Establishing the transcriptional programme for blood: the SCL stem cell enhancer is regulated by a multiprotein complex containing Ets and GATA factors

Berthold Göttgens, Aristotelis Nastos, Sarah Kinston, Sandie Piltz, Eric C.M. Delabesse, Maureen Stanley, Maria‐Jose Sanchez, Aldo Ciau‐Uitz, Roger Patient, Anthony R. Green

Author Affiliations

  1. Berthold Göttgens*,1,
  2. Aristotelis Nastos2,
  3. Sarah Kinston1,
  4. Sandie Piltz1,
  5. Eric C.M. Delabesse1,
  6. Maureen Stanley1,
  7. Maria‐Jose Sanchez1,
  8. Aldo Ciau‐Uitz2,
  9. Roger Patient2 and
  10. Anthony R. Green1
  1. 1 University of Cambridge Department of Haematology, Cambridge Institute for Medical Research, Hills Road, Cambridge, CB2 2XY, UK
  2. 2 Institute of Genetics, Nottingham University, Queen's Medical Centre, Nottingham, NG7 2UH, UK
  1. *Corresponding author. E‐mail: bg200{at}


Stem cells are a central feature of metazoan biology. Haematopoietic stem cells (HSCs) represent the best‐characterized example of this phenomenon, but the molecular mechanisms responsible for their formation remain obscure. The stem cell leukaemia (SCL) gene encodes a basic helix–loop–helix (bHLH) transcription factor with an essential role in specifying HSCs. Here we have addressed the transcriptional hierarchy responsible for HSC formation by characterizing an SCL 3′ enhancer that targets expression to HSCs and endothelium and their bipotential precursors, the haemangioblast. We have identified three critical motifs, which are essential for enhancer function and bind GATA‐2, Fli‐1 and Elf‐1 in vivo. Our results suggest that these transcription factors are key components of an enhanceosome responsible for activating SCL transcription and establishing the transcriptional programme required for HSC formation.


One of the fundamental issues facing current biology concerns the molecular mechanisms whereby multipotent stem cells give rise to distinct differentiated progeny. This issue is central to the development of multicellular organisms from a single fertilized egg and also to the ability of adult stem cells to support tissue regeneration. Interest has been heightened by results from several laboratories, which suggest that adult stem cells have a much broader developmental potential than previously recognized, and that differentiated cells can regain aspects of stem cell function (for reviews see Fuchs and Segre, 2000; Blau et al., 2001; Lovell‐Badge, 2001; Morrison, 2001). Haematopoiesis is the best‐characterized stem cell system, with functional assays and molecular markers for multiple stages of differentiation from haematopoietic stem cells (HSCs) through a plethora of progenitors to multiple mature cell types (Metcalf, 1998; Orkin, 2000; Weissman, 2000). However, the molecular mechanisms responsible for establishing the transcriptional programme for blood remain obscure.

In the mouse, haematopoiesis is first detected both in the yolk sac blood islands and also within the body of the post‐implantation embryo, before subsequent sequential transfers to the fetal liver, thymus, spleen and bone marrow (Dzierzak and Medvinsky, 1995; Cumano and Godin, 2001; Palis and Yoder, 2001). The development of blood and endothelium is intimately linked throughout vertebrate development. The close association between the development of blood and endothelial cells in the avian and murine yolk sac gave rise to the concept of the haemangioblast, a bipotent precursor of both cell types (Sabin, 1920; Wagner, 1980). This idea has been reinforced by the observation that murine embryonic stem (ES) cells can give rise to Flk‐1+ cells capable of generating both blood and endothelial progeny (Choi et al., 1998; Nishikawa et al., 1998a). Moreover, single Flk‐1‐positive cells from avian embryos can develop into either haematopoietic or endothelial colonies (Eichmann et al., 1997), and numbers of both endothelial and haematopoietic cells are severely reduced in the zebrafish mutant cloche (Stainier et al., 1995; Liao et al., 1997). In the body of the amphibian, avian or murine embryo, blood cells arise as clusters of cells attached to the endothelium of arteries (Dieterlen‐Lievre, 1975; Garcia‐Porrero et al., 1995; Ciau‐Uitz et al., 2000; de Bruijn et al., 2000), and it has been suggested that differentiated endothelial cells may directly generate blood progenitors (Jaffredo et al., 1998, 2000; Nishikawa et al., 1998b). Interestingly, Runx1−/− embryos exhibit normal primitive erythropoiesis but fail to develop both intra‐arterial clusters and definitive haematopoiesis (Okuda et al., 1996; Wang et al., 1996; North et al., 1999; Mukouyama et al., 2000). Runx1 may therefore regulate production of blood progenitors from haemogenic endothelium or formation of the latter from a mesodermal precursor. In contrast, two other transcription factors, encoded by the stem cell leukaemia (SCL) and LMO‐2 genes, are essential for the development of both primitive erythropoiesis and definitive haematopoiesis (Warren et al., 1994; Porcher et al., 1996; Robb et al., 1996).

