, membranes were incubated for 1.5 h in the presence of a 1/20,000 dilution of peroxidase-linked goat anti-rabbit (Sigma) or anti-mouse (Jackson Laboratories) immunoglobulin G antibody in TBS-T buffer supplemented with 5% milk powder. After six 10-min washes in TBS-T, membranes were developed with an enhanced chemiluminescence detection kit (Super Signal [Pierce] or ECL-Plus [Amersham]).

Preparation of nuclear extracts. Nuclei from 2 × 10⁷ cells were collected by centrifugation after incubation for 15 min on ice in 400 µl with buffer containing 10 mM HEPES (pH 7.5), 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride, and 2% aprotinin. The nuclei were resuspended in 100 µl of 20 mM HEPES (pH 7.5)-25% glycerol-0.42 M NaCl-1.5 mM MgCl₂-0.2 mM EDTA (pH 7.5)-0.5 mM DTT-0.5 mM phenylmethylsulfonyl fluoride-2% aprotinin and incubated for an additional 40 min on ice. The samples were centrifuged to recover the supernatants corresponding to the nuclear extracts.

Electrophoretic mobility shift assays (EMSAs). Double-stranded deoxyoligonucleotide probes were end labeled with T4 polynucleotide kinase and [-³²P]ATP. Nuclear extracts (10 µg of proteins) were preincubated for 10 min on ice with 0.8 µg of poly(dI-dC) (Pharmacia) in a buffer containing 5 mM Tris (pH 7.4), 25 mM NaCl, 0.5 mM DTT, 0.5 mM EDTA, and 1% Ficoll in a final volume of 20 µl. After addition of the probe, the samples were incubated for an additional 15 min on ice and loaded on a 4% nondenaturing polyacrylamide gel in 0.25× TBE (1× TBE is 0.089 M Tris, 0.089 M boric acid, and 0.0025 M EDTA). In the competition studies, a 500-fold molar excess of unlabeled oligonucleotide was added with the probe. In some experiments, 2 µl of preimmune serum or specific antiserum was mixed into the binding reaction mixture just before addition of the probe. Oligonucleotides sequences were as follows: Comp Spi, 5'-CCGGCCCCGGCCAGTTCCTCGATTCCGCGCTTCCTC-3'; Comp Spi mut, 5'-CCGGCCCCGGCCAGTTGGTCGATTCCGCGCTTCCTC-3' (mutations are underlined); E74, 5'-AATAACCGGAAGTAACTC-3' (40); and probe GATA-1, 5'-GATCTCCGGCAACTGATAAGGATTCCCTG-3'. The sequences of other oligonucleotides are given in Fig. 8.

Cell culture. Cells lines were provided by the following colleagues: IW32, NN10, and CTFP10 cell lines by F. Wendling; 745-A cells by F. Moreau Gachelin; CB7, 27.17, and 28.9 cells by Y. Ben-David; F4-12B2 by G. Uzan; and clones 605 (Spi-1/PU.1 transfected) or 615 (control Neo^r cells) derived from the DS19 cell line by A. Skoultchi and G. Rao (39). All cell lines were cultured in Iscove modified Eagle medium (IMEM; GIBCO) supplemented with 10% heat-denatured fetal calf serum (FCS; Boehringer Mannheim) and penicillin-streptomycin (GIBCO) at 37°C and 5% CO₂. HMBA (Sigma) was added at 5 mM (final concentration) to induce erythroid differentiation of SFFV-transformed cell lines. Erythroid differentiation was determined by counting the percentage of hemoglobin-producing cells following staining with by acidic benzidine reagent.

Transient transfections. The 270/41 (by reference to base A of the ATG initiator of the Fli-1 protein taken as +1) sequence of the mouse Fli-1 gene was amplified by PCR and subcloned in the sense orientation immediately upstream of the luciferase (luc) coding sequence into vector pGL2-basic (Promega). Mutated versions of the resulting construct were obtained either by PCR or by the use of restriction sites. All constructs were verified by sequencing before transfection. For cotransfection experiments, the murine Spi-1/PU.1 or Fli-1 cDNA was subcloned under the control of the cytomegalovirus (CMV) promoter into plasmid pHOOK-3 (Invitrogen). Transfections of 745-A cells were performed with DAC-30 (Eurogentec). Briefly, 10⁶ cells were reseeded in 1 ml of 10% FCS-IMEM, and 1 ml of IMEM (without serum and without antibiotics) containing 2 µg of test plasmid, 0.5 µg of CMV-gal plasmid, and 15 µg of DAC-30 (Eurogentec) was added to the cells for 6 h at 37°C in 5% CO₂; 1 ml of 20% FCS-IMEM medium was then added, and cells were cultured for a further 24 h. Luciferase and -galactosidase activities were determined on cell lysates by luminometry (Victor 1420; Wallac), using luciferase and -galactosidase kit assays purchased from Boehringer Mannheim. Final results (except for Fig. 9) of luciferase activities were corrected for -galactosidase activities.

