The SCL +40 Enhancer Targets the Midbrain Together with Primitive and Definitive Hematopoiesis and Is Regulated by SCL and GATA Proteins

Abstract

The SCL/Tal-1 gene encodes a basic helix-loop-helix transcription factor with key roles in hematopoietic and neural development. SCL is expressed in, and required for, both primitive and definitive erythropoiesis. Thus far, we have identified only one erythroid SCL enhancer. Located 40 kb downstream of exon 1a, the +40 enhancer displays activity in primitive erythroblasts. We demonstrate here that a 3.7-kb fragment containing this element also targets expression to the midbrain, a known site of endogenous SCL expression. Although the 3.7-kb construct was active in primitive, but not definitive, erythroblasts, a larger 5.0-kb fragment, encompassing the 3.7-kb region, was active in both fetal and adult definitive hematopoietic cells. This included Ter119⁺ erythroid cells along with fetal liver erythroid and myeloid progenitors. Unlike two other SCL hematopoietic enhancers (+18/19 and −4), +40 enhancer transgenes were inactive in the endothelium. A conserved 400-bp core region, essential for both hematopoietic and midbrain +40 enhancer activity in embryos, relied on two GATA/E-box motifs and was bound in vivo by GATA-1 and SCL in erythroid cells. These results suggest a model in which the SCL +18/19 and/or −4 enhancers initiate SCL expression in early mesodermal derivatives capable of generating blood and endothelium, with subsequent activation of the +40 enhancer via an autoregulatory loop.

The vertebrate hematopoietic system is highly conserved across evolution, as are the expression patterns and functions of its key regulatory genes. The stem cell leukemia gene SCL (or Tal-1) encodes a basic helix-loop-helix transcription factor which is essential for the development of primitive and definitive hematopoiesis (29, 31, 36). Although SCL is not required for the self-renewal or long-term reconstituting activity of adult hematopoietic stem cells (HSCs), those lacking SCL have impaired short-term repopulating capacity (6, 17, 21, 35). Continued SCL expression is also important for normal erythroid and megakaryocytic differentiation (17, 21).

Appropriate transcriptional regulation is essential for the biological functions of SCL, and we have therefore systematically investigated the mechanisms regulating murine SCL transcription. Six independent enhancers have been identified to date, each directing expression to a specific subdomain of the normal SCL expression pattern (7, 11, 12, 33, 38). The SCL +18/19 stem cell enhancer (numbering reflects the distance in kilobases from the start of SCL exon 1a) directs transgene expression to HSCs and progenitors in both adult and fetal liver, as well as to embryonic endothelium (33, 34, 37). The −4 enhancer is active in embryonic endothelial (and hematopoietic) cells (12), but not adult hematopoietic cells. Ets and GATA family transcription factors have been implicated in controlling SCL expression in HSCs and progenitors through these two enhancers (12, 13).

SCL is also expressed in both the primitive and definitive erythroid lineages (9, 14). Since none of the known SCL enhancers directed transgene expression to erythroid cells beyond early progenitor stages (11, 13, 33, 34, 38), we postulated the existence of a distinct erythroid enhancer, responsible for maintaining SCL expression following erythroid commitment (34). A systematic survey of histone acetylation across the SCL locus resulted in identification of the +40 enhancer, a novel element with erythroid specificity in vitro and in transgenic mice (7). A 3.7-kb fragment containing the +40 region, directed transgene expression to primitive, but not definitive, erythroblasts in vivo.

In addition to its pivotal role in hematopoiesis, SCL exhibits a conserved pattern of expression in the central nervous system (CNS), and particularly in specific regions of the diencephalon, mesencephalon, metencephalon, and spinal cord (9, 38, 39). In the spinal cord, SCL promotes glial and neuronal progenitors to adopt astrocyte and V2b interneuron fates, respectively (23). SCL is also essential for the development of neurons in CNS regions known to express SCL (5). Conditional deletion of SCL in neurons resulted in growth retardation and premature death, along with behavioral and visual reflex abnormalities (5). Three different SCL enhancers that target expression to distinct SCL-expressing regions of the CNS have been identified (11, 38).

In this paper, we describe a detailed analysis of the +40 enhancer. In addition to its hematopoietic activity, the 3.7-kb +40 enhancer targets specific midbrain neurons in a pattern consistent with endogenous SCL expression and indistinguishable from that of SCL promoter 1a. Investigation of a 5-kb +40 enhancer fragment revealed activity in adult definitive erythroid and progenitor cells, as well as in the primitive erythroid cells targeted by the 3.7-kb enhancer. We have also identified a conserved 400-bp +40 enhancer core region, and our results demonstrate that its activity requires two conserved GATA/E-box motifs found therein. Moreover, both GATA-1 and SCL proteins bind the core in vivo in erythroid cells, suggesting that SCL transcription is subject to positive autoregulation via the +40 enhancer.

MATERIALS AND METHODS

Generation of +40 constructs.

The 3.7-kb enhancer fragment (+40/3.7), SV/lac/+40 and SV intron/lac/+40 transgenes [referred to here as SV (intron) lac/+40/3.7] were as described previously (7). Shorter +40 fragments were PCR amplified (Extensor hi-fidelity mix; ABgene, Epsom, United Kingdom) from genomic 129 embryonic stem cell DNA and cloned into the SalI and KpnI sites of SV intron/lac/+40/3.7 after removal of the 3.7-kb fragment. To generate the deletion constructs, short fragments located immediately 5′ or 3′ of the 0.4-kb core were PCR amplified using 3′ or 5′ primers (respectively) containing an RsrII site and subcloned into SV intron/lac/+40/3.7 or SV/luc/+40/3.7 using RsrII and a unique upstream or downstream site. Point mutants were generated by site-directed mutagenesis (QuikChange; Stratagene, La Jolla, CA). Fragments containing the desired mutations were subcloned into the original 3.7-kb enhancer. A short fragment, running from nucleotide 97342 of GenBank sequence AJ297131 (the 5′ end of the 5.0-kb enhancer +40/5.0) to the RsrII site internal to +40/3.7, was PCR amplified and inserted into SV intron/lac/+40/3.7 and SV/luc/+40/3.7, resulting in SV intron/lac/+40/5.0 and SV/luc/+40/5.0. All PCR products used were sequenced prior to their use in cloning. Luciferase reporter constructs were made in pBluescript II (Stratagene, La Jolla, CA) and contained the simian virus 40 (SV40) minimal promoter, firefly luciferase coding region, and SV40 poly(A) (all from pGL2-Promoter; Promega, Southampton, United Kingdom) with or without +40 fragments. Further details are available from the authors on request.