The SCL gene (also known as TAL‐1) encodes a basic helix–loop–helix (bHLH) protein and is normally expressed in blood progenitors, in endothelium and within specific regions of the central nervous system, a pattern of expression that is highly conserved across vertebrate species from mammals to teleost fish (reviewed in Begley and Green, 1999). Within the blood and endothelial system, SCL is expressed in haemangioblasts, HSCs, a subset of haematopoietic lineages and, at lower levels, in angioblasts and at least some mature endothelial cells (Green et al., 1992; Mouthon et al., 1993; Kallianpur et al., 1994; Drake et al., 1997; Gering et al., 1998; Liao et al., 1998; Mead et al., 1998; Sinclair et al., 1999; Akashi et al., 2000; Ciau‐Uitz et al., 2000; Robertson et al., 2000). Targeted mutation of the SCL gene has shown that it is essential for the development of all haematopoietic lineages (Porcher et al., 1996; Robb et al., 1996). Although SCL is expressed in haemangioblasts found in frog and zebrafish systems (Gering et al., 1998; Ciau‐Uitz et al., 2000) or generated during murine ES cell differentiation (Robertson et al., 2000), SCL−/− mouse embryos and ES cells both generate endothelial cells (Visvader et al., 1998; Robertson et al., 2000), suggesting that SCL is required for lineage commitment to blood cell formation. Consistent with this concept, ectopic expression of SCL during zebrafish development results in excessive formation of haemangioblasts and blood cells (Gering et al., 1998). The defective remodelling of primary vascular networks observed in SCL−/− embryos may reflect a distinct later function of SCL (Visvader et al., 1998) or may represent a consequence of the absence of haematopoietic progenitors (Takakura et al., 2000).

Current evidence therefore demonstrates that SCL plays a pivotal role in the normal development of both blood and endothelium. This focuses attention on the mechanisms whereby transcription of SCL itself is initiated and maintained, and our laboratory has undertaken a systematic analysis of the transcriptional regulation of the murine SCL locus. Both human and murine SCL are transcribed from two lineage‐specific promoters (Aplan et al., 1990; Begley et al., 1994; Lecointe et al., 1994; Bockamp et al., 1995, 1997, 1998). A survey of the chromatin structure surrounding the murine SCL gene has revealed a panel of DNase I‐hypersensitive sites associated with enhancer or silencer activity in transfection assays (Göttgens et al., 1997). Transgenic reporter assays subsequently identified five independent enhancers, each of which targets expression to a specific subdomain of the normal SCL expression pattern (Sanchez et al., 1999; Sinclair et al., 1999; Göttgens et al., 2000).

A 3′ enhancer contained within a 5.5 kb fragment displayed particularly striking properties. It was active in the region of E7.5 extraembryonic mesoderm that gives rise to the yolk sac, and subsequently directed reporter gene expression to endothelial and blood cells within yolk sac blood islands of E8 embryos (Sanchez et al., 1999). Within the embryo proper, the enhancer was active in endothelial cells and also in haematopoietic progenitors at multiple sites and times, including E8 para‐aortic splanchnopleura, E11 AGM region and E11 fetal liver (Sanchez et al., 1999). The 3′ enhancer targeted expression to the vast majority of long‐term repopulating HSCs from adult bone marrow and fetal liver (Sanchez et al., 2001). Moreover, expression of SCL under control of this stem cell enhancer in transgenic mice selectively rescued the formation of early haematopoietic progenitors in SCL−/− embryos (Sanchez et al., 2001). These data suggest that the SCL 3′ enhancer functions as a nodal point for the integration of signals responsible for establishing the transcriptional programme for blood cell development.

Here we define a 641 bp conserved core enhancer responsible for targeting expression to blood progenitors and endothelium in transgenic mice, and to dorsal lateral plate mesoderm in transgenic frogs. Three critical transcription factor‐binding sites are identified, each of which is essential for enhancer function in vitro and in vivo. We demonstrate that GATA‐2, Fli‐1 and Elf‐1 bind to the critical motifs in vitro and within intact haematopoietic progenitor cells. Finally, we show that the three transcription factors are capable of forming a complex, that the three corresponding motifs all lie on one face of the DNA helix and that this arrangement is essential for enhancer function. These data strongly suggest that GATA‐2, Fli‐1 and Elf‐1 are key components of an enhanceosome responsible for directing expression to haemangioblasts together with their haematopoietic and endothelial progeny.


Chromatin structure of the SCL stem cell enhancer

Our previous analysis of the regulation of the murine SCL gene identified a 5.5 kb fragment 3′ to the SCL‐coding region in which we mapped the approximate positions of two regions of DNase I hypersensitivity (Göttgens et al., 1997). Transgenic analysis showed that this fragment directed expression of a lacZ reporter gene to endothelium and blood progenitors throughout ontogeny (Sanchez et al., 1999), as well as to the vast majority of long‐term repopulating HSCs from fetal liver and adult bone marrow (Sanchez et al., 2001). In order to refine the position of the enhancer, we mapped regions of open chromatin in two SCL‐expressing multipotent haematopoietic cell lines, EML‐C1 and 416B (Figure 1), using a restriction endonuclease accessibility assay with a resolution of ∼100 bp (Göttgens et al., 2001). EML‐C1 cells contained three regions of accessible chromatin (+17, +18 and +19; numbers refer to the distance in kb from the SCL promoter), two of which (+18 and +19) were also identified in 416B cells.

Figure 1.

Identification of candidate core enhancer regions. (A) Restriction endonuclease accessibility assays identify three regions of accessible chromatin [+17, +18 and +19; numbering corresponds to distances in kb from the start of exon 1a in two haematopoietic progenitor cell lines (EML‐C1, 416B)]. (B) Diagram of the murine SCL locus indicating the three regions of open chromatin above a homology profile of a human/mouse sequence alignment. All sites for AvaII, HaeIII and HinfI are indicated. The summary shows the three regions accessible to endonucleases (black boxes labelled +17, +18 and +19) which correspond to peaks within the homology profile.

