Despite the major functions of the basic helix–loop–helix transcription factor TAL‐1 in hematopoiesis and T‐cell leukemogenesis, no TAL‐1 target gene has been identified. Using immunoprecipitation of genomic fragments bound to TAL‐1 in the chromatin of murine erythro‐leukemia (MEL) cells, we found that 10% of the immunoselected fragments contained a CAGATG or a CAGGTG E‐box, followed by a GATA site. We studied one of these fragments containing two E‐boxes, CAGATG and CAGGTC, followed by a GATA motif, and showed that TAL‐1 binds to the CAGGTG E‐box with an affinity modulated by the CAGATG or the GATA site, and that the CAGGTG–GATA motif exhibits positive transcriptional activity in MEL but not in HeLa cells. This immunoselected sequence is located within an intron of a new gene co‐expressed with TAL‐1 in endothelial and erythroid cells, but not expressed in fibroblasts or adult liver where no TAL‐1 mRNA was detected. Finally, in vitro differentiation of embryonic stem cells towards the erythro/megakaryocytic pathways showed that the TAL‐1 target gene expression followed TAL‐1 and GATA‐1 expression. These results establish that TAL‐1 is likely to activate its target genes through a complex that binds an E‐box–GATA motif and define the first gene regulated by TAL‐1.
The tal‐1 gene (also known as scl or TCL‐5), was identified via a chromosomal translocation involving human chromosome 1p32 in T‐cell acute lymphocytic leukemia (T‐ALL) (Begley et al., 1989; Finger et al., 1989; Bernard et al., 1990; Chen et al., 1990). Further studies have shown that aberrant activation of the tal‐1 gene occurs in most patients with T‐ALL (Bash et al., 1995) with diverse molecular mechanisms, including chromosomal translocations, interstitial deletions and transactivation without detectable chromosomal alteration. Finally transgenic mice have demonstrated a direct role of TAL‐1 in T‐cell leukemogenesis (Condorelli et al., 1996; Kelliher et al., 1996; Larson et al., 1996; Aplan et al., 1997). The normal function of TAL‐1 is the regulation of hematopoiesis (Green, 1996; Shivdasani and Orkin, 1996) and embryonic angiogenesis (Visvader et al., 1998). Mice homozygous for deletion of the tal‐1 gene die in utero by embryonic day 10 as a result of the absence of blood formation (Robb et al., 1995; Shivdasani et al., 1995), indicating that TAL‐1 has a critical function very early during hematopoietic differentiation. Furthermore, tal‐1 null embryonic stem (ES) cells failed to produce any hematopoietic lineage and to contribute to hematopoiesis in chimeric mice, implying that TAL‐1 is also critical for definitive hematopoiesis (Porcher et al., 1996; Robb et al., 1996). Finally, transgenic rescue of hematopoietic defects of tal‐1−/− embryos has established a role for TAL‐1 in angiogenesis (Visvader et al., 1998).
TAL‐1 belongs to the bHLH class of transcription factors (Baer, 1993). The bHLH domain allows both the formation of protein dimers and sequence‐specific DNA recognition, and bHLH dimers bind to E‐box motifs which have the general sequence CANNTG (Murre et al., 1989). TAL‐1 proteins cannot bind to DNA as homodimers but are able to bind DNA after dimerization with the ubiquitously expressed E2A gene products, E47 and E12 (Hsu et al., 1991, 1994b). CASTing experiments have shown that the TAL‐1/E2A heterodimers bind to the preferred sequence 5′‐AACAGATGGT‐3′ (Hsu et al., 1994a). Using an artificial reporter gene containing multiple copies of the TAL‐1/E2A binding sequence, transcriptional activity of the TAL‐1/E47 heterodimer was examined in transiently transfected murine fibroblasts. While the E47 homodimer strongly activated the reporter gene, the TAL‐1/E47 heterodimer was much less active, suggesting either a negative or a weak positive regulatory role of TAL‐1 (Hsu et al., 1994c).