Stable transfections. Vector pEF-LAC-Fli was obtained by substituting the chloramphenicol acetyltransferase cassette in vector pEF-LAC-CAT (12) (kindly provided by H. Itoh) with the murine Fli-1 cDNA under control of the pEF-1 promoter. Stable transfection of 745-A cells was performed with 2 µg of DNA and DAC-30 as described for transient transfection; 24 h following the addition of DNA, transfected cells were cloned by limiting dilution into 96-well plates and selected in the presence of G418 (1 mg/ml; GIBCO). Vector pMTCI-Fli was obtained by subcloning the murine Fli-1 cDNA under control of the mouse metallothionein-1 promoter into vector pMTCI (a gift from F. Grigniani). Vector pMTCI is derived from the original vector pHEB0MT (17) in which a 513-bp cassette including the chimeric intron, multiple cloning site, and simian virus 40 late poly(A) signal isolated from vector PCI (Promega) was inserted into the single BamHI site downstream of the mouse metallothionein 1 promoter. Stable transfection of 745-A was performed as described for vector pEF-LAC-Fli except that selection was done in the presence of hygromycin B (1 mg/ml; GIBCO). Induction of Fli-1 expression in clones stably transfected by vector pMTCI Fli was obtained by the addition of 200 µM ZnCl₂.

	RESULTS

Top Abstract Introduction Materials and methods Results Discussion References

Fli-1 transcripts and proteins are expressed at comparable levels in erythroleukemic cell lines carrying a Spi-1- or Fli-1-activated locus. Different erythroleukemic cell lines established from mice infected by either the SFFV/F-MuLV complex (CTFP10, 745-A, and 28.9) or F-MuLV virus alone (CB7, IW32, and NN10) were collected from different laboratories. Total RNAs were prepared and analyzed by Northern blotting using Spi-1/PU.1 and Fli-1 cDNA probes sequentially. As expected, Spi-1/PU.1 transcripts were detected in all cell lines established from mice infected with the SFFV/F-MuLV complex but not in cell lines established from mice infected with F-MuLV (Fig. 1A, middle). Similarly, Fli-1 transcripts were detected in all cell lines established from mice infected with F-MuLV. However, Fli-1 transcripts were also detected in all cell lines established from SFFV-infected mice (Fig. 1A, top). Accurate analysis of Fli-1 transcripts detected on Northern blots after longer times of gel electrophoresis revealed that all tested SFFV-transformed cell lines contained a single 4-kb Fli-1 transcript whereas all three F-MuLV cell lines examined displayed a larger Fli-1 RNA signal (4 to 4.2 kb) made of at least two major size-related transcripts (Fig. 1A, top). By Western blot analysis, two Spi-1 proteins, a 44-kDa species and a 37-kDa species, were detected in all SFFV-transformed cell lines but not in F-MuLV-transformed cell lines (Fig. 1B, middle). Two Fli-1 proteins, a 51-kDa species and a 48-kDa species, were detected in all tested cell lines (Fig. 1B, top). Overall, these data indicated that both F-MuLV- and SFFV-transformed erythroleukemic cell lines express the Fli-1 gene. Interestingly, although the levels of Fli-1 expression varied among the different cell lines, they did not appear to be correlated with cell line origin (for instance, the SFFV-transformed cell line 745-A [lane 4] displayed levels of Fli-1 expression similar to that displayed by the F-MuLV cell line IW32 [lane 2]).

View larger version (46K): [in this window] [in a new window]   FIG. 1.   Fli-1 and Spi-1 gene expression in erythroleukemic cell lines established from mice infected by F-MuLV or SFFV. (A) Total RNA (10 µg) prepared from the indicated cell lines was subjected to electrophoresis, transferred to a membrane, and successively hybridized with the Fli-1 (top), Spi-1/PU.1 (middle), and 18S rRNA (bottom) radiolabeled probes. (B) Western blot analysis was performed with 20 µg of total cell proteins of the indicated cell lines by using anti-Fli-1 (top) or anti-Spi-1 (middle) antibodies. Photograph of a duplicate of the gel stained by Coomassie blue is shown (bottom) as a control for equal loading of all cell protein samples.