Stable transfection assays.

F4N (8) and BW5147 (30) cell lines were grown in RPMI 1640 with l-glutamine (Invitrogen, Carlsbad, CA) as described previously (7). Transfection assays of enhancer activity using luciferase reporter constructs were done as described previously (7). Experiments were repeated at least three times per cell line or experimental series. Results are expressed as mean luciferase activity (in relative light units) ± standard deviation, where the mean of the four pools of the SV/luc “promoter alone”-transfected cells was normalized to 1 and the activity in each of the four pools transfected with the various +40 construct-containing plasmids compared.

Generation and analysis of transgenic mice.

Transgene fragment preparation, generation of transgenic mice, PCR-based genotyping, and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) staining were all performed as described previously (7). Animals bearing the +40/3.7 fragment and its derivatives were additionally genotyped using transgene-specific primers SV40-forward (5′-GGT CTG GTG TCA AAA ATA ATA A-3′) and +40-reverse (5′-ACT ACT TGT CAT TTG TTG CTA A-3′). Details of PCR protocols are available on request. X-Gal-stained embryos were embedded in paraffin, sectioned (5 μm), and counterstained with neutral red (Sigma, St. Louis, MO).

Flow cytometric analysis of β-Gal activity.

Single-cell suspensions from hematopoietic tissues were analyzed for β-galactosidase (β-Gal) activity as described previously (26), except that no red cell lysis was done. A FACScalibur running CellQuest software (BD Biosciences Immunocytometry Systems, San Jose, CA) was used for data acquisition and analysis. Phycoerythrin-conjugated rat anti-mouse Ter119, anti-Mac-1, anti-c-Kit, and rat immunoglobulin G2b (IgG2b) isotype control antibodies, along with biotinylated anti-CD34 and streptavidin-allophycocyanin were purchased from BD Biosciences Pharmingen (San Diego, CA). A forward scatter gate excluded debris, whereas dead cells were excluded by propidium iodide (Sigma) uptake. Fetal blood samples were obtained as described previously (27). Briefly, pregnant females were sacrificed, and the closed uterus was washed. The uterine wall and yolk sac were removed without disrupting the umbilical cord. Umbilical vessels were then clamped and severed, and the embryo was washed in phosphate-buffered saline containing 2% fetal calf serum (PBS-2% FCS). After transfer to fresh PBS-2% FCS, the jugular veins and cervical arteries were severed, and the embryo was allowed to bleed out. For fetal blood analyses, between 0.5 × 10⁴ and 2 × 10⁴ viable cells were examined. Similarly, between 2 × 10⁴ and 10 × 10⁴ viable fetal liver cells were analyzed. For adult tissues, between 0.5× 10⁴ and 2 × 10⁴ viable cells were examined.

In vitro assays of colony-forming activity in sorted populations.

Bone marrow and fetal liver suspensions were stained as described above, except that sterility was maintained throughout. Approximately 2 × 10⁴ cells each was then analyzed. For fetal liver, β-Gal-positive (β-Gal⁺) samples were pooled before sorting, and the genotypes of individual embryos were confirmed retrospectively by PCR on yolk sac DNA. β-Gal⁺ and β-Gal-negative (β-Gal⁻) populations were sorted into PBS-5% FCS using a MOFLO cell sorter (Dako, Ely, United Kingdom). In SV intron/lac/+40/5.0 transgenic bone marrow, the average percentages of gated cells in the β-Gal⁺ and β-Gal⁻ fractions were 1.3% and 97%, respectively. Using fetal liver, the corresponding values were 1.6% and 87% for SV intron/lac/+40/3.7 and 16% and 69% for SV intron/lac/+40/5.0, respectively. A sample of each sorted fraction was taken for reanalysis. Although purities varied between individual experiments, regardless of the tissue sorted, the β-Gal⁻ fraction was over 97% pure, such that contamination of the β-Gal⁺ sort population with β-Gal⁻ cells cannot account for the vast majority of colonies being formed from the β-Gal⁺ sorted population. Viable cell counts were done on the remainder, and 0.5 × 10⁴ to 1.25 × 10⁴ viable cells were plated in Methocult (M3434; Stem Cell Technologies, Vancouver, Canada) according to the manufacturer's protocol. A stained, unsorted sample was treated similarly. A total of 12.5 × 10³ unfractionated, 5 × 10³ β-Gal⁺ or 12.5 × 10³ β-Gal⁻ viable bone marrow cells were plated, respectively. Similarly, 10⁴ unfractionated, 10³ β-Gal⁺ or 10⁴ β-Gal⁻ fetal liver cells were plated. Erythroid and myeloid (granulocyte-macrophage CFU) colonies were scored after 8 to 10 days, and the distributions of colony-forming cells in the sorted β-Gal⁺ and β-Gal⁻ fractions were calculated as described previously (33).

Chromatin immunoprecipitation assays.