The pivotal functions of TAL‐1 may be accomplished by DNA binding and tight co‐operation with at least two partners, the LIM‐only protein LMO2 and GATA‐1. TAL‐1 and LMO2 genes are co‐expressed during primitive and definitive hematopoiesis (Green et al., 1992; Kallianpur et al., 1994; Robb et al., 1996), are activated in T‐ALL (Begley et al., 1989; Boehm et al., 1991; Rabbitts, 1994), and the two proteins are functionally related. Furthermore, like TAL‐1, LMO2 is also a T‐cell oncogene when expressed as a transgene in mice (Boehm et al., 1991; Royer‐Pokora et al., 1991), and LMO2−/− and TAL‐1−/− mice display the same phenotype (Warren et al., 1994). The other potential partner of TAL‐1 is GATA‐1, a transcription factor present in all stages of vertebrate erythropoiesis. GATA‐1 binds to the WGATAR motif present in the promoters and enhancers of several erythroid‐specific genes (reviewed in Weiss and Orkin, 1995), and exerts positive or negative transactivation effects (Aird et al., 1994; Raich et al., 1995; Weiss and Orkin, 1995; Briegel et al., 1996). An essential role for GATA‐1 in erythropoiesis has been established through gene disruption. Indeed GATA‐1−/− ES cells fail to contribute to the mature erythroid compartment in chimeric mice (Pevny et al., 1991) while expression of a normal GATA‐1 transgene in the mutant ES cells rescues erythroid development (Simon et al., 1992). The absence of erythropoiesis obtained in the TAL‐1, LMO2 and GATA‐1 null mutations suggests that these three proteins have closely related roles in erythroid differentiation. As LMO2 can bind to both GATA‐1 and TAL‐1 in erythroid cell lines, it has been postulated that it may act as a physical bridge between TAL‐1 and GATA‐1 proteins to settle a multimeric complex (Osada et al., 1995). Recent CASTing experiments have defined a CAGGTG‐GATA bipartite DNA motif (Wadman et al., 1997) that binds a multimolecular complex including TAL‐1, E47, LMO2, GATA‐1 and the LIM‐domain binding protein Ldb1 (Agulnick et al., 1996; Jurata et al., 1996), and generates a positive transcriptional activity (Wadman et al., 1997). Therefore, co‐operation between TAL‐1, LMO2 and a member of the GATA family may be involved in the regulation of TAL‐1 target genes.
Despite its major role in hematopoiesis (Green, 1996; Shivdasani and Orkin, 1996) and leukemogenesis (Aplan et al., 1992a, 1997; Condorelli et al., 1996; Kelliher et al., 1996; Larson et al., 1996), as well as the identification of interacting partners (Valge‐Archer et al., 1994; Wadman et al., 1994; Osada et al., 1995), no TAL‐1 target gene has been identified so far. Furthermore, the sequence initially defined to bind TAL‐1 in vitro (Hsu et al., 1994a) and the newly identified in vitro target (Wadman et al., 1997) have not been found in regulatory regions of genes involved in hematopoiesis regulation. In order to identify in vivo TAL‐1 binding sites as well as target genes regulated by TAL‐1 protein, we set up a strategy based on immunoselection of DNA–protein complexes in chromatin, and we describe here one of the isolated target sequence that binds TAL‐1 and is located within an intron of a new gene whose expression is consistent with TAL‐1 regulation.
Isolation of in vivo TAL‐1 target genomic fragments by chromatin immunopurification
To identify TAL‐1 target sequences, we developed an in vivo DNA–protein immunopurification strategy that enriches for short fragments bound to TAL‐1 in chromatin (Figure 1). We chose the murine erythro‐leukemia (MEL) cell line for this immunopurification method for two reasons. First, MEL cells have been widely used to study the molecular and cellular mechanisms that regulate terminal erythroid differentiation (Visvader et al., 1991; Aplan et al., 1992b; Green et al., 1992; Murrell et al., 1995) and secondly, TAL‐1 has major functions in this cell line (Aplan et al., 1992b). We obtained 450 ng of immunoprecipitated DNA, starting from 3×108 cells, i.e. 0.00025% of the initial DNA, and the average insert size was 500 bp. One fifth of the immunoselected material was cloned and 1000 recombinant clones from the library were ordered, among which 60 were sequenced. As expected from the absence of known TAL‐1 target genes, none of the cloned fragments sequenced was recorded in sequence databases but interestingly, a high proportion (5/60, i.e. 8.3%) of potential TAL‐1 target sequences contained a GATA motif close to an E box of the CAGATG or the CAGGTG type (Table I). As TAL‐1 and GATA‐1 are considered to be functionally related in erythroid differentiation (for review see Shivdasani and Orkin, 1996), we reasoned that these fragments would be of the utmost interest as potential TAL‐1 targets in MEL cells. We therefore selected one 474 bp fragment, whose sequence is shown in Figure 2, for further studies for the following reasons. (i) It contained two E‐boxes in tandem spaced by two nucleotides, one of which is of the CAGATG type and the other of the CAGGTG type. Furthermore, the second E box (CAGGTG) lies 13 bp 5′ from a consensus GATA binding site. (ii) This fragment was located in a genomic region highly sensitive to micrococcal nuclease in MEL cells, indicating an open chromatin configuration around the immunoprecipitated region. The high micrococcal nuclease sensitivity of the immunoprecipitated fragment was assessed by comparison with the CD2 locus which is not transcriptionally activated in MEL cells, and needed 5‐ to 10‐fold more nuclease to be digested (data not shown). Therefore, this fragment was a good candidate as a TAL‐1 binding site, and we attempted first to determine if the region containing the two potential TAL‐1 binding sites and the GATA motif could be a regulatory region.