We have previously reported the complete sequence of the mouse and human SCL loci (Göttgens et al., 2000), thus allowing the precise position of accessible restriction endonuclease sites to be placed on the mouse sequence and compared with a human/mouse homology profile. This analysis revealed two major peaks of high sequence homology corresponding to the +18 and +19 regions of open chromatin, as well as several minor peaks, one of which corresponded to the +17 region. These results therefore identified the +17, +18 and +19 regions as prime candidates for the location of the SCL stem cell enhancer.

The +19 region is both necessary and sufficient for SCL stem cell enhancer activity in transgenic mouse embryos

The position of the stem cell enhancer was refined by comparing expression of lacZ reporter constructs in F0 transgenic embryos. Embryos were analysed at E11.5, which allowed enhancer activity to be assessed in yolk sac, fetal liver and the AGM region at a single time point. We have demonstrated previously that the 5.5 kb SCL 3′ fragment within the SV/lac/Bg5.5 construct confers stem cell enhancer activity on otherwise inert minimal promoters in F0 transgenic assays (Sanchez et al., 1999). We now tested a 2.5 kb fragment containing the +17, +18 and +19 sites, which gave rise to X‐gal staining in yolk sac, fetal liver and endothelium in a pattern that was the same as that produced by the original 5.5 kb enhancer fragment (Figure 2, compare A and B; Table I). Histological sections revealed that both constructs targeted haematopoietic and endothelial cells in fetal liver and dorsal aorta, including clusters of round cells attached to the ventral floor of the aorta which are thought to represent haematopoietic stem cells (Peault, 1996) (Figure 2I).

Figure 2.

The +19 region is both necessary and sufficient for SCL stem cell enhancer activity in transgenic mouse embryos. Top: transgenic reporter constructs in relation to the mouse SCL locus. (AH) Representative E11.5 transgenic embryos for the constructs indicated in the top part of the figure. (I) Sections of fetal liver (fl) and dorsal aorta (da) of the embryos shown in (A), (B), (E) and (G).

View this table:
Table 1. Activity of SCL stem cell enhancer constructs in transgenic mouse embryos

The activity of the 2.5 kb fragment was not altered by deletion of the +17 or +18 regions (Figure 2, compare B with C and D; Table I), but deletion of the +19 region resulted in complete loss of specific X‐gal staining (Figure 2, compare B with E; Table I). Constructs containing both the +18 and +19 regions or the +19 region alone displayed the same activity as the 2.5 kb fragment in both whole‐mount analysis (Figure 2, compare B with F and G; Table I) and histological sections (Figure 2I). Taken together, these results demonstrate that a 641 bp region containing the +19 site was both necessary and sufficient for activity of the stem cell enhancer.

Sequences containing conserved Myb, Ets and GATA sites are essential for enhancer activity in transgenic mice

The activity of the 641 bp core enhancer was abolished by a 3′ deletion of 227 bp (Figure 2H; Table I). Human/mouse sequence comparisons revealed several blocks of homology within this critical 3′ region, one containing conserved Myb and Ets family‐binding sites, a second containing a conserved Ets site and a third containing a conserved GATA site (Figure 3A, regions 1–3, respectively). Mutations were generated in each region and their effect on enhancer activity was assessed at E11.5 in F0 transgenic embryos.

Figure 3.

Identification of three critical regions in the +19 enhancer. (A) Human/mouse sequence alignment of the +19 enhancer. Highlighted regions 1–3 correspond to the three homology blocks that were mutated (mut1, mut2 and mut3 mutant sequences shown under each homology block). The arrowhead indicates the position of the deletion removing all three homology blocks (SA0.4 fragment, see Figure 2). (B) Representative E11.5 transgenic embryos for the constructs indicated below each panel. SV/lac/SH0.6 contains the wild‐type 641 bp core enhancer; SV/lac/SH0.6 mut1, mut2 and mut3 contain core enhancers with mutation of regions 1, 2 and 3, respectively. The arrow in the right hand panel indicates the endothelial staining in the vitelline vessel.

A mutation in region 1 that altered both the Myb and Ets consensus binding sites completely abolished enhancer activity, as did a mutation altering the Ets site of region 2 (Figure 3B; Table I). A mutation of region 3 that altered the conserved GATA site also abolished X‐gal staining in fetal liver and yolk sac, and markedly reduced endothelial staining. Residual weak X‐gal staining was evident in the endothelium of the vitelline and some intraembryonic vessels but was absent from yolk sac blood island endothelium (Figure 3B; Table I). Only a minority of dorsal aorta endothelial cells were positive, and we did not observe any staining in clusters of cells attached to the floor of the aorta wall (data not shown). These results therefore demonstrate that all three regions are essential for activity of the +19 core enhancer in vivo.

The +19 enhancer directs expression to dorsal lateral plate mesoderm in transgenic frogs

Expression of the +19 enhancer in blood and endothelial tissues raises the possibility that it may be active in the putative bipotential progenitor of these tissues, the so‐called haemangioblast. Haemangioblasts have not been described in mouse embryos, but lineage and gene expression analyses in Xenopus embryos have identified a population of cells in the dorsal lateral plate (DLP) mesoderm which gives rise to adult embryonic blood and the major vessels (Ciau‐Uitz et al., 2000). We therefore tested the activity of the mouse +19 enhancer in the DLP region of transgenic frog embryos. An SV40 minimal promoter driving green fluorescent protein (GFP) gave rise to a variable pattern of expression in the somites, head or ventral regions of stage 26–32 embryos (for example see Figure 4A). Inclusion of the murine 2.5 kb enhancer fragment (containing the +17, +18 and +19 regions) resulted in consistent expression in the region of the DLP (Figure 4, compare A with B, arrow). DLP staining was also obtained when the 2.5 kb enhancer fragment was included in GFP reporter constructs with the HSV‐TK or Xenopus cytoskeletal actin minimal promoters (data not shown), and so, as in transgenic mice (Sanchez et al., 1999), specificity was provided by the enhancer. Sequential in situ hybridization demonstrated that GFP (turquoise) was co‐expressed with endogenous SCL (purple) in the region of the DLP in stage 26 embryos (Figure 4C–E), a time point at which the DLP is known to contain cells co‐expressing SCL and endothelial genes, i.e. putative haemangioblasts (Mead et al., 1998; Ciau‐Uitz et al., 2000).