Fli-1 gene transcription is initiated at two different alternative regions in SFFV- and F-MuLV-transformed cells. We next tested whether the different Fli-1 transcripts observed by Northern blot analysis were differently spliced transcripts or reflected the presence of different transcription initiation/termination sites within the Fli-1 gene. RNase protection experiments using antisense RNA probes from different coding regions of the Fli-1 mRNA revealed no detectable difference between transcripts present in SFFV- and F-MuLV-derived cell lines (data not shown). We therefore tested whether the 4- to 4.2-kb Fli-1 transcripts present in F-MuLV cells differ in their 5' or 3' untranslated ends. We first performed 5' RACE analysis using the NFli3' oligonucleotide and poly(A)⁺ mRNA isolated from either CB7 (F-MuLV-transformed) or 745-A (SFFV-transformed) cells. Two major fragments of 460 and 270 bp were amplified from CB7 RNA, whereas only the 270-bp fragment was obtained from 745-A RNA (Fig. 2A). Cloning and sequencing of these cDNAs revealed that the shorter 270-bp fragments amplified from F-MuLV- or SFFV-transformed cells were identical. The longer 460-bp fragment present in F-MuLV cells contained the whole sequence of the 270-bp fragment and additional sequences from upstream mouse genomic sequence (Fig. 3). The same labeled oligonucleotide NFli3' was also used in primer extension experiments performed on total RNA from several other F-MuLV- or SFFV-transformed cells (Fig. 2B). Several Fli-1 extension products extending up to the 200 region were detected in all SFFV-transformed cells (Fig. 2B, lanes 4 to 10), while additional prominent longer extension products up to the 400 region were specifically observed in all three tested F-MuLV-transformed cell lines (Fig. 2B, lanes 1 to 3). Taken together, these data indicated that two types of transcripts, initiated either in the 200 or the 400 region, were differently produced in SFFV- and F-MuLV-transformed cells.

View larger version (79K): [in this window] [in a new window]   FIG. 2.   Analysis of the Fli-1 transcription initiation site(s) by 5' RACE (A) and primer extension (B) techniques, using primer NFli3'. (A) Electrophoretic analysis of 5' RACE reaction products (see Materials and Methods) obtained by using poly(A)+ mRNA prepared from CB7 or 745-A cells. Arrowheads indicate the main reaction products. Sizes of molecular weight markers are shown in base pairs on the left. (B) Electrophoretic analysis of extension products obtained from total RNA prepared from the indicated cell lines. The relative (to the ATG codon) positions of main extension products in F-MuLV- and SFFV-transformed lines are shown on the right. Sizes of molecular weight markers (pUC19/HpaII) are shown in base pairs on the left.

View larger version (27K): [in this window] [in a new window]   FIG. 3.   DNA sequence of the 5' region of the mouse Fli-1 gene (1). Coordinates are shown relative to ATG initiation codon of the 51-kDa Fli-1 protein. The primers used to characterize the 5' ends of Fli-1 transcripts are shown in boldface. The 5' ends of major 5' RACE products are indicated by unfilled and filled stars for the 745-A and CB7 cell lines, respectively. Major and minor transcription initiation sites further characterized by S1 and RNase mapping are indicated by large and thin arrowheads, respectively.

View larger version (52K): [in this window] [in a new window]   FIG. 4.   Analysis of Fli-1 transcription initiation sites by S1 nuclease protection assay (A) or RNase mapping (B and C) in transformed and normal murine cells. (A) Electrophoretic analysis of S1-protected fragments obtained from total RNA of F-MuLV-transformed CB7 cells and 5'-end-labeled DNA probe 1 (1298/156) (lane 3). The Maxam-Gilbert C+T and G+A ladders of the same probe are shown in lanes 1 and 2, respectively. Positions of C and T nucleotides in the mouse Fli-1 sequence (Fig. 3) are shown on the left. (B) Electrophoretic analysis of RNase-protected fragments detected with antisense RNA probe 2 (1298/144) and total RNA prepared from SFFV-transformed 745-A cells (lane 3) and F-MuLV-induced CB7 cells (lane 4). (C) Electrophoretic analysis of RNase-protected fragments detected with antisense RNA probe 3 (262/+52) and total RNA prepared from the F-MuLV (lanes 2 to 4)- or SFFV (lanes 5-7)-transformed cells and normal mouse spleen (lane 8). The coordinates of major protected fragments on the Fli-1 sequence are shown on the right.

View larger version (15K): [in this window] [in a new window]   FIG. 5.   Map of Fli-1 transcription initiation sites determined by 5' RACE, S1 nuclease protection, or RNase protection assays in F-MuLV-transformed, SFFV-transformed lines, or normal mouse spleen cells. The 5' region of the Fli-1 gene is schematically shown at the top. The relative abundance of the transcripts initiated at these different sites in tested cells is illustrated by the thickness of the vertical arrows. Ex1, exon 1 coding sequence.

Spi-1/PU.1 activates Fli-1 transcription initiation at the 204 site in SFFV lines. To explain the differences in pattern of Fli-1 transcripts in SFFV- and F-MuLV-transformed cell lines, we hypothesized that overexpression of Spi-1/PU.1 protein enhanced Fli-1 gene transcription from the 204 site in such cells. If this is the case, any variation of Spi-1/PU.1 level would result in a corresponding change in Fli-1 gene transcription initiated at the 204 site in SFFV-transformed cells.