The immunoprecipitated fragment binds TAL‐1 in vitro
We studied the ability of the two E‐boxes to bind TAL‐1 using MEL nuclear extracts. A complex of similar mobility as the TAL‐1/E2A‐containing complex was observed when an oligonucleotide containing the two E‐boxes was used in electromobility shift assays (EMSAs) (Figure 3, lanes 1 and 3) and supershift experiments using an anti‐TAL‐1 monoclonal antibody showed that this complex actually contains TAL‐1 (Figure 3, lanes 2 and 4). To identify which E box binds TAL‐1, we first used oligonucleotides that contained only the CAGATG E‐box (Figure 3, lane 5) or the CAGATG E‐box followed by a mutated CAGGTG E‐box (Figure 3, lane 6). No TAL‐1‐containing complex could be obtained with these probes, suggesting that TAL‐1 binds to the CAGGTG E‐box. However, a 20 bp oligonucleotide (the length of the TAL‐1/E2A oligonucleotide) that contained the CAGGTG E‐box could not efficiently bind TAL‐1 (Figure 3, lane 7). As the E‐box1–E‐box2 oligonucleotide contained two more nucleotides at its 3′‐end, we used a 22 bp oligonucleotide that contained only the CAGGTG E‐box and the same 3′‐end nucleotides, and showed that this oligonucleotide binds TAL‐1 (Figure 3, lanes 8 and 9). Finally, we showed that an oligonucleotide containing the CAGGTG E‐box and a mutated CAGATG E‐box also binds TAL‐1 (Figure 3, lanes 10 and 11), demonstrating that, in vitro, TAL‐1 can bind the CAGGTG E‐box located in the immunoprecipitated fragment.
The GATA‐1 site increases TAL‐1 binding affinity
We then explored the role of the GATA motif located 13 bp downstream from the CAGGTG E‐box. Gel‐shift experiments showed that the probe containing the two E‐boxes and the GATA site generated numerous complexes (Figure 4A, lane 4). One complex contained TAL‐1, and TAL‐1 affinity for this composite sequence was higher than for the E‐box1–E‐box2 probe, and similar to the one observed for the TAL‐1 consensus probe (Figure 4A, lanes 2, 3 and 4). We determined whether the GATA binding site can modulate the TAL‐1 binding affinity to the CAGGTG E‐box. Competition experiments were performed using the TAL‐1/E2A oligonucleotide as a probe and increasing amounts of wild‐type or mutant E‐box1–E‐box2–GATA oligonucleotides as competitors. A typical result of such an experiment is shown in Figure 4B, and indicated that the E‐box1–E‐box2–GATA oligonucleotide (Figure 4B, lane 5) was more efficient in competition than the E‐box1–E‐box2–mutantGATA oligonucleotide (Figure 4B, lane 3) and less efficient than the mutant E‐box1–E‐box2–GATA oligonucleotide (Figure 4B, lane 8). Quantification of the TAL‐1‐containing complex indicated a ratio of 2 between the E‐box1–E‐box2–GATA and the E‐box1–E‐box2–mutantGATA oligonucleotides, and a ratio of 1.5 between the mutant E‐box1–E‐box2–GATA and the E‐box1–E‐box2–GATA oligonucleotides. Altogether these results showed that TAL‐1 binds to the CAGGTG E‐box and that the TAL‐1 binding affinity can be modulated positively (by the GATA binding site) or negatively (by the CAGATG E‐box).