Figure 4.

The +19 core enhancer is active in transgenic frogs and murine haematopoietic progenitor cells and critically depends on regions 1, 2 and 3. (A) An example of the variable ectopic expression patterns seen in a minority of transgenic stage 32 embryos with the SV/GFP construct (ectopic expression was never seen in the DLP region). (B) SV/GFP/H2.5 transgenic embryo showing reporter gene expression in the DLP region (arrow). (CE) Stage 27 SV/GFP/H2.5 transgenic embryo stained for GFP (BCIP, turquoise) and endogenous SCL (BM Purple). d/e indicates the plane of the section shown in (D) and (E) which demonstrate that the purple staining (endogenous SCL) is contained within the area of transgene expression (turquoise) in the DLP (arrow). (F) Summary of transgenic frog experiments using wild‐type and mutant versions of the SH0.6 core enhancer. DLP, embryos showing DLP staining with or without ectopic expression; ectopic, embryos showing ectopic expression only. The number in each bar represents the number of GFP‐expressing embryos for each category. (G) The SV/luc/SH0.6 construct was active in 416B but not BW5147 cells in stable transfection assays. (H) Stable transfection assays in 416B cells demonstrate that the SH0.6 enhancer activity critically depends on the 227 bp 3′ fragment deleted in the SA0.4 construct, and that the three conserved regions mutated in the mut1, mut2 and mut3 constructs are each critical for enhancer function.

As shown in Figure 4F, the 641 bp fragment containing the +19 region was sufficient to target expression to the DLP in 14 of 20 (70%) GFP‐expressing stage 28–32 embryos. Enhancer activity in the DLP was abolished by mutation of regions 1 or 2, and was observed in only one of 47 GFP‐expressing embryos after mutation of region 3. These data therefore demonstrate that the SCL +19 enhancer can target expression to haemangioblasts in frogs and that the same three regions are critical.

The +19 region functions as a chromatin‐dependent enhancer in haematopoietic progenitor cell lines

In order to perform biochemical analysis of the +19 enhancer, it was important to develop a cellular assay for enhancer function. Luciferase reporter assays were therefore established using 416B cells, an SCL‐expressing cell line derived from a multipotent haematopoietic progenitor, and BW5147 cells, a T‐cell line that does not express SCL and has no open chromatin in the +19 region (Göttgens et al., 1997). In transient transfection experiments, inclusion of the 641 bp +19 core enhancer only provided a marginal increase in the activity of a luciferase reporter driven by the SV40 minimal promoter (data not shown). Stable transfection studies were then performed since the activity of some enhancers is only revealed following integration into chromatin (Tapscott et al., 1992; May and Enver, 1995). Compared with an enhancerless control construct, the 641 bp +19 enhancer resulted in a 21‐fold increase in luciferase activity in 416B cells but was inactive in BW5147 cells (Figure 4G). These experiments demonstrate that the +19 region functions as an integration‐dependent and cell type‐specific enhancer in transfection assays.

Enhancer activity in 416B cells was abolished by the 227 bp deletion, which removes conserved regions 1, 2 and 3 (Figure 4H, SV/luc/SA0.4). Moreover, mutation of any one of the three conserved regions also abrogated enhancer function (Figure 4H). Thus, as in mouse and frog embryos, the activity of the +19 core enhancer was critically dependent on the integrity of regions 1, 2 and 3.

Elf‐1, Fli‐1 and GATA‐2 bind to the +19 enhancer in vitro and in vivo

In order to identify transcription factors capable of binding to the +19 core enhancer, gel shift assays using 416B extracts were performed. Region 1 contains conserved myb (TAACAG) and ets (GGAT) sites. A region 1 oligo was bound by two specific complexes (Figure 5A, I and II), which could be competed by a wild‐type oligo (Figure 5A, compare lanes 2 and 3) but not by a mutant oligo with altered myb and ets sites (Figure 5A, compare lanes 2 and 4). Binding of both complexes I and II was reduced by competition with an oligo containing only an altered myb site (Figure 5A, compare lane 4 with lane 5), but not by an oligo containing only an altered ets site (Figure 5A, compare lane 4 with lane 6). These results suggest that complexes I and II contained Ets family transcription factors.

Figure 5.