Chromatin immunoprecipitation assays were performed as described previously (7) in the F4N erythroleukemia cell line (8), using rabbit anti-SCL (a kind gift from C. Porcher, Oxford, United Kingdom), rabbit anti-GATA-1 (AB11963; Abcam, Cambridge, United Kingdom), and nonspecific rabbit IgG (Sigma-Aldrich). Enrichment was determined using SYBR green real-time quantitative PCR (Stratagene) as described previously (7). The following primer pairs were used to amplify the SCL −16 transcriptionally inactive negative control, the α-globin HS-8 positive control, and the +40 enhancer regions, respectively: NegF (5′-GCA ATG AAC CTC CGA ACT GG-3′) and NegR (5′-CGT CTA AGA AGG TGC CCA CAG-3′), PosF (5′-GTC TCC CTT AGG TAG AGT AG-3′) and PosR (5′-GTG GCT CTT TCT TGG AGA GG-3′), and +40F (5′-TCC TAA AGC CTT GGT GCC TG-3′) and +40R (5′-GAG CTG GCG ATA AGG AAG AGG-3′). The levels of enrichment with the specific antibodies were normalized and calculated as the change in enrichment compared to that for nonspecific IgG. Results shown are a representative experiment of three independent experiments.

RESULTS

Characterization of hematopoietic and midbrain expression driven by the 3.7-kb +40 enhancer element.

We recently described a novel regulatory element for SCL, the +40 enhancer, which targets in vivo lacZ transgene activity to distinct sites of endogenous SCL expression, namely, primitive erythroid cells of the peripheral blood, along with the midbrain (7). To investigate the onset of transgene expression, we analyzed β-Gal activity in transgenic embryos carrying the SV intron/lac/+40/3.7 transgene (Fig. 1A) during early mouse embryonic development. As shown in Fig. 1B, panel i, by embryonic day 7.5 (E7.5), the SV intron/lac/+40/3.7 transgene is expressed in a ring of cells within the extraembryonic mesoderm of whole-mount embryos. Sagittal sections revealed that staining was confined to small groups of cells lying at the junction between extraembryonic tissue and the embryo proper (data not shown). At E8.5, staining was limited to the allantois, yolk sac blood islands, and to cells within the developing blood vessels of the embryo proper (Fig. 1B, panels ii and iii), including the heart rudiment (Fig. 1B, panel iv) and paired dorsal aortae (Fig. 1B, panel v). By contrast to previously characterized SCL regulatory elements, there was no apparent β-Gal activity in the endothelium lining the dorsal aorta or other major blood vessels (Fig. 1B, panels v and vi, respectively) (see also references 12, 13, 33, and 38). Analysis of whole-mount embryos from E9.5 onwards indicated that hematopoietic transgene expression was localized to blood vessels (Fig. 1B, panels vii to ix). At E11.5, blood vessel and yolk sac endothelium (Fig. 1B, panels x and xi, respectively), as well as endocardium (Fig. 1B, panel xii), appeared β-Gal⁻. Generalized staining of circulating blood cells made it difficult to identify clusters of hematopoietic cells on the ventral wall of the dorsal aorta thought to contain HSCs or to unequivocally detect staining in them (data not shown). As noted previously, expression in circulating erythroid cells declined sharply with the shift from primitive to definitive erythropoiesis after E12.5 (data not shown) (7).

FIG. 1.

+40/3.7 transgene expression during mouse development. Enhancer activity in blood coincides with the initiation of hematopoiesis. (A) Diagram of the SV (intron) lac/+40/3.7 transgene(s). (B) Whole mounts and sections from X-Gal-stained SV intron/lac/+40/3.7 transgenic embryos (line 4092) at various stages of development. (i) A whole-mount E7.5 transgenic embryo showing staining in the extraembryonic mesoderm. (ii) An E8.5 embryo. (iii and vi) Cross sections of the same embryo with high-power views (iv and vi, respectively) showing staining in cells located in the yolk sac blood vessel (Bv) lumen (iii), the heart (Ht) rudiment (iii and iv), and the dorsal aortae (DA) (iii and v). Arrowheads in panels v and vi indicate the endothelium surrounding the dorsal aortae (v) and other major blood vessels (vi). (vi) Cells within the lumen of a large blood vessel (Bv), but not the surrounding endothelial lining (arrowhead), were β-galactosidase positive. (vii) An E9.5 embryo with staining localized to the blood vessels, but not midbrain. (viii) An E10.5 embryo showing staining in both blood and midbrain. (ix) An E11.5 transgenic embryo. An arrowhead indicates the diencephalic extension. (x to xii) Sagittal sections through an E11.5 embryo showing a major blood vessel (the arrowhead indicates the endothelial lining) (x), yolk sac (xi), and the developing heart (xii). High-power lateral (xiii) and dorsal (xiv) views of the E11.5 embryo shown in panel ix. Lines marked a and b indicate the plane of the corresponding coronal sections shown in panels xv and xvi, respectively. (xiv to xvi) Stained intertectal axonal projections crossing at the midline (arrows), diencephalic extensions (arrowheads), and cells within blood vessels (Bv) are indicated. (xvii) An X-Gal-stained whole-mount E18.5 brain with staining in the mesencephalon. (xviii to xx) Whole-mount X-Gal-stained brains from adult SV intron/lac/+40/3.7 transgenic (line 4092) (xviii and xix) and SCL-lacZ knock-in mice (line 252) (xx), both displaying β-galactosidase activity in the superior (SC) and inferior (IC) colliculi. Note the staining in the blood vessels (Bv) of the SCL-lacZ knock-in mouse, but not in the SV intron/lac/+40/3.7 transgenic mouse.