TAL‐1 binds to the in vivo target as a part of a multicomponent complex present only in normal erythroid cells
As our observations of MEL cells might not reflect the TAL‐1 binding during hematopoiesis, we investigated the binding of nuclear proteins obtained from cultures of purified human hematopoietic progenitor cells (HPCs) in liquid suspension (Condorelli et al., 1995). In this system HPCs are induced to undergo erythroid or granulopoietic determination in the first culture week, and terminal differentiation in the second week. In all experiments, nuclear extracts prepared from cultures at day 9 were used. The E‐box1–E‐box2–GATA oligonucleotide generated multiple complexes that were detected with nuclear extracts obtained from erythroid cells, and not with nuclear extracts obtained from granulocytic cells, indicating specific erythroid complexes (Figure 5, compare lanes 1 and 2, and lanes 3 and 4). One of the complexes (indicated by an arrow in Figure 5) has a mobility similar to the complex previously described (Wadman et al., 1997) using a CAGGTG type E‐box followed by a GATA‐site oligonucleotide, and shown to contain LMO2, GATA‐1, TAL‐1, E2A and Ldb1 (Figure 5, lanes 1 and 3). Interestingly, no TAL‐1/E2A complex was observed, indicating that TAL‐1 is essentially tied up in the multicomponent complex present in erythroid cells and not in granulocytic cells.
The in vivo immunoprecipitated TAL‐1 target sequence behaves as a cis regulatory acting sequence with positive transcriptional activity
To determine if the immunoprecipitated TAL‐1 target sequence is a transcriptional regulatory region we used transient transfection experiments. Transcriptional activity of the E‐box1–E‐box2–GATA sequence cloned 5′ to the minimal promoter from the glycophorin B gene (Rahuel et al., 1992) was studied using the Dual Luciferase™ Reporter Assay System (Figure 6). Contrary to other studies that have used tandems of identical TAL‐1/E2A binding sites of the CAGATG type (Hsu et al., 1994c; Nielsen et al., 1996) or of the CAGGTG type followed by a GATA site (Wadman et al., 1997), we chose to work with only one copy of each oligonucleotide, which is closer to the in vivo context. Transcriptional activities of the wild‐type or mutated E‐box1–E‐box2–GATA–glycophorin B (GpB) luciferase reporter constructs were compared in erythrocytic (MEL), erythro/megakaryocytic (HEL) and non‐hematopoietic (HeLa) cells. No transcriptional activity was detected in HeLa cells that do not express TAL‐1, while similar transcriptional activities were obtained in MEL and HEL cells that express TAL‐1 (Figure 6) showing a correlation between the transcriptional activity of this sequence and TAL‐1 expression. The wild‐type E‐box1–E‐box2 GATA–GpB construct produced a 2‐fold increase of luciferase activity as compared with the GpB minimal construct, and mutation of E‐box1 produced a 4‐fold increase in luciferase activity indicating that the CAGGTG–GATA motif acts as a positive cis‐acting sequence whose function is negatively modulated by E‐box1 (Figure 6). Double mutations that abolished both the E‐box1 and E‐box2, or the E‐box1 and GATA binding sites resulted in background transcriptional activity demonstrating a co‐operativity between TAL‐1 and GATA‐1 (Figure 6). Similar results were obtained when oligonucleotides were inserted in opposite orientation indicating an orientation‐independent transcriptional activity (data not shown). These data showed that the region containing the two E‐boxes and the GATA site from the immunoprecipitated fragment behaves as a cis regulatory acting sequence with a positive transcriptional activity that can be tuned by the repressive activity mediated by E‐box1 (CAGATG).