GATA‐2, Fli‐1 and Elf‐1 bind to the +19 core enhancer. (A) Fli‐1 and Elf‐1 bind to region 1 in band shift analysis using 416B nuclear extracts. I and II correspond to complexes containing Fli‐1 and Elf‐1 as demonstrated by supershift analysis (compare lane 2 with lanes 7 and 11). Oligo sequences used are shown below the autoradiograph. Antibodies and competitor oligos are as indicated (con, IgG control antibody; wt, wild‐type; 1, mut1; 1m, mut1m; 1e, mut1e). (B) Elf‐1 binds to region 2 in band shift analysis using 416B nuclear extracts. I, II and III correspond to complexes specifically binding to region 2. Complex III was shown to contain Elf‐1 by supershift analysis (compare lanes 5 and 6). Oligo sequences are shown below the autoradiograph. Antibodies and competitor oligos are as indicated (wt, wild‐type; 2, mut2; 2e, mut2e). (C) GATA proteins bind to region 3 in band shift analysis using 416B nuclear extracts. I, II and III correspond to complexes binding specifically to the GATA motif present in region 3. Oligo sequences are shown below the autoradiograph (wt, wild‐type; 3, mut3; Gc, GATA consensus sequence; see Bockamp et al., 1995). (D) GATA‐2, Fli‐1 and Elf‐1 bind to the +19 core enhancer but not to exon 6 sequences in vivo. Chromatin immunoprecipitation was performed using the indicated antibodies. The DNA content of the immunoprecipitates was analysed by real‐time PCR using primers for the +19 core enhancer or for an untranslated region of exon 6. The +19 core enhancer, but not exon 6 sequences, were enriched in immunoprecipitates obtained with antibodies to GATA‐2, Fli‐1 and Elf‐1 (Input, positive control; Rabbit IgG, negative control).

RT–PCR was used to assess expression of 19 Ets family members in 416B cells. Ten Ets family members were found to be expressed (Ets1, Ets2, Fli‐1, Pu.1, Elf‐1, Tel, GABPα, Erg, Elk4 and Mef/Elf4), but nine others were not detected (SpiB, Elk1, Er71, Erm, Er81, Pea3, Elf2/Nerf, Ehf and Elf3/Esx) (data not shown). Antibodies to Ets1, Erg, Elf1, Fli‐1, Pu.1 and GABPα were used in supershift experiments. Ets1, Erg, Pu.1 and GABPα antibodies did not alter complex I or II (Figure 5A, lanes 9 and 10; data not shown). However, a Fli‐1 antibody gave rise to a low mobility complex and a concomitant loss of complex II (Figure 5A, compare lane 2 with lane 7). Moreover, an antibody specific to Elf‐1 generated a low mobility complex accompanied by a weakening of complex I (Figure 5A, compare lane 2 with lane 11). These data therefore strongly suggest that complexes I and II contain Elf‐1 and Fli‐1, respectively.

Region 2 contained a conserved ets motif (GGAA). Specific mutation of the ets site (Figure 5B, mut2e) abolished enhancer activity as effectively as the original more extensive mutation (Figure 5B, mut2) in stable reporter assays performed in 416B cells, thus demonstrating the essential role of the Ets motif (data not shown). Gel shift assays performed using a region 2 oligo identified three complexes, all of which were bound by a wild‐type competitor oligo (Figure 5B, compare lanes 2 and 3). A competition oligo containing the extensive mut2 mutation was able to bind complex I, but not complexes II and III (Figure 5B, compare lanes 2 and 4). Only complex III was unable to bind to a competitor oligo containing a specific mutation of the ets site (Figure 5B, compare lanes 2 and 5), thus demonstrating that complex III binds to the ets site. Antibodies to Ets1, Erg, Fli‐1, Pu.1 and GABPα did not alter complex III (Figure 5B, compare lanes 5 and 7; data not shown). In contrast, inclusion of an antibody to Elf‐1 generated a low mobility complex with weakening of complex III (Figure 5B, compare lanes 5 and 6). These results suggest that complex III contains Elf‐1.

Region 3 contains a conserved GATA motif. Previous northern blotting analysis had shown that our 416B cells express readily detectable levels of GATA‐2, but low or undetectable levels of GATA‐1 and GATA‐3 (Bockamp et al., 1997). Expression of GATA‐4, GATA‐5 and GATA‐6 was not detectable by RT–PCR (data not shown). Gel shift analysis using conventional nuclear extracts failed to detect a specific complex binding to a region 3 oligo (data not shown). However, three complexes were identified using a rapid nuclear extract protocol (see Materials and methods). All three were bound by wild‐type competitor oligo and an oligo containing a GATA consensus site (Gc), but not by an oligo with the region 3 mutation (Figure 5C, compare lane 2 with lanes 3, 4 and 5). These results therefore suggest that a GATA protein (probably GATA‐2) binds to the GATA site in region 3.

Gel shift analysis identifies proteins capable of binding to a target sequence in vitro, but the results do not necessarily reflect the pattern of binding within intact cells. To address this issue, chromatin immunoprecipitation (ChIP) studies were performed. Briefly, proteins bound to DNA were cross‐linked using formaldehyde and, after shearing by sonication, protein–DNA complexes were immunoprecipitated using control antibodies and antibodies to specific transcription factors. The DNA content of immunoprecipitates was then analysed by quantitative real‐time PCR (Figure 5D). As a negative control, primers were designed for an untranslated region of exon 6 of the mouse SCL gene, a region not thought to play a role in transcriptional regulation. This region of exon 6 was not enriched in immunoprecipitates obtained using antibodies to Fli‐1, Elf‐1, GATA‐2 or Pu.1 relative to those obtained using rabbit IgG. In contrast, the +19 core enhancer was clearly enriched in immunoprecipitates obtained with antibodies to Fli‐1, Elf‐1 and GATA‐2 (but not Pu.1) relative to immunoprecipitates obtained using rabbit IgG. We therefore conclude that Fli‐1, Elf‐1 and GATA‐2 all bind to the +19 core enhancer in vitro and also within intact 416B cells.