In the midbrain, lacZ activity was first detected at E10.5 (Fig. 1B, panel viii). At E11.5, expression was evident in the anterolateral midbrain (Fig. 1B, panels ix, xiii, and xvi) with diencephalic extensions (Fig. 1B, panel ix) and axonal projections to the roof of the midbrain (Fig. 1B, panels xiv and xvi). By E18.5, the β-Gal⁺ axonal projections had spread caudally to cover the tectum (Fig. 1B, panel xvi). Midbrain expression was maintained in adult mice and was particularly prominent in the superior (SC) and inferior (IC) colliculi (Fig. 1B, panels xviii and xix). The same pattern was seen in three SV (intron) lac/+40/3.7 lines (line 4092, 4102, and 4274) and it appeared identical to that previously observed for another SCL neural element which lies immediately upstream of exon 1a (termed promoter 1a; see reference 38). With one important exception, the pattern also matched that observed in brain from fetal and adult mice with lacZ knocked into the SCL locus (Fig. 1B, panel xx) (9, 39). Whereas the knock-in construct gave rise to lacZ expression in blood vessels (Fig. 1B, panel xx), consistent with SCL expression in adult endothelium and our previous identification of two SCL endothelial enhancers (12, 33, 38), the +40/3.7 construct did not target adult blood vessels (Fig. 1B, panels xviii and xix) or the cells therein (data not shown; also see Fig. 6C, panel vii).

FIG. 6.

The +40/5.0-kb enhancer fragment targets additional definitive hematopoietic sites in vivo. (A) Schematic representation of a four-way sequence alignment of the MAP17 gene downstream region in humans (Homo sapiens [Hs]), dogs (Canis familiaris [Cf]), mice (Mus musculus [Mm]), and rats (Rattus norvegicus [Rn]), showing peaks of sequence homology (modified from Delabesse et al. [7]) and the position of the +40/5.0 fragment compared to the +40/3.7 region. Red boxes, coding exons; beige boxes, untranslated exons; blue boxes, repeat sequences. (B) In vitro luciferase assays of +40 fragment enhancer activity in murine F4N erythroid and BW5147 T-lymphoid cells stably transfected with luciferase reporter constructs containing either the SV40 minimal promoter alone (SV/luc) or the promoter with the +40/3.7-kb (SV/luc/+40/3.7) or the +40/5.0-kb (SV/luc/+40/5.0) fragment, respectively. Histograms represent the mean (± SD) luciferase activity (in relative light units [RLU]) of four independent pools, normalized to the mean result obtained with SV/luc. The results shown are representative of those observed in at least three independent experiments per cell line. (C) β-Galactosidase activity in SV intron/lac/+40/5.0 transgenic mice. Whole-mount E12.5 (line 5400) (i) and E7.5 (line 5405) (ii) transgenic embryos stained with X-Gal. In panel ii, maternally derived tissue (Mat) and the embryo proper (Em) are indicated. (iii) A sagittal section through another X-Gal-stained E7.5 embryo showing the allantois (Al) and defined clusters of expressing cells in the region of extraembryonic mesoderm which gives rise to yolk sac blood islands (Bi). (iv) A sagittal section through an X-Gal-stained E12.5 transgenic embryo (line 5405) showing staining in the pretectum (PT), tegmentum (Tg), and tectum (T). (v) A whole-mount X-Gal-stained transgenic adult brain (line 5400). (vi and vii) X-Gal-stained peripheral blood samples from an adult SV intron/lac/+40/5.0 (line 5402) transgenic mouse (vi) and SV/lac/+40/3.7 (line 4274) transgenic mouse (vii). At E12.5 (i and iv) and in the adult brain (v), staining was localized to blood vessels and midbrain regions (particularly the superior and inferior colliculi of adult brain). Examples shown in panels i and v and vi are representative of at least three transgenic embryos or adult mice from each of three independent lines (lines 5400, 5402, and 5405), respectively. Panel ii is representative of seven transgenic embryos from line 5405. (D) Flow cytometric analysis (FACS) of β-galactosidase activity in peripheral blood from a line 5405 SV intron/lac/+40/5.0 transgenic embryo (solid lines) and nontransgenic littermate (dashed lines) at E12.5 and E14.5 as indicated. Similar results were obtained from at least three embryos of each genotype for each time point. Analogous data were also obtained at E12.5 in an additional independent transgenic line (line 5400). (E) FACS analysis of β-galactosidase activity and Ter119 expression in E12.5 fetal liver (FL) and adult bone marrow (BM) from an SV intron/lac/+40/5.0 transgenic embryo (line 5405) or mouse (line 5402), respectively. Cells were stained simultaneously for β-galactosidase activity and the erythroid lineage marker Ter119. Similar fetal liver results were obtained with at least four transgenic embryos. Comparable bone marrow results were obtained with at least three transgenic mice from each of three independent SV intron/lac/+40/5.0 transgenic lines (lines 5400, 5402, and 5405).

The 3.7-kb +40 enhancer is active in fetal liver hematopoietic progenitors.

We have previously noted that the 3.7-kb +40 enhancer was inactive in the vast majority of definitive hematopoietic cells in the fetal liver at E14.5, a time point at which most cells are definitive erythroblasts (7). This was surprising since most erythroid regulatory elements do not discriminate between the primitive and definitive erythroid lineages. Here we confirm our previous observation (Table 1) but also demonstrate the existence of a small, but significant, β-Gal⁺ population in both transgenic E12.5 (Fig. 2A and Table 1) and E14.5 (Table 1) fetal liver. At E12.5, the greater part of this β-Gal⁺ population did not express the erythroid lineage surface marker Ter119, but a significant proportion expressed Mac-1, CD34, and/or c-Kit, all markers found on multipotent progenitors and HSCs from fetal liver (Fig. 2A and Table 1). LacZ expression in adult bone marrow and spleen was not significantly different from LacZ expression in nontransgenic controls (Table 1).

TABLE 1.