The immunoprecipitated sequence is located within an intron of a new gene
Using the immunoprecipitated sequence as a probe, we screened a murine genomic library and got several phages that encompassed the probe. One of them, containing a 18 kb insert, was cut with various restriction enzymes and the DNA fragments that overlapped or lay around the immunoprecipitated sequence were used in an exon‐trapping assay. We found that a 5 kb NcoI–NcoI fragment (Figure 7A) that started within the immunoprecipitated sequence contained three exons. Neither the nucleotide sequence nor the polypeptide encoded by these exons (Figure 7A) were found in databases, indicating that these exons belong to a new transcription unit. However, the encoded polypeptide displayed a significant homology to the N‐terminus of the recently cloned otogelin (Cohen‐Salmon et al., 1997) i.e. it contained a cysteine‐rich domain that precedes the beginning of a von Willebrand Factor (vWF) D‐type domain (Figure 7A). Fine mapping of the three exons defined their genomic organization and located the immunoprecipitated sequence 56 bp 5′ from the first exon identified (Figure 7B). Finally, the sequence of the DNA that flanked the three exons showed intron–exon boundaries that are consistent with the GT‐AG rule applied to the splicing point (Figure 7B). Therefore, we conclude that the immunoprecipitated sequence is located in an intron of a new gene.
The expression of the identified gene is consistent with TAL‐1 regulation
The transcriptional status of TAL‐1 and the target gene we have identified was evaluated by reverse transcription‐coupled PCR (RT–PCR). We found a co‐expression of these two genes in MEL cells, fetal liver and in a murine stromal cell line, M2‐10 B4, that contains endothelial cells (Lemoine et al., 1988) (Figure 8A). On the other hand, the putative TAL‐1 target gene was not expressed in NIH 3T3 fibroblastic cell line or in adult liver where indeed no TAL‐1 mRNA could be detected (Figure 8A). Finally, we used an in vitro differentiation assay of ES cells towards the erythro/megakaryocytic pathways to study the expression of TAL‐1, LMO2, GATA‐1 and of the TAL‐1‐associated transcription unit. Gene expression in these cultures was also assayed by RT–PCR analysis at days 0, 3, 5, 6 and 9. As shown in Figure 8B, the TAL‐1 target gene started to be detected at day 6, i.e. 1 day after the appearance of TAL‐1, GATA‐1 and LMO2 mRNAs. Taken together, these data show that the transcription unit we have identified is expressed in a manner consistent with TAL‐1 regulation.
Immunoselection of TAL‐1 genomic binding sites
To identify TAL‐1 targets we set up an immunoprecipitation assay of TAL‐1 bound DNA in chromatin which may reflect the in vivo TAL‐1 DNA‐binding activity. As we and others have already discussed (Gould et al., 1990; Deveaux et al., 1997; Yamamoto et al., 1997), this method has several advantages as it does not select against low‐affinity binding sites, and may take into account the complexity provided by adjacent co‐operating binding sites. Immunoprecipitation of DNA–protein complexes has successfully identified in vivo target genes for several transcription factors in Drosophila (Gould and White, 1992; Graba et al., 1992) as well as in mammals (Bigler and Eisenman, 1995; Grandori et al., 1996; Deveaux et al., 1997), and it seems that the use of these in vivo strategies is of highest interest when addressing the question of relevant targets for transcription factors. Target sequences identified by immunoprecipitation were often different from those identified in vitro, with changes in core adjacent bases of the motif or even insertions in the core site. This was described for targets of UBX, the product of the Drosophila homeotic gene Ultrabithorax (Gould et al., 1990), thyroid hormone receptor (Bigler and Eisenman, 1994, 1995), Pax‐2 (Phelps and Dressler, 1996), MYC‐MAX (Grandori et al., 1996) and NF‐E2 (Deveaux et al., 1997) targets. Our work indicated that an E‐box–GATA association might be important for TAL‐1 function in vivo as ∼10% of the immunoprecipitated fragments showed a GATA binding site associated to an E‐box. This bipartite DNA motif, previously defined in vitro (Wadman et al., 1997), might thus represent an hallmark for genes regulated by both TAL‐1 and GATA proteins. Previous studies have shown that the association of GATA and Sp1, or GATA and Ets binding sites are respectively crucial for erythrocytic‐ (Frampton et al., 1990) or megakaryocytic‐ (Lemarchandel et al., 1993) specific expression, and the close association of TAL‐1 and GATA binding sites might now define a particular set of hematopoietic specific genes.