A haematopoietic stem cell enhanceosome

Our results identify three critical transcription factor‐binding sites in the +19 core enhancer. Mutation of each individual site abolishes enhancer activity, suggesting that the corresponding transcription factors participate in a multimeric complex necessary for enhancer function. To investigate whether the three proteins identified in the previous section were capable of interacting, GST pull‐down experiments were performed. In vitro translated Elf‐1, Fli‐1 and GATA‐2 were all capable of binding GST–Elf‐1 and GST–GATA‐2 fusion proteins (Figure 6A and B, compare lanes 2 and 3).

Figure 6.

Formation of a multiprotein complex on the +19 core enhancer. (A) GATA‐2, Fli‐1 and Elf‐1 proteins interact with Elf‐1. In vitro translated GATA‐2, Fli‐1 and Elf‐1 are bound specifically by GST–Elf‐1 in pull‐down experiments (1, input; 2, adsorption to GST alone; 3, adsorption to GST–Elf‐1). (B) GATA‐2, Fli‐1 and Elf‐1 proteins interact with GATA‐2. In vitro translated GATA‐2, Fli‐1 and Elf‐1 are bound specifically by GST–GATA‐2 in pull‐down experiments (1, input; 2, adsorption to GST alone; 3, adsorption to GST–GATA‐2). (C) The activity of the +19 enhancer requires the binding sites for GATA‐2, Fli‐1 and Elf‐1 to be on the same surface of the DNA helix. A 6 bp insertion between regions 1 and 2 (SV/luc/SH0.6mut4) or between regions 2 and 3 (SV/luc/SH0.6mut5) severely reduces enhancer activity in stable transfection assays in 416B cells. (D) A model for the role of the +19 enhanceosome in the development of HSCs and blood. It is anticipated that different GATA and Ets family members may participate in distinct cell types or in response to specific signals. SCL expression is essential for lineage commitment to blood but not endothelium.

The concept of a multimeric protein complex is also consistent with the spacing between regions 1, 2 and 3 (18 and 21 bp), which suggests that transcription factors binding to these regions would all lie on the same face of the DNA helix. To assess whether this arrangement was critical for activity of the +19 core enhancer, we generated two new reporter constructs with a 6 bp insertion between regions 1 and 2 or between regions 2 and 3 (SV/luc/SH0.6mut4 and SV/luc/SH0.6mut5, respectively). Both constructs showed greatly diminished activity when assayed in 416B cells (Figure 6C). Taken together, our results therefore suggest a model in which Fli‐1, Elf‐1 and GATA‐2 form three DNA‐binding components of an enhanceosome responsible for establishing the transcriptional programme necessary for HSC and blood cell development (Figure 6D).


Activation of SCL expression is essential for the initiation of haematopoiesis and the formation of HSCs, but not for the specification of endothelial cells. The SCL stem cell enhancer is active in haemangioblasts together with their haematopoietic and endothelial progeny, and expression of SCL under control of the stem cell enhancer results in selective rescue of early haematopoietic progenitors in SCL−/− embryos. This enhancer therefore functions as a pivotal element responsible for integrating signals necessary for lineage commitment to HSCs and blood cell formation. Our results identify a core +19 enhancer containing three key motifs, each of which is essential for enhancer function in haemangioblasts, blood progenitors and endothelial cells. In addition, our data identify GATA‐2, Fli‐1 and Elf‐1 as critical components of a novel multiprotein complex with a central role in establishing the transcriptional programme for blood.

Formation of a multiprotein complex on the +19 enhancer is essential for targeting expression to blood progenitors and endothelium

Several aspects of our data point to the assembly on the +19 enhancer of a multiprotein complex with a central role in blood and endothelial development. Mutation of any one of three critical binding sites resulted in marked loss of enhancer function, an observation which is reminiscent of the interferon‐γ (IFN‐β) enhanceosome (reviewed in Carey, 1998; Merika and Thanos, 2001). The intact IFN‐β enhancer displays a highly specific response to viral infection, a property which cannot be reproduced by individual components of the enhancer. Its biological function therefore represents more than the sum of its parts and involves the integration of multiple regulatory inputs in a precise way. Our results also demonstrate that activity of the +19 enhancer requires all three critical transcription factor‐binding sites to be on the same face of the DNA helix, analogous to the prototypic IFN‐β and T‐cell receptor α (TCRα) enhanceosomes. In both of these cases, an architectural protein (HMG‐I or LEF‐1, respectively) bends the DNA and thereby reduces the energetic threshold for complex assembly (Giese et al., 1995; Kim and Maniatis, 1997). It may therefore be relevant that Elf‐1 has been shown to interact with HMG‐I(Y) (John et al., 1995).

The IFN‐β and TCRα enhancers, together with the hypersensitive site 2 (HS2) of the β‐globin locus control region (LCR), are all classical enhancers and are active in transient transfection assays. In contrast, the SCL +19 enhancer falls into the category of chromatin‐dependent enhancers, which require stable integration to reveal their activity. Analogous elements include the HS3 and HS4 of the β‐globin LCR (for a review see Li et al., 1999) together with enhancers from the MyoD and CD34 genes (Tapscott et al., 1992; May and Enver, 1995). The molecular basis for chromatin dependence is poorly understood. Part of the explanation may lie in the observation that some promoters are active in the absence of an enhancer in transient transfection assays but not in stable transfection experiments (Göttgens et al., 1997; Li et al., 2001). The function of a linked enhancer may be masked in the former and only become evident in stable assays. In this case, the difference between classical and chromatin‐dependent enhancers may be partly quantitative, with classical enhancers being more powerful and therefore detectable in transient assays above the promoter activity. Alternatively, the two categories of enhancer may be qualitatively distinct, with chromatin‐dependent enhancers requiring some aspect of chromosomal chromatin for enhanceosome assembly or function. This would be consistent with the observations that chromatin assembly on transiently transfected DNA is incomplete (Jeong and Stein, 1994) and that some enhancers exhibit distinctive chromatin‐opening properties (Ellis et al., 1996). In this context, the presence of a critical GATA site in the +19 enhancer may be significant since GATA proteins interact with histone‐modifying proteins (Blobel et al., 1998; Wada et al., 2000; Dai and Markham, 2001) and possess an evolutionarily conserved chromatin remodelling activity (Boyes and Felsenfeld, 1996; Boyes et al., 1998; Muro‐Pastor et al., 1999).