Analysis of in vivo +40 transgene activity in hematopoietic populations

Genotype	Tissue	Total β-Gal+ (%)a	Ter119+ β-Gal+ (%)b	Mac-1+ β-Gal+ (%)b	CD34+ β-Gal+ (%)b	c-Kit+ β-Gal+ (%)b
+40/3.7	E12.5 FLc	1.47 ± 0.46*	16 ± 5.5	45 ± 5.1	65 ± 7.8	91 ± 5.2
	E14.5 FLc	0.36 ± 0.09*	36 ± 11	ND	ND	86 ± 3.8
	BMe	0.09 ± 0.03
	Spleene	0.08 ± 0.01
+40/5.0	E12.5 FLd	9.73 ± 2.05*	51 ± 11	16 ± 5.8	33 ± 6.4	60 ± 11
	E14.5 FLd	4.59 ± 1.37*	53 ± 10	26 ± 8.6	23 ± 7.5	68 ± 11
	BMe	1.66 ± 0.32*	57 ± 31	ND	ND	ND
	Spleene	1.75 ± 0.85***	37 ± 17	ND	ND	ND
SCL 3′E	BMe	1.93 ± 0.56**	9.1 ± 3.6	ND	ND	ND
	Spleene	2.12 ± 0.52**	5.4 ± 2.0	ND	ND	ND

^aTotal β-galactosidase-positive (β-Gal⁺) cells (mean ± standard deviation [SD]) as a percentage of total viable cells analyzed. In analogous samples from nontransgenic controls analyzed concomitantly, <0.22% of the total viable cells were β-Gal⁺. Significant differences between the nontransgenic and transgenic populations, where P < 0.001, P < 0.005, or P < 0.02, are marked with one, two, or three asterisks, respectively. Otherwise, there was no significant difference in a t test (P > 0.05). ND, not done.

^bTotal Ter119, Mac-1, CD34, or c-Kit marker-positive, β-Gal⁺ cells (mean ± SD) as a percentage of total β-Gal⁺ cells analyzed.

^cFetal liver (FL) results from line 4092 embryos.

^dFetal liver (FL) results from line 5405 embryos.

^eBone marrow (BM) and spleen were from adult (>12 weeks of age) SV intron/lac/+40/3.7 (line 4092; also representative of lines 4102 and 4274), SV intron/lac/+40/5.0 (line 5402; also representative of lines 5400 and 5405), and SCL 3′E (+6E5/lacZ/3′E, line 2269) transgenic mice. For all results shown, at least three embryos or mice of each genotype were analyzed.

FIG. 2.

The +40/3.7 enhancer is active in E12.5 fetal liver, and a 700-bp subfragment is sufficient for +40 activity in vivo. (A) Flow cytometric analysis (FACS) of β-galactosidase activity and cell surface marker expression in E12.5 fetal liver cells from a line 4092 SV intron/lac/+40/3.7 transgenic (Tg) embryo (top row) and nontransgenic (non-Tg) littermate (bottom row). Cells were stained simultaneously for β-Gal activity and the erythroid lineage marker Ter119, Mac-1, CD34, or c-Kit as indicated. Comparable data were obtained in two independent experiments in which 10 embryos of each genotype were analyzed. (B) Schematic representation of a four-way sequence alignment of the MAP17 gene downstream region in humans (Homo sapiens [Hs]), dogs (Canis familiaris [Cf]), mice (Mus musculus [Mm]), and rats (Rattus norvegicus [Rn], showing peaks of sequence homology (modified from Delabesse et al. [7]). Red boxes, coding exons; beige boxes, untranslated exons; blue boxes, repeat sequences. The relative positions of the various +40 fragments incorporated into lacZ reporter constructs are shown below the graph of sequence homology. (C) Whole-mount X-Gal-stained E12.5 transgenic embryos bearing the SV intron/lac/+40/3.7 (line 4092) (i), SV intron/lac/+40/0.7 (line 6320) (ii), and SV intron/lac/+40/0.4 (iii) (F₀ transgenic) constructs. (D) FACS analysis of β-Gal activity in adult bone marrow (BM) from a nontransgenic littermate, an SV intron/lac/+40/3.7 transgenic (line 4092), and two SV intron/lac/+40/0.7 transgenic mice representing independent lines (6281 and 6320). Cells were stained simultaneously for β-Gal activity and the erythroid lineage marker Ter119. Similar results were obtained in an additional two SV intron/lac/+40/0.7 transgenic lines (lines 6278 and 6559).

To confirm progenitor function, β-Gal⁺ cells were sorted from E12.5 fetal liver and assayed for their ability to form definitive hematopoietic colonies in vitro. As shown in Table 2, compared to the β-Gal⁻ population, the β-Gal⁺ population was enriched (∼20-fold) for both erythroid and myeloid colonies. Our results demonstrate that the 3.7-kb +40 element was active in approximately 27% of sorted erythroid progenitors and 26% of sorted myeloid progenitors.

TABLE 2.

Hematopoietic progenitors in transgenic fetal liver and adult bone marrow

Tissue, genotype, and cell populationa	No. of erythroid coloniesb	% Total sorted erythroid progenitorsc	No. of myeloid coloniesb	% Total sorted myeloid progenitorsc
E12.5 FL
SV intron/lac/+40/3.7
β-Gal+ sorted	21 ± 11	27	127 ± 37	26
β-Gal− sorted	1.1 ± 0.7	73	6.9 ± 3.8	74
Stained, unsorted	1.3 ± 0.3		11 ± 1.2
SV intron/lac/+40/5.0
β-Gal+ sorted	11 ± 4.1	83	70 ± 16	72
β-Gal− sorted	0.5 ± 0.5	17	6.3 ± 1.7	28
Stained, unsorted	1.7 ± 0.7		11 ± 1.8
BM
SV intron/lac/+40/5.0
β-Gal+ sorted	26 ± 17	12	65 ± 34	8
β-Gal− sorted	2.5 ± 2.1	88	9.9 ± 7.3	92
Stained, unsorted	3.1 ± 2.0		10 ± 4.7

^aFL, fetal liver; BM, bone marrow.

^bThe mean number (± standard deviation) of erythroid colonies (erythroid burst-forming unit and granulocyte-sythrocyte-monocyte-megakaryocyte CFU) or myeloid colonies (granulocyte-macrophage colony-forming cells) formed per 1,000 plated E12.5 fetal liver or 5,000 bone marrow cells, respectively.