Identification of a functional in vivo TAL‐1 binding site
In order to define whether one of these immunoselected fragments represents a transcriptional regulatory sequence, we focused our study on a sequence that contains two E‐boxes followed 13 bp downstream by a GATA motif. The TAL‐1/E2A heterodimer could bind the CAGGTG E‐box, which is not the preferred in vitro TAL‐1/E2A consensus binding site initially defined (Hsu et al., 1994a), and its binding affinity could be modulated positively or negatively by adjacent DNA binding sites. Thus, CAGATG and CAGGTG are TAL‐1 binding sites depending on adjacent bases or on closely associated DNA‐binding site(s), and TAL‐1 binding affinity may be tuned by other transcriptional factors to achieve fine regulation of the corresponding target gene(s). Finally, and in agreement with a previous study (Wadman et al., 1997), we found that an oligomeric complex bound the CAGGTG motif followed 13 bp downstream by a GATA binding site.
Transfection experiments indicated that this complex behaved as a transcriptional transactivating complex in erythroid but not in HeLa cells, and that GATA‐1 and TAL‐1 are required simultaneously for efficient transcriptional activity. We also showed that the upstream E‐box acted as a negative cis‐acting sequence and thus could tune the TAL‐1‐containing complex transcriptional activity. One potential candidate for binding to this E‐box is the newly identified and broadly expressed MIST‐1 bHLH protein (Lemercier et al., 1997) which is able to bind the CAGATG sequence, is present in hematopoietic cells (C.Lemercier, personnal communication) and can act as a repressor (Lemercier et al., 1998). We tested MIST‐1 binding to E‐box1 and preliminary data indicated that both E‐box1 and E‐box1–E‐box2 oligonucleotides were able to bind MIST‐1 (data not shown), suggesting that MIST‐1 could be involved in the repressive activity mediated by E‐box1. A similar regulation has been described in the rat prothymosin‐α intron enhancer, where a second E‐box element, adjacent to the E‐box recognized by MYC, binds a putative repressor protein and is a critical determinant of specificity (Desbarats et al., 1996).
Characterization of a TAL‐1 target gene
The fragment we have immunopurified in MEL cells contains a composite E‐box1–E‐box2–GATA motif which can bind a TAL‐1‐containing complex present in normal erythroid cells and can activate the transcription of an indicator gene erythroid cell line. Therefore, this motif confers to the in vivo immunoselected fragment a potential regulatory transcriptional role mediated by TAL‐1 at least in the erythroid lineage. Using an exon‐trapping assay, we have isolated three consecutive exons, one of which was located 56 bp downstream of the immunoprecipitated region. The close association of the identified exons with the TAL‐1 binding region, together with the co‐expression of TAL‐1 and the identified gene in the tissues or cell lines tested, showed that this immunoselected region belongs to an intron of a novel gene potentially regulated by TAL‐1. The 107 amino acid polypeptide encoded by the three exons showed a striking similarity (51% homology on a stretch of 74 amino acids) with the N‐terminus of otogelin, a newly cloned secreted glycoprotein (2910 amino acid length) of the mucin‐related family, specifically expressed in the inner ear and of yet unknown function (Cohen‐Salmon et al., 1997). The homology concerns amino acid stretches specific to otogelin and the beginning of a vWF D‐type domain which is involved in multimerization of the von Willebrand factor (Voorberg et al., 1990). Cloning of the full‐length cDNA and characterization of the corresponding protein is now needed to establish the type of protein encoded by this gene. The precise role of this TAL‐1 target gene as a mediator of TAL‐1 biological function in erythroid lineage also remains to be established. Our preliminary data indicate that this gene is downregulated during terminal differentiation of MEL cells, although TAL‐1 mRNA increased. This apparent discrepancy might be explained by the function(s) of LMO2 and Ldb1 which are part of the multimeric complex that was shown to bind to the bipartite E‐box–GATA motif (Wadman et al., 1997) and whose expression also decreased during MEL cells differentiation (Visvader et al., 1997).
Finally, TAL‐1 target gene expression in a cell line expressing endothelial markers is in accordance with the major role of TAL‐1 in embryonic angiogenesis (Visvader et al., 1998) and therefore suggests a possible role of this target gene in vascular development. Whether GATA‐2 may substitute for GATA‐1 activity in regulating this TAL‐1 target gene in endothelial cells remains to be established.
Materials and methods
The MEL cell line (strain 745) and the HeLa cell line were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin and 10% fetal calf serum (FCS). The human erythro‐megakaryocytic cell line (HEL) was cultured in RPMI 1640 supplemented with antibiotics and 10% FCS. In all experiments, exponentially growing cells were used.