View larger version (41K): [in this window] [in a new window]   FIG. 6.   Kinetic analysis of Fli-1 transcript and Spi-1/PU.1 protein levels in HMBA-treated 745-A cells. (A) RNase protection assay using a mixture of the antisense Fli-1 RNA probe 3 (262/+52) and the antisense RNA probe HPRT. The protected fragments corresponding to the major Fli-1 transcripts initiated at positions 204 and 398 as well as the HPRT transcript are indicated on the right. (B) Western blot analysis of Spi-1/PU.1 proteins. The positions of major protein bands (determined by comparison with molecular weight standards [not shown]) are indicated on the right. The time of HMBA (5 mM) treatment is indicated at the top. A photograph of the upper part of the gel stained by Coomassie blue is shown (bottom) as a control for equal loading of protein samples.

View larger version (50K): [in this window] [in a new window]   FIG. 7.   Quantitative analysis of 204 and 398 Fli-1 transcripts and Spi-1/PU.1 DNA binding activity in Spi-1/PU.1-overexpressing DS19/605 (lanes 1 and 2) and control DS19/615 (lanes 3 and 4) cells treated (lanes 2 and 4) or not treated (lanes 1 and 3) for 3 days with 5 mM HMBA. (A) Top, autoradiogram of gel with protected Fli-1 and HPRT fragments (internal control) detected by RNase protection assay using antisense Fli-1 RNA probe 3 (262/+52) and antisense HPRT probe. Fli-1 and HPRT transcripts corresponding to detected protected fragments are indicated on the right. Bottom, quantitation of 204 and 398 Fli-1 transcripts. The signals corresponding to 204, 398 Fli-1, and HPRT transcripts were quantified by phosphorimager analysis. The relative abundance of 204 Fli-1 transcripts (to HPRT transcripts) in untreated control DS19/615 cells was arbitrarily chosen as 100. (B) Top left, EMSA of Spi-1/PU.1 DNA binding activity, using labeled E74 probe and equal amounts of protein nuclear extracts prepared from the indicated cells. The position of Spi-1/PU.1 complex, identified by Spi-1/PU.1 antiserum (Fig. 11D), is indicated on the right. Top right, EMSA of GATA-1 DNA binding activity, using labeled GATA-1 probe and equal amounts of protein nuclear extracts prepared from the indicated cells. The position of the GATA-1 complex is indicated on the right. Bottom, the signals corresponding to Spi-1/PU.1 and GATA-1 complexes were quantified by phosphorimager analysis. The relative abundances of Spi-1/PU.1 complexes, standardized to the corresponding abundances of GATA-1 complexes, are reported as percentages of the Spi-1/GATA-1 ratio determined for untreated control DS19/615 cells, taken as 100%.

Characterization of a Spi-1/PU.1-responsive promoter in the mouse Fli-1 gene. Recently the nucleotide sequence of the 5' noncoding region of the Xenopus laevis Fli-1 gene was determined (30). Alignment of this sequence to known 5' sequences of the mouse and human Fli-1 genes (1) revealed two highly conserved blocks (more than 70% homology among the three species) around nucleotides 500 and 200 of the mouse sequence (Fig. 8 and data not shown). Interestingly, this 200 conserved region includes several consensus motifs for known transcription factors of the GATA, IRF, and ETS families. Among these consensus DNA binding sites, two EBSs are perfectly conserved among human, mouse, and Xenopus sequences. Since both EBSs are located just upstream to the Spi-1/PU.1-regulated 204 Fli-1 transcription initiation site, we tested the functionality of these sites. We fused the conserved 270/41 region of the mouse Fli-1 gene, containing both EBSs, to a luc reporter gene and transfected the resulting reporter construct into either untreated or HMBA-treated 745-A cells (Fig. 9). In untreated cells, the tested construct expressed a significant reporter activity which was 25- to 30-fold higher (Fig. 9, lane 2) than that obtained with the promoterless vector (Fig. 9, lane 1). This activity was stimulated by cotransfecting a Spi-1/PU.1 expression vector, and the extent of stimulation increased with the amount of cotransfected plasmid (0.5 or 1 µg) (Fig. 9, lanes 3 and 4). In HMBA-treated cells, we observed a significant reduction of Fli-1 promoter activity (Fig. 9, lane 5) which was rescued by cotransfecting the Spi-1/PU.1 expression vector (Fig. 9, lanes 6 and 7). Importantly, transfection of the tk-luc reporter construct in the same experiments revealed that the tk promoter activity is affected neither by HMBA nor by cotransfected Spi-1 expression vector, indicating the specific property of the 270/41 promoter (Fig. 9, lanes 8 to 11). These results establish that the 270/41 region of the Fli-1 gene contains promoter activity which can be modulated by Spi-1/PU.1, thus reproducing the regulation of endogenous 204 Fli-1 transcripts by Spi-1/PU.1 (see above).