^cThe distribution of erythroid (erythroid burst-forming unit and granulocyte-sythrocyte-monocyte-megakaryocyte CFU) or myeloid (granulocyte-macrophage CFU) progenitors in each fraction, given as a percentage of total sorted (erythroid or myeloid) progenitors. Data shown are pooled from two independent fetal liver experiments each (lines 4092 and 5405) and three bone marrow (line 5402) experiments. The difference between the results obtained with the β-Gal⁺ and β-Gal⁻ sorted populations was significant in all cases (P < 0.000001).

The minimal region required for +40 activity centers on an area of high sequence homology containing multiple conserved transcription factor binding motifs.

To refine a core region responsible for enhancer activity, we generated reporter constructs containing subfragments of the 3.7-kb region. These centered on a region of high homology in human/dog/mouse/rat sequence alignments (Fig. 2B) (7). The whole 3.7-kb fragment directed expression to E12.5 embryonic midbrain and blood, as did both 0.9-kb and 0.7-kb subfragments (Fig. 2C, panels i and ii, and Table 3). However, no activity was detected in either brain or blood at E12.5 using a smaller 0.4-kb subfragment (Fig. 2C, panel iii, and Table 3). In addition, the level of activity (measured by rapidity and intensity of X-Gal staining) observed with SV intron/lac/+40/0.9 and SV intron/lac/+40/0.7 appeared generally lower than with SV (intron) lac/+40/3.7, suggesting that while specificity was maintained with the shorter +40 regions, maximal enhancer function was lost. As with SV (intron) lac/+40/3.7 transgenic mice, there was no detectable β-Gal activity in adult bone marrow from mice representing four independent SV intron/lac/+40/0.7 transgenic lines (Fig. 2D and data not shown). These data demonstrate that sequences within the 0.7-kb fragment are sufficient to direct β-Gal expression to both midbrain and primitive erythroblasts.

TABLE 3.

Analysis of +40 regions required for in vivo activity in E12.5 embryos

Construct	No. of embryos with expression of the construct/total no. of embryosa
	Blood	Midbrain	No expression	Ectopic expression
SV (intron) lac/+40/3.7	4/6	5/6	0/6	3/6
SV intron/lac/+40/0.9	1b/3	3/3	0/3	3/3
SV intron/lac/+40/0.7	2b/8	5c/8	3/8	1/8
SV intron/lac/+40/0.4	0/4	0/4	3/4	1/4
SV intron/lac/+40/3.7 ΔGE1 + 2	0/10	0/10	4/10	6d/10
SV intron/lac/+40/3.7 ΔGE1	2/6	3c/6	2/6	3/6
SV intron/lac/+40/3.7 ΔGE2	4/13	1/13	7/13	3d/13
SV intron/lac/+40/3.7 mGE1 + 2	0/11	0/11	4/11	7d/11
SV intron/lac/+40/3.7 mGE1	4/8	4/8	2/8	2/8
SV intron/lac/+40/3.7 mGE2	2/4	0/4	1/4	1/4
SV intron/lac/+40/5.0	9/10	9/10	1/10	3/10

^aResults represent four SV (intron) lac/+40/3.7 transgenic lines and two F₀ transgenic E12.5 embryos, five SV intron/lac/+40/0.7 transgenic lines and three F₀ transgenic E12.5 embryos, and three SV intron/lac/+40/5.0 transgenic lines and seven F₀ transgenic E12.5 embryos, respectively. All other results denote F₀ transgenic embryos analyzed at E12.5.

^bWeak staining and/or staining in only a proportion of circulating blood cells.

^cWeak staining in midbrain.

^dEctopic staining in the brains of some (1 of 10 SV intron/lac/+40/3.7 ΔGE1 + 2, 1 of 13 SV intron/lac/+40/3.7 ΔGE2, and 3 of 11 SV intron/lac/+40/3.7 mGE1 + 2) embryos, which was apparently not related to the endogenous SCL midbrain expression pattern.

A four-way human/dog/mouse/rat sequence alignment of this region is shown in Fig. 3. The 0.7-kb fragment and the 3′ boundary of the 0.4-kb fragment are indicated. Several conserved consensus binding sites for hematopoietic transcription factors were identified, including two GATA sites, two E-boxes (CANNTG), and two Ets (GGAW) sites. Both GATA sites lie in close proximity to an E-box. In all four species, the 5′ GATA site and E-box were separated by 9 bp, and the 3′ GATA site and E-box were separated by 6 bp. This spacing indicates that, for each GATA/E-box motif, DNA-bound GATA and basic helix-loop-helix proteins would be on the same face of the double helix and capable of interacting.

FIG. 3.

The +40 core region contains multiple conserved transcription factor recognition sequences. A four-way sequence alignment of the +40/0.7 region (boxed) in human SCL (Homo sapiens [hsscl]), dog SCL (Canis familiaris [cfscl]), mouse SCL (Mus musculus [mmscl]), and rat SCL (Rattus norvegicus [rnscl]) genes, showing potential Ets, GATA, and E-box (helix-loop-helix factor) binding sites. An RsrII restriction enzyme recognition site located between the two GATA/E-box motifs and present only in the mouse sequence is indicated. Completely conserved residues are shown by white letters on a black background, and residues conserved in three of the four species are shown by white letters on a gray background. The +40/0.4 core region fragment shares the same 5′ boundary as the +40/0.7 fragment, and its 3′ border is indicated by a black arrowhead. Gaps introduced to maximize alignment are indicated by the dashes.

A 400-bp core region is required for 3.7-kb enhancer activity in vivo and contains GATA/E-box motifs with both essential and redundant functions.