For the in vitro differentiation of ES cells (clone CJ7) towards the erythrocytic and megakaryocytic lineages, ES cells were grown in 0.9% α‐methylcellulose containing monothioglycerol and the following growth factors: interleukin 3, GM‐CSF, interleukin 6, interleukin 11, murine stem‐cell factor, thrombopoietin and erythropoietin. The cells were cultured in an humidified incubator at 37°C with 5% CO2 for 9 days.
Hematopoietic progenitor cell (HPC) purification and liquid differentiation culture
HPC purification and the induction of their differentiation using growth factors in liquid cultures were performed exactly as previously described (Condorelli et al., 1995).
In vivo immunoselection of genomic targets for TAL‐1 in MEL nuclei
To isolate functionally relevant in vivo TAL‐1 binding sites we used an immunopurification method of DNA–protein complexes in chromatin. A schematic diagram of the strategy is depicted in Figure 1. Exponentially growing MEL cells (3×108) were lysed in buffer A (10 mM Tris–HCL pH 7.9, 10 mM NaCl, 1.5 mM MgCl2, 1 mM DTT) supplemented with 0.1% Triton X‐100 and several protease inhibitors (2 μg/ml aprotinin, leupeptin and pepstatin, 0.1 mM PMSF), and washed twice in buffer A. To cross‐link in vivo DNA–protein complexes, the intact nuclei were resuspended in 10 mM Na‐phosphate buffer pH 7.4 containing 2% paraformaldehyde, incubated for 30 min at 4°C with gentle rocking, and then washed three times in buffer A. Nuclei were resuspended in buffer A supplemented with 2 mM CaCl2 and incubated with 20 U/ml Micrococcal nuclease (Pharmacia) at 37°C for 5 min. The reaction was stopped with 5 mM EDTA. 1% NP‐40 and 2% sarcosine (N‐Methylaminoacetic acid) were added and soluble chromatin was harvested by centrifugation at 15 000 r.p.m. for 15 min at 4°C, then loaded onto a CsCl step‐gradient (1.75, 1.5, 1.35 g/cm3). The gradient was centrifuged for 40 h at 20°C at 27 000 r.p.m. in a SW41 rotor (Beckman). Fractions were collected from the bottom of the tube and OD260/OD280 ratio was measured. Fractions containing nucleoprotein complexes were selected and dialyzed against dialysis buffer (10 mM Tris, 1 mM EDTA, 0.5 mM PMSF, pH 7.5). The pooled selected chromatin fractions were subjected to immunoprecipitation, using the anti‐TAL‐1 antibody (2TL136 Mab, 10 μg/ml, a gift from D.Mathieu) (Pulford et al., 1995) in the dialysis buffer supplemented with 1% NP‐40 for 4 h at 4°C. After addition of protein G Plus–Agarose (Santa Cruz Biotechnology), incubation was continued for an additional 4 h at 4°C. The protein G–agarose beads were washed with 10 mM Na‐phosphate buffer pH 7.0 containing 1 M NaCl and 0.1% NP‐40 and washed again with the above buffer without NaCl. Washing with high‐ and low‐salt solutions was repeated for two more cycles. The bound nucleoprotein complexes were eluted with 3 M NaSCN containing 10 mM Na‐phosphate buffer pH 7.0 for 1 h at 4°C with gentle rocking. Immunoprecipitate was shown to be enriched in TAL‐1 protein by Western blotting. Eluate was dialyzed overnight against 0.1 M Na‐phosphate buffer pH 7.0, 1 mM EDTA. The dialyzed sample was incubated with 0.1 mg/ml proteinase K (Boehringer Mannheim), 50 mM Tris–HCl buffer pH 8, 100 mM EDTA, 0.3% sodium dodecyl sulfate (SDS) at 55°C for 3 h, extracted twice with phenol‐chloroform‐isoamylalcohol and ethanol‐precipitated in presence of dextran T40 as a carrier. The purified DNA was blunt‐ended by T4 DNA polymerase and one fifth of the material was cloned into EcoRV digested pBluescript SK‐vector (Stratagene) using library‐efficiency Escherichia coli XL1‐Blue competent cells (Stratagene).