The specificity of the SCL +19 enhancer is also striking for two reasons. First, a 641 bp core enhancer is sufficient to direct expression to blood progenitors and endothelium at multiple sites including yolk sac, para‐aortic splanchno pleure/AGM region and fetal liver. This suggests either that there exists an unexpected uniformity in the signalling pathways responsible for generating blood and endothelium in the yolk sac, and at multiple sites and times within the embryo, or that the +19 enhancer is able to integrate diverse signals operating at distinct anatomical sites. Secondly, we have been unable to separate haematopoietic and endothelial activities of the +19 enhancer, since the three critical motifs within the +19 enhancer are each essential for targeting expression to both lineages. This argues that formation of a multiprotein complex on the +19 enhancer is necessary for directing expression to haemangioblasts and also for subsequent enhancer activity in both cell types. Furthermore, it is likely that the composition of the +19 enhanceosome will alter during development, with various GATA and Ets family members participating in different cell types, or in response to distinct signals.

GATA‐2, Fli‐1 and Elf‐1 regulate the SCL +19 enhancer

The expression patterns and functions of GATA‐2, Fli‐1 and Elf‐1, where known, are consistent with a role for all three proteins in the regulation of SCL expression during blood and endothelial development. In the mouse, GATA‐2 is expressed in the yolk sac and AGM region and is known to be present in haematopoietic progenitors, mast cells, megakaryocytes and early erythroid cells (Lee et al., 1991; Cross et al., 1994; Minegishi et al., 1999), all cell types in which SCL is also expressed. Moreover, GATA‐2 and SCL are co‐expressed during frog and zebrafish development in lateral mesodermal cells that give rise to blood and endothelium (Gering et al., 1998; Brown et al., 2000; Ciau‐Uitz et al., 2000). There are also several similarities in the functions of GATA‐2 and SCL. GATA‐2 is essential for normal levels of primitive erythropoiesis and for all lineages of definitive haematopoiesis, the latter observation suggesting a defect at the level of HSC (Tsai et al., 1994a).

Haematopoietic differentiation of GATA‐2−/− ES cells is markedly reduced, particularly for those colony types that depend on stem cell factor (Tsai et al., 1994a). Interestingly, c‐kit, the receptor for stem cell factor, has been identified as a target gene of SCL (Krosl et al., 1998), and both SCL and GATA‐2 have been implicated in regulating the proliferation of haematopoietic progenitors (Green et al., 1991; Briegel et al., 1993).

Fli‐1 is also co‐expressed with SCL and GATA‐2 in the lateral mesoderm of frog and zebrafish embryos (Meyer et al., 1995; Brown et al., 2000). In the mouse, Fli‐1 is expressed in developing endothelial and haematopoietic cells (Truong and Ben‐David, 2000; Vlaeminck‐Guillem et al., 2000). Moreover, chicken Fli‐1 is expressed in endothelial cells and splanchnopleural mesoderm including haematopoietic clusters attached to the wall of the dorsal aorta (Mager et al., 1998). Three targeted mutations of the Fli‐1 gene have been reported. The first resulted in mice with no profound phenotype (Melet et al., 1996) and subsequently was realized to represent a hypomorph (Bartel et al., 2000). More recently, two other targeted alleles have been generated, one in which the Ets DNA‐binding domain was replaced by lacZ (Hart et al., 2000) and the other in which a C‐terminal activation domain was deleted (Spyropoulos et al., 2000). The latter two mutations are both embryonically lethal, fail to produce mature megakaryocytes and give rise to haemorrhage within the central nervous system, which may reflect a platelet and/or endothelial defect.

Less is known about the expression pattern and function of Elf‐1. However, murine Elf‐1 is widely expressed in haematopoietic cells (Davis and Roussel, 1996), and chicken Elf‐1 is expressed in embryonic blood and endothelial cells (Dube et al., 2001). Moreover, zebrafish Elf‐1 is co‐expressed with SCL, GATA‐2 and Fli‐1 in the region of the lateral mesoderm that gives rise to blood and endothelium (M.Gering and R.Patient, unpublished observations). No overt haematopoietic defect is seen in Elf‐1−/− mice (Garrett‐Sinha et al., 2001), but a detailed description of these mice awaits publication.

The complete absence of primitive and definitive haematopoiesis that typifies the SCL−/− phenotype is not reproduced by targeted mutation of GATA‐2, Fli‐1 or Elf‐1, and yet alteration of the binding site for any one of these transcription factors abolishes activity of the +19 enhancer. This could reflect the existence of additional SCL regulatory elements that can substitute for the +19 enhancer. Although our data do not exclude this possibility, detailed analysis of ∼55 kb surrounding the murine SCL locus and extending to the 5′‐ and 3′‐flanking genes has not revealed any other enhancer with significant haematopoietic activity in vivo (Sanchez et al., 1999; Sinclair et al., 1999; Göttgens et al., 2000). Instead, it seems more likely that altered levels of related transcription factors can compensate to varying degrees for the absence of GATA‐2, Fli‐1 or Elf‐1. Thus, targeted mutation of GATA‐1 results in a compensatory increase in levels of GATA‐2 (Weiss et al., 1994), and the sequences of the Ets proteins Erg and Mef/Elf‐4 are very similar to those of Fli‐1 and Elf‐1, respectively (Miyazaki et al., 1996; Baltzinger et al., 1999).