Although apparently not sufficient for +40 transgene activity in vivo, the high level of sequence conservation within the +40/0.4-kb core indicated that this region might be necessary for +40 function. We therefore deleted from the original +40/3.7-kb fragment, either the proximal ∼200 bp (from the beginning of the +40/0.4 region to an internal RsrII site) containing the 5′ GATA/E-box motif, the distal ∼200 bp (from the internal RsrII site to the end of the +40/0.4 region) containing the 3′ GATA/E-box motif, or the entire 400-bp +40/0.4 region (Fig. 3 and 4). Luciferase reporter constructs containing each of these deletion fragments (SV/luc/+40/3.7 ΔGE1, SV/luc/+40/3.7 ΔGE2 and SV/luc/+40/3.7 ΔGE1 + 2) were tested for enhancer activity by stable transfection using the erythroid F4N cell line (Fig. 4A). Deletion of the entire 400-bp core region or its 5′ half resulted in almost complete loss of enhancer function, whereas deletion of the 3′ 200 bp had little or no effect (Fig. 4A).

FIG. 4.

The +40/0.4 core region, containing GATA/E-box motifs, is required for in vivo activity. (A) Schematic representation of the SV intron/luc/+40/3.7 deletion constructs and in vitro luciferase assays of their enhancer activity in stably transfected F4N cells compared to the original +40/3.7 wild-type construct. Histograms represent the mean (plus standard deviation [error bar]) luciferase activity (in relative light units [RLU]) of four independent pools, normalized to the mean result obtained with SV/luc. The results shown are representative of those observed in at least three independent experiments. (B) Whole-mount X-Gal-stained E12.5 F₀ transgenic embryos bearing the SV intron/lac/+40/3.7 ΔGE1 + 2 (i), SV intron/lac/+40/3.7 ΔGE1 (ii), or SV intron/lac/+40/3.7 ΔGE2 (iii) deletion constructs in which the entire +40/0.4-kb region or the 5′ or the 3′ 0.2 kb, respectively, had been deleted. Midbrain staining in the SV intron/lac/+40/3.7 ΔGE1 embryo shown in panel ii is indicated by an arrowhead.

The same +40 deletion fragments were used to generate corresponding lacZ reporter constructs (SV intron/lac/+40/3.7 ΔGE1, SV intron/lac/+40/3.7 ΔGE2, and SV intron/lac/+40/3.7 ΔGE1 + 2). F₀ transgenic embryos were generated and analyzed for β-Gal activity at E12.5. Consistent with our transfection data, transgene activity was not detected in either the midbrain or blood of any of 10 embryos bearing the SV intron/lac/+40/3.7 ΔGE1 + 2 transgene (Fig. 4B, panel i, and Table 3). These data demonstrate that the 400-bp core region is necessary for appropriate transgene action in vivo. One-third of the SV intron/lac/+40/3.7 ΔGE1 transgenic embryos analyzed displayed β-Gal activity in both blood and midbrain (Fig. 4B, panel ii, and Table 3), indicating that the 5′ half of the 400-bp core region is not required for enhancer activity in either tissue. Similarly, 4 of the 13 SV intron/lac/+40/3.7 ΔGE2 transgenic embryos had β-Gal activity in peripheral blood, whereas 1 of 13 had appropriate expression in the midbrain (Fig. 4B, panel iii, and Table 3). Taken together, these data demonstrate that the 5′ and 3′ halves of the 400-bp core region have redundant functions in determining the cell type specificity of enhancer action in blood (primitive erythropoiesis) and midbrain.

To investigate the importance of individual binding sites within the 0.4-kb core region, we mutated individually, or in combination, each of the two conserved GATA and E-box recognition sites (Fig. 5A) and tested their enhancer activity in vitro in F4N cells (Fig. 5B). Compared to the original 3.7-kb fragment, there was an almost complete loss of activity when all four sites were mutated, suggesting that one or more of these sites are essential for enhancer activity in this cell line. Mutation of either the 5′ GATA (data not shown), 5′ E-box (data not shown), or both sites (Fig. 5B), also resulted in almost complete loss of enhancer activity, whereas the effect of mutating the 3′ GATA and/or E-box sites was minimal (Fig. 5B and data not shown). These data are consistent with the results observed using deletions within the 3.7-kb fragment (Fig. 4A).

FIG. 5.

GATA/E-box motifs, present in the +40 enhancer core, are necessary for its in vivo activity. (A) Sequences of the two wild-type GATA/E-box motifs found within the +40/0.4 core with the corresponding mutated sequences (mGE1 and mGE2) shown below. The individual residues that differ between the wild-type +40/3.7 and mutated fragments are shown in white letters on a black background. (B) Schematic representations of the SV intron/luc/+40/3.7 mutation constructs and in vitro luciferase assays of their enhancer activity in stably transfected F4N cells compared to the original +40/3.7 wild-type construct. Histograms represent the mean (plus standard deviation [SD] [error bar]) luciferase activity (in relative light units [RLU]) of four independent pools, normalized to the mean result obtained with SV/luc. The results shown are representative of those observed in at least three independent experiments. (C) Whole-mount X-Gal-stained E12.5 F₀ transgenic embryos bearing the SV intron/lac/+40/3.7 mGE1 + 2 (i), SV intron/lac/+40/3.7 mGE1 (ii), or SV intron/lac/+40/3.7 mGE2 (iii) mutation constructs in which both motifs, the 5′ GATA/E-box motif or the 3′ GATA/E-box motif, respectively, were mutated. (D and E) Chromatin immunoprecipitation of GATA-1 (D) and SCL (E) proteins in the F4N erythroid cell line, analyzed by real-time PCR. The results are presented as mean change in enrichment (plus SD [error bar]) obtained using the GATA-1 and SCL antibodies compared to rabbit IgG. The transcriptionally inactive SCL −16 region and the α-globin HS-8 were used as negative (Neg) and positive (Pos) controls, respectively.