DNA sequencing and sequence analysis
DNA sequencing was carried out using the PRISM Ready Reaction Dye Deoxy Terminator Cycle Sequencing Kit (Perkin‐Elmer) on the ABI Model 377 DNA Sequencer (Applied Biosystems).
Construction of reporter plasmids and transient transfection experiments
The glycophorin B minimal promoter (GpB‐37) (Rahuel et al., 1992) was ligated between the MluI and BglII sites of the firefly luciferase reporter vector pGL3‐Basic (Promega) to generate the plasmid pGL3‐GpB. One copy of pairs of complementary oligonucleotides (45 bp) corresponding to the wild‐type or mutated E‐box1–E‐box2–GATA sequence was then inserted into the Acc65I site of pGL3‐GpB to create the luciferase reporter E‐box1–E‐box2–GATA–GpB or its mutated forms.
Transcriptional activity was tested by transient transfection of MEL, HEL or HeLa cells, using the Dual‐Luciferase™ reporter Assay System (Promega). For each transfection, 107 log‐phase cells were washed in phosphate‐buffered saline (PBS), pelleted and resuspended in 0.2 ml of ice cold PBS with 50 mM HEPES pH 7.4, containing 10 μg of supercoiled plasmid construct and 100 ng of pRL‐TK control vector for normalization. Electroporations (250 V, 960 μF) were performed using the Bio‐Rad Gene Pulser (Bio‐Rad). The cells were resuspended in fresh medium, incubated for 24 h at 37°C, then harvested and lysed in 200 μl of Passive Lysis Buffer (Promega). Quantification of the luminescent signals from each of the two luciferase reporter enzymes was performed on 2–10 μl aliquots of supernatant using an automated luminometer equipped with two reagent injectors (LUMAT LB 9507, Berthold).
Cloning of the genomic DNA fragment that encompassed the immunoselected sequence and detection of the transcription unit linked to this sequence
The 474 bp immunoselected DNA fragment was 32P‐labelled and used as a probe to isolate phages from a λGEM12 mouse genomic library (a kind gift from Généthon, Evry, France) using standard methods (Sambrook et al., 1989). A 18 kb genomic DNA fragment that contained the immunoprecipitated sequence was selected and cut with various restriction enzymes to isolate subfragments (4–6 kb) that were used in an exon‐trapping assay (Exon Trapping System, Gibco‐BRL). Data base searches for nucleotides and amino acids sequence alignment were achieved using the N.C.B.I. Blast program (Altschul et al., 1997).
Total RNAs were isolated from cell lines or tissues using Trizol (Gibco‐BRL), and random priming cDNAs were synthesized from 1 μg of total RNA. A fraction (2%) of the cDNA was used for the HPRT PCR and the amount of cDNA used for each of the specific amplification was normalized by the amount of the HPRT PCR product. Products were visualized after separation on a 1.5% agarose gel by ethidium bromide staining and hybridized with a 32P‐labelled internal oligonucleotide to assess the specificity of the product obtained.
Sequence of oligonucleotides
Oligonucleotides used as probes and competitors in EMSAs, and in the construction of reporter plasmids are shown in Table II. TAL‐1, GATA‐1, LMO2 and HPRT oligonucleotides used for RT–PCR are described elsewhere (Keller et al., 1993). The location of TAL‐1 target gene oligonucleotides used for expression studies is indicated in Figure 7A.
Nuclear extracts and DNA binding assays
Nuclear extracts were prepared from MEL cell line and from normal erythroid and granulocytic cells in culture (Schreiber et al., 1989), and DNA binding assays were performed essentially as described (Hsu et al., 1994a; Condorelli et al., 1995; Wadman et al., 1997). Quantification of the TAL‐1‐containing complex was done using a PhosphoImager (Molecular Dynamics).
We would like to thank Dr G.Uzan for the kind gift of ES cells. We are grateful to Dr S.Chrétien (INTS, Paris, France) and to the members of the INSERM U.474, especially to Dr D.Beaupain, Dr M.‐C.Garel, Dr V.Mignotte and Dr N.Raich for constant support and helpful discussions during this work. We thank Anne‐Marie Dulac for typing the manuscript. This work was supported by the Institut National de la Santé et de la Recherche Médicale, by the Ligue Nationale contre le Cancer and by the ARC. L.Maouche‐Chrétien was supported by the Fondation de France.
- Copyright © 1998 European Molecular Biology Organization