View larger version (25K): [in this window] [in a new window]   FIG. 8.   Phylogenetic conservation of the 200 region of the Fli-1 gene shown by the alignment of 5' sequences of the mouse (1), human (1), and Xenopus Fli-1 (30) genes. The region spanning bp 280 to 180 of mouse and human Fli-1 genes, corresponding to bp 210 to 110 bp of the Xenopus homologous gene, is shown. Conserved sequences are boxed. Consensus DNA binding sites for GATA, ETS, or IRF family transcription factors are indicated above sequences by horizontal arrows. The position of the major transcription initiation 204 site found in SFFV-transformed cells is shown by a vertical arrow. Oligonucleotides (probes) used in EMSA (see text) are shown below the sequences.

View larger version (35K): [in this window] [in a new window]   FIG. 9.   Spi-1/PU.1-regulated promoter activity of the 270/41 region of the mouse Fli-1 gene in 745-A cells. 745-A cells (106) were transfected with equal amounts (2 µg) of either pGL2-basic, tk-luc, or 270/41 Fli-1 luc reporter construct together with the indicated amount of either Spi-1 expression vector pHOOK-Spi-1 or empty vector pHOOK in the presence (hatched boxes) or absence (empty boxes) of 5 mM HMBA as indicated. Luciferase activities were determined on lysates prepared from equal numbers of cells from each lot of transfected cells 24 h following the addition of DNA. The experiment was repeated three times, and the results are given as means of relative luciferase activities standardized for each repetition by the activity (taken as 100%) obtained for cells transfected with the 270/41 Fli-1 luc construct in the absence of both pHOOK-Spi-1 and HMBA (lane 2).

View larger version (22K): [in this window] [in a new window]   FIG. 10.   Identification of sequences contributing to promoter activity of the 270/41 mouse Fli-1 region. Reporter constructs containing the deleted (top) or mutated (in indicated sites; bottom) version of the 270/41 Fli-1 region were transfected into 745-A cells. The luciferase (Luc) signals were measured and normalized by -galactosidase activities of a cotransfected CMV-gal construct. The normalized signal of the parental 270/41 reporter construct was arbitrarily chosen as 100. Representative data from three different experiments are shown.

In vitro binding of Spi-1/PU.1 protein to conserved EBSs of the 270/41 Fli-1 promoter. To test whether Spi-1/PU.1 actually interacts with either or both of the two EBSs in vitro, different oligonucleotides spanning these two binding sites were used in EMSAs (Fig. 8). As shown in Fig. 11A, using a radiolabeled oligonucleotide spanning the 5' EBS and 745-A nuclear cell extracts, we observed three shifted complexes, C1 to C3 (Fig. 11A, lane 1). These complexes were specific because they were competed by an excess of unlabeled 5' EBS probe (Fig. 11A, lane 2). Specifically, the faster-migrating complex C3, but not C1 or C2, was competed by an excess of a cold oligonucleotide that binds Spi-1/PU.1 (40) (Fig. 11A, lane 3). No competition was observed in the presence of an excess of the same oligonucleotide mutated in the Spi-1/PU.1 binding site (Fig. 11A, lane 4). Moreover, addition of Spi-1/PU.1 antiserum, but not preimmune serum, to binding reactions specifically eliminated complex C3 (compare lanes 5 with lanes 6 and 7). Complex C3 was not observed in EMSAs using 745-A extracts and the mutated 5' EBS (GGAAGGTA) as a probe (Fig. 11B; compare lanes 4 and 5). Thus, we conclude that the conserved 5' EBS can bind Spi-1/PU.1 protein in vitro. When using HMBA-treated 745-A cell extracts, we found a drastic decrease of Spi-1/PU.1 binding to the 5' EBS (Fig. 11A; compare lanes 9 and 10). This decrease correlated with the decrease of Spi-1/PU.1 protein detected by Western blot analysis (Fig. 6B).

View larger version (51K): [in this window] [in a new window]   FIG. 11.   In vitro binding of Spi-1/PU.1 to EBSs of the 270/41 Fli-1 region. Radiolabeled oligonucleotides spanning the 5' EBS (A and B) or 3' EBS (C) of the 270/41 Fli-1 region (Fig. 8), or radiolabeled oligonucleotide E74 containing a consensus EBS (D), were tested in EMSA using nuclear extracts prepared from 745-A cells treated or not with 5 mM HMBA as indicated. The presence (+) and/or absence () of the indicated specific or nonimmune serum or excess of unlabeled oligonucleotides (competitors) is indicated. Major observed complexes C1 to C3 (in panel A) or C (in panel C) are indicated by arrows. The identified Spi-1/PU.1 and Fli-1 complexes are marked. In panel D, two different preparations of 745-A cell extracts were used in EMSAs shown in lanes 1 to 3 and 4 to 9. Free probe is not visible in panel D. The sequences of the different oligonucleotides used either as probes or as competitors are given in Fig. 8 and in Materials and Methods.