A number of signalling pathways activated by growth factors or cellular stress converge on Ets and GATA transcription factors, and include the MAP kinase, phosphatidylinositol 3‐kinase and phospholipase C cascades (Towatari et al., 1995; Yordy and Muise‐Helmericks, 2000; Liang et al., 2001). Cross‐talk has been reported between these pathways and the BMP pathway (reviewed in von Bubnoff and Cho, 2001), which specifies blood and endothelial lineages during embryonic development. Identification of GATA‐2, Fli‐1 and Elf‐1 as key components of the SCL +19 enhanceosome will facilitate analysis of the signalling pathways responsible for HSC formation and thereby allow the development of novel approaches to the manipulation of stem cells.

Materials and methods

Restriction endonuclease accessibility assay

EML‐C1 and 416B cells were maintained as described previously (Tsai et al., 1994b; Bockamp et al., 1997). The restriction endonuclease assay was performed as described (Göttgens et al., 2001) using SacI digestion and a 400 bp SacI–ScaI fragment just 3′ of exon 6 as a probe.

Transgenic analysis

Transgenic reporter constructs were generated using standard cloning procedures and details are available upon request. SV/lac/H2.5 plasmid contained a HindIII fragment from 76 955 to 79 480. The Δ17 deletion was generated by PCR and removed 111 bp ranging from 77 582 to 77 693; the Δ18 deletion removed a 401 bp NcoI–XmnI fragment from 78 359 to 78 760; and the Δ19 deletion removed a 292 bp ApaI–HindIII fragment from 79 188 to 79 480. The NH1.1 fragment ranged from 78 359 to 79 480, the SH0.6 fragment from 78 840 to 79 480 and the SA0.4 fragment from 78 840 to 79 188 (numbering refers to the SCL locus database entry AJ131017). All mutations were generated using standard procedures and verified by DNA sequencing. F0 transgenic mouse embryos were prepared and analysed as described (Sanchez et al., 1999). Transgenic Xenopus embryos were generated using the method of Kroll and Amaya (1996) with one modification. Sperm was frozen to improve permeability and the percentage of transgenic embryos (Sparrow et al., 2000).

Whole‐mount in situ hybridization was performed according to Bertwistle et al. (1996). When double in situ hybridization was performed, embryos were hybridized with probes differentially labelled with fluorescein or digoxigenin. The first colour was developed using BCIP (Boehringer Mannheim), then BM Purple (Boehringer Mannheim) was used for the second colour reaction. The gene with the strongest expression was chosen for the first staining.

Reporter assays and EMSA analysis

Transient/stable transfections and luciferase assays were performed as described previously (Göttgens et al., 1997). For the rapid nuclear extract preparation, 2 × 106 cells were collected by centrifugation, washed in phosphate‐buffered saline (PBS) and resuspended in 400 μl of buffer A [10 mM HEPES pH 7.9, 10 mM KCl, 0.1 mM EGTA, 1 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride (PMSF)], and incubated at 4°C for 15 min. After adding 25 μl of 10% NP‐40, the sample was mixed and briefly centrifuged. The pellet was resuspended in 25 μl of buffer C (10 mM HEPES pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM DTT, 1 mM PMSF) and incubated for 30 min followed by a 5 min centrifugation. A 1 μl aliquot of the resultant supernatant was used per track for electrophoretic mobility shift assay (EMSA) analysis. Standard nuclear extracts were prepared and EMSAs performed as described (Bockamp et al., 1998). Oligonucleotides (wild‐type and mutant) used for EMSA are shown in Figure 5. Antibodies for supershift analysis were purchased from Santa Cruz Biotechnology Inc.

ChIP assay

ChIP assays were performed as described (Forsberg et al., 2000). Antibodies were purchased from Santa Cruz Biotechnology Inc. Precipitates were analysed by conventional and real‐time PCR using the SYBR green intercalant dye according to the manufacturer's instructions (ABI7700 machine; Applied Biosystems, Foster City, CA). The primers used were: Exon6‐FW, CCATACTCTTGCCAAGGC TACC; Exon6‐REV, AGCAGTCCTACATGGGCCTAAA; +19‐FW, CCATACTCTTGCCAAGGCTACC; and +19REV, AGCAGTCCTAC ATGGGCCTAAA.

GST pull‐down assay

Full‐length human Elf‐1 and GATA‐2 were subcloned into pGEX‐4T. GST fusion proteins were expressed and GST pull‐down assays performed as described (Kerr et al., 1993). Human Elf‐1, mouse Fli‐1 and human GATA‐2 proteins were prepared by in vitro transcription/translation using the TNT system (Promega).


We are grateful to Drs E.Bresnick, E.O.Bockamp and M.Walmsley for advice with the chromatin immunoprecipitation, EMSA and Xenopus experiments, respectively. This work was supported by the Wellcome Trust, the Medical Research Council, the British Heart Foundation, the Pre‐Leukaemia Society, the Association pour la Recherche sur le Cancer (ARC) and the Fondation de France (Comité Leucémie).