View larger version (51K): [in this window] [in a new window]   FIG. 12.   Deregulated overexpression of Fli-1 in 745-A cells inhibits their HMBA-induced erythroid differentiation. 745-A cells were stably transfected with empty vector pEF-LAC-CAT, Fli-1 constitutive expression vector pEF-LAC-Fli (A to C), or Fli-1 Zn-inducible Fli-1 expression vector pMTCI Fli (D to F), as indicated above the lanes, and individual clones were selected and amplified under appropriate antibiotic selection (G418 [1 mg/ml] for pEF-LAC-CAT or pEF-LAC-Fli and hygromycin [1 mg/ml] for pMTCI Fli). Individual clones or untransfected cells were grown for 3 (A to C) or 4 (D to F) days in the presence or absence of 5 mM HMBA with or without 200 µM ZnCl2 as indicated. The percentage of benzidine-positive differentiated cells was then determined, and total RNA and total protein cell lysates were prepared for Fli-1 RNA and protein analyses, respectively. (A and D) Northern blot sequentially hybridized with the Fli-1 probe (top) and the 18S rRNA probe (bottom). The positions of endogenous (En.) and exogenous (Ex.) Fli-1 transcripts are indicated on the right. (B and E) Results of Western blot analysis of Fli-1 proteins (top) and Grb2 proteins revealed on the same blot and taken as an internal standard (bottom). (C and F) Results of quantitative analyses of Fli-1 protein levels performed by densitometric tracing of the autoradiographs shown in panels B and E, respectively. Results are expressed as percentages of the Fli-1/Grb2 ratio determined for pEF-LAC-CAT1-transfected (C) and untransfected (F) 745-A cells, respectively. Numbers in parentheses indicate percentages of benzidine-positive cells.

Deregulated overexpression of Fli-1 inhibits HMBA-induced differentiation of SFFV-transformed cells. It has been shown recently that deregulated overexpression of Spi-1/PU.1 proteins inhibits HMBA-induced erythroid differentiation in SFFV-transformed erythroleukemic cell lines (39, 49). Our finding that in these cells Fli-1 is also highly expressed and that transcription of the Fli-1 gene is under positive control by Spi-1/PU.1 raised the possibility that deregulated expression of Fli-1 in Spi-1/PU.1-overexpressing SFFV cells is responsible for inhibition of differentiation. Indeed, the resistance of Spi-1/PU.1-overexpressing SFFV-transformed DS19/605 cells to HMBA-induced erythroid differentiation correlates with the maintenance of elevated levels of 204 Fli-1 transcripts (Fig. 7A) as well as Fli-1 protein (data not shown). To test this hypothesis directly, we determined whether overexpression of a Fli-1 transgene in SFFV-transformed cells affected their capacity to undergo erythroid differentiation. The murine Fli-1 cDNA was cloned into the pEF-LAC vector (12), which places Fli-1 expression under the control of the human elongation factor 1 promoter. Control experiments established that pEF-LAC-Fli produces high levels of Fli-1 proteins when transiently transfected into 3T3 cells (data not shown). SFFV-transformed cells from cell line 745-A were transfected with pEF-LAC-Fli or pEF-LAC-CAT (control vector harboring a CAT cDNA in place of the Fli-1 cDNA), and several G418-resistant clones were selected. Among 10 individual G418-resistant clones obtained after transfection with pEF-LAC-Fli, we identified two clones, pEF-LAC-Fli1 and pEF-LAC-Fli2, expressing high levels of unrearranged exogenous Fli-1 transcripts of the expected length (Fig. 12A, lanes 3 to 6). Interestingly, these two clones were the only ones which reproducibly displayed reduced numbers of benzidine-positive cells (21 to 23%) after 3 days of HMBA treatment. By comparison, nonexpressing clones and clones transfected with control vector pEF-LAC-CAT produced 60 to 70% benzidine-positive cells after 3 days of HMBA treatment. We then compared the level of Fli-1 protein expressed in these two poorly differentiating clones to that expressed in two randomly chosen, well-differentiating control clones, pEF-LAC-Fli7 and pEF-LAC-CAT1. This comparison was done by Western blot analysis using the Grb2 protein signal revealed on the same blot as an internal standard (Fig. 12B). Quantitative analysis of the results indicated that the two poorly differentiating clones pEF-LAC-Fli1 and p-EF-LAC-Fli2 express two- to threefold more Fli-1 protein than control cells both before and after HMBA treatment (Fig. 12C). Taken together, these results indicate that deregulated overexpression of Fli-1 protein in 745-A cells inhibits HMBA-induced erythroid differentiation. However, given that the level of Fli-1 protein in the overexpressing clones displaying reduced differentiation capacity is only two- to threefold higher than that of the endogenous protein (Fig. 12C), we decided to further confirm this finding by using an inducible Fli-1 expression vector. For that purpose, the murine Fli-1 cDNA was cloned into the vector pMTCI, which places Fli-1 expression under the control of the mouse methallothionein 1 promoter. We derived several hygromycin-resistant clones of 745-A cells transfected with the pMTCI Fli-1 expression vector. Among these clones, we identified one, pMTCI Fli11, which expresses detectable and inducible amounts of exogenous Fli-1 transcripts in the presence of HMBA (Fig. 12D, lanes 2 and 3). As expected, pMTCI Fli11 displayed a reproducible reduction of differentiation capacity (50% benzidine-positive cells) compared to the differentiation capacity of untransfected 745-A cells (86% benzidine-positive cells) after 4 days of HMBA treatment in the absence of Zn. More interestingly, the differentiation capacity of pMTCI Fli11 cells was further reduced threefold (19% of benzidine-positive cells) in the presence of Zn while the addition of Zn minimally affected the differentiation capacity of untransfected 745-A cells (80% benzidine-positive cells). By Western blot analysis, we found that the levels of Fli-1 protein in pMTCI Fli11 cells were three times higher than the levels in untransfected 745-A cells in the presence of HMBA and absence of Zn (Fig. 12E [compare lanes 2 and 5] and 12F) and that these levels further increased more than threefold in the presence of Zn (Fig. 12E [compare lanes 2 and 3] and 12F), thus exceeding by 10-fold the levels observed in untransfected cells (Fig. 12E [compare lanes 3 and 6] and 12F). Taken together, these results establish that deregulated overexpression of Fli-1 inhibits HMBA-induced differentiation of 745-A cells and that this inhibition is proportional to the level of Fli-1 protein produced.

	DISCUSSION

Top Abstract Introduction Materials and methods Results Discussion References

The present data reveal that deregulation of Fli-1 gene expression is a common property of erythroleukemic cell lines derived from mice infected with either the Friend viral complex (SFFV plus F-MuLV) or F-MuLV alone. Moreover, we provide evidence that the ETS transcription factor Spi-1/PU.1, which is overexpressed in SFFV-derived erythroleukemic cell lines, triggers expression of the Fli-1 gene through the activation of a new Fli-1 promoter located at position 204 from the translational start site. Therefore, Fli-1 gene transcription is regulated by at least two different promoters, one of which is positively regulated by the ETS family transcription factor Spi-1/PU.1 in SFFV-derived erythroleukemic cells. Indeed, we showed that in SFFV-derived cells, modulation of Spi-1/PU.1 levels (by HMBA treatment or stable Spi-1/PU.1 transgene expression) results in coordinate variations in 204 Fli-1 transcript levels. This new transcription initiation site is located in a DNA region (270/41) which is well conserved among the human, mouse, and frog Fli-1 genes and which exhibited promoter activity (when linked to a reporter gene) in transient transfection assays. Importantly, we demonstrated that the promoter activity of this Fli-1 region was completely dependent on the integrity of two conserved EBSs (Fig. 10), which preferentially interact with Spi-1/PU.1 in vitro (Fig. 11 and data not shown). In addition, we showed that in transient transfection assays, this promoter region was transactivated when cotransfected with a Spi-1/PU.1 expression vector (Fig. 9). Thus, these data clearly establish the positive transcriptional regulation of the Fli-1 gene by the ETS family transcription factor Spi-1/PU.1 through the specific EBS-dependent promoter in SFFV-transformed cells.

Intriguingly, we also observed that, in 745-A cells, the level of 398 Fli-1 transcripts are markedly upregulated during HMBA treatment (Fig. 6A), suggesting that Spi-1/PU.1 may be simultaneously involved in negative control of the upstream promoter. However, in the other studied SFFV-derived cell line, DS19, the levels of the 398 Fli-1 transcripts are only slightly upregulated by the downregulation of Spi-1/PU.1 levels induced by HMBA treatment and also only slightly downregulated by the upregulation of Spi-1/PU.1 levels following stable transfection of Spi-1/PU.1 expression vector (Fig. 7A). Thus, although these data do not exclude the possibility that Spi-1 is simultaneously involved in negative control of the upstream promoter, the difference in the regulation of 398 transcripts observed between 745-A and DS19 cells suggests that the upstream promoter is also most probably regulated by other factors, the expression of which is differentially regulated by HMBA in 745-A cells or in DS19 cells.

Specificity of the Spi-1/PU.1Fli-1 regulatory cascade. To our knowledge, our results are among the first data demonstrating that one member of the ETS family of transcription factors is involved in the transcriptional regulation of another. Although Spi-1/PU.1 enhanced the expression of Fli-1 in SFFV-derived cells, no increase of other ETS family proteins, such as Ets1, Ets2, Elf-1, or GABP/, could be detected in EMSAs compared to levels in F-MuLV-transformed cells and other hematopoietic cell lines (data not shown). Thus, the Fli-1 gene seems to be specifically activated by Spi-1/PU.1 in SFFV-transformed cells. These results are interesting in light of the presence of functional EBSs in promoters of a number of ETS factors, including those tested above (2, 8). This suggests a high specificity of the Spi-1Fli-1 regulatory cascade, likely due to sequences surrounding the EBSs and/or Spi-1-associated transcription factors which may contribute to the unique 270/41 Fli-1 promoter transactivation by Spi-1 in SFFV-transformed cells.

Role of the Spi-1/PU.1Fli-1 regulatory cascade in erythroleukemia. Our present findings reveal that deregulation of Fli-1 gene expression is a common event in transformation of murine erythroid progenitor cells induced either by the Friend viral complex or by F-MuLV alone. Although the target genes of Fli-1 involved in transformation of erythroid progenitors remain unknown, the recent demonstration that Fli-1 overexpression inhibits apoptosis (50), together with our observations that it inhibits erythroid differentiation, indicate that Fli-1 is involved in the control of a key transformation event. The same antiapoptotic or antidifferentiating effects were described previously for avian primary erythroblasts (38) and SFFV-derived erythroleukemic cells overexpressing Spi-1/PU.1 (39, 49). In light of our finding of activation of Fli-1 in Spi-1/PU.1-overexpressing cells, it is highly probable that at least some of the effects earlier ascribed to Spi-1/PU.1 are indeed mediated by Spi-1/PU.1-activated Fli-1. Recent data support and extend this hypothesis. Indeed, it was found that overexpression of the N-terminal part of Fli-1 inhibited the retinoic acid-induced differentiation of myeloblastic leukemia HL-60 cells. Furthermore, this effect correlated with a negative effect of Fli-1 on transcription of retinoic acid-activated genes, and complex formation between Fli-1, hormone receptor, and unidentified partner was detected in vitro (9). Negative interference with the nuclear hormone response has already been described for Spi-1/PU.1 (16), raising the possibility that this Spi-1/PU.1 effect is mediated indirectly by activation of the Fli-1 gene. Interestingly, the v-erbA oncogene harbored by avian erythroleukemia virus, which induces erythroleukemia in chickens, is known to encode a dominant negative version of the normal thyroid hormone receptor, further suggesting that inhibition of the nuclear hormone response may be a general mechanism leading to erythroleukemia induced by either the Friend viral complex, F-MuLV, or avian erythroleukemia virus.

Biological relevance of the Spi-1/PU.1Fli-1 regulatory cascade. In the present study, we also show that the 204 transcription initiation site of the Fli-1 gene is also used in normal spleen cells, indicating that its use is not an abnormal property of Friend erythroleukemic cells. Both Spi-1/PU.1 and Fli-1 are known to be expressed in multiple hematopoietic lineages. Given the highly pleiotropic effect of the inactivation of Spi-1/PU.1 gene on mouse hematopoiesis (14, 31, 44, 45), our finding of a positive regulation of Fli-1 by Spi-1/PU.1 raises the possibility that some of the normal functions of Spi-1/PU.1 in hematopoiesis can be mediated by its effect on Fli-1 gene regulation. There are at least three hematopoietic lineages in which Spi-1/PU.1 and Fli-1 expression patterns overlap. The first one is the megakaryocytic lineage, in which high levels of both Spi-1/PU.1 and Fli-1 transcripts have been found (24, 32). Interestingly, in a recent study (11), we observed that the stimulation of the erythromegakaryocytic UT7-mpl cells by thrombopoietin is associated with a sequential increase in Spi-1/PU.1 and Fli-1 protein levels. It is therefore tempting to speculate that the Spi-1/PU.1Fli-1 regulatory cascade identified in the present study might belong to a normal transduction regulatory cascade. Spi-1/PU.1 and Fli-1 expression patterns also overlap in the B-lymphoid and the myeloid lineages (reviewed in reference 14). It will be interesting to investigate whether Spi-1/PU.1 can also regulate the expression of Fli-1 in these two other lineages and to determine if Spi-1/PU.1 and Fli-1 contribute to a sustained inhibition of erythroid commitment of myeloid or lymphoid progenitors.