Transgenic analysis was then performed using mutant SV intron/lac/+40/3.7 constructs in which either the 5′ GATA/E-box motif, the 3′ GATA/E-box motif, or all four sites were mutated (Fig. 5C and Table 3). All 11 F₀ SV intron/lac/+40/3.7 mGE1 + 2 transgenic E12.5 embryos, in which both 5′ and 3′ GATA/E-box motifs were mutated, lacked appropriate expression in both blood and midbrain (Fig. 5C, panel i, and Table 3). Mutation of the 5′ GATA/E-box motif resulted in expression in the midbrain and blood in four out of eight transgenic E12.5 embryos (Fig. 5C, panel ii, and Table 3). In contrast, mutation of the 3′ GATA/E-box motif resulted in expression in blood (in two of four transgenic E12.5 embryos), but not in the midbrain (Fig. 5C, panel iii, and Table 3). Again, these observations are in line with those obtained using deletion constructs (Fig. 4B and Table 3).

To identify proteins bound to the conserved GATA/E-box motifs in vivo, we performed chromatin immunoprecipitation in the erythroid cell line F4N. Antibodies against GATA-1 and SCL were used. In erythroid cells, these proteins are part of transcription activation complexes that specifically target GATA/E-box motifs (32, 42). A 3.5-fold enrichment for the +40 enhancer genomic region was observed when using anti-GATA-1 antibody, compared to nonspecific rabbit IgG (Fig. 5D). By contrast, there was no enrichment for a transcriptionally inactive negative-control region, located in the 16-kb region upstream of SCL exon 1 (7). An α-globin HS-8 positive-control region, containing a GATA/E-box motif known to be bound by SCL and GATA-1 (3), showed a 3.6-fold enrichment. Using an anti-SCL antibody, a 20-fold enrichment for the +40 enhancer was seen (Fig. 5E). Again, no enrichment was observed for the −16 negative control, whereas the α-globin HS-8 region showed a 14-fold enrichment. These results confirm that, in F4N erythroid cells, GATA/E-box motifs present in the +40 enhancer are physically bound by GATA-1 and SCL.

Taken together, our results demonstrate that at least one +40 GATA/E-box motif is required for blood activity of the +40 enhancer. Furthermore, out of a total of 38 independent transgenic embryos, in which the 3′ GATA/E-box motif was either mutated or absent, only one embryo displayed β-Gal activity in the midbrain. This observation suggests that the 3′ GATA/E-box motif, although not absolutely essential, is particularly important for midbrain enhancer function and that the 5′ GATA/E-box motif can only partially compensate for its loss.

A larger +40/5.0-kb enhancer fragment directs expression to adult blood and definitive bone marrow progenitors.

We have previously reported that the 3.7-kb +40 enhancer is active in primitive, but not definitive, erythroblasts (7), and yet, SCL is expressed in both the primitive and definitive erythroid lineages. A region of relatively high sequence conservation lies immediately upstream of the 3.7-kb fragment (Fig. 6A), indicating that additional upstream sequences may be required for full +40 erythroid enhancer activity. To address this possibility, luciferase and lacZ reporter constructs were generated in which 1.3 kb of upstream sequence was added to the original 3.7-kb fragment (Fig. 6A). In stable transfection assays using the erythroid cell line F4N, the 5.0-kb enhancer was ∼10-fold more active than the 3.7-kb enhancer (Fig. 6B). By contrast, neither the 5.0-kb nor the 3.7-kb enhancer fragments were active in the T-cell line BW5147 that does not express SCL (Fig. 6B). The 5.0-kb fragment therefore maintains the lineage specificity previously noted for the 3.7-kb fragment (7).

In transgenic midgestation embryos, the 5.0-kb fragment gave rise to the same pattern of expression as that of the 3.7-kb fragment. However, SV intron/lac/+40/5.0 embryos displayed an apparently higher level of β-Gal activity (indicated by quicker, more intense X-Gal staining) and less inappropriate ectopic expression than 3.7-kb enhancer counterparts (Table 3). In particular, activity was noted in the midbrain and blood in 9 of 10 independent E12.5 F₀ transgenic embryos or lines (Fig. 6C, panel i, and Table 3). Transgene activity was already evident in the extraembryonic mesoderm at E7.5 (Fig. 6C, panels ii and iii). At E12.5, midbrain staining was evident in the pretectum, tegmentum, and tectum (Fig. 6C, panel iv). Midbrain expression continued postnatally, particularly in the superior and inferior colliculi of the adult brain (Fig. 6C, panel v). Further, in adult mice representing three independent transgenic lines, and in stark contrast to the +40/3.7 transgenic mice, there was abundant staining in blood vessels (compare Fig. 6C, panel v, with 1B, panel xviii), specifically in circulating blood cells (compare panels vi and vii in Fig. 6C).

Flow cytometric analysis (fluorescence-activated cell sorting [FACS]) analysis of β-Gal activity in peripheral blood established that the 5.0-kb fragment directed expression to circulating erythroid cells at both E12.5 and E14.5 (Fig. 6D), time points when primitive and definitive erythroid cells predominate, respectively. By contrast, the 3.7-kb enhancer was not active in peripheral blood at E14.5 (7; also data not shown). In E12.5 fetal liver, the 5.0-kb enhancer gave rise to β-Gal activity in a greater proportion of cells than the 3.7-kb enhancer (∼10% versus ∼1.5%, respectively) (Table 1; compare Fig. 6E to 2A). With the 5.0-kb enhancer, approximately half of the β-Gal⁺ cells were also Ter119⁺ (Table 1) and significant proportions were positive for Mac-1, CD34, and c-Kit (Table 1). In E14.5 fetal liver, a similar pattern was observed, with the 5.0-kb fragment again far more active than the 3.7-kb fragment (4.6% versus 0.36%; see also Table 1). Moreover, in marked contrast to the 3.7-kb fragment, the 5.0-kb fragment gave rise to clear activity (particularly in Ter119⁺ erythroid cells) in adult bone marrow (Fig. 6E and Table 1) and spleen (Table 1), but not thymus (data not shown).

To confirm progenitor function, β-Gal⁺ cells were sorted from fetal liver or adult bone marrow and subjected to in vitro colony-forming assays. The 5.0-kb enhancer was active in ∼83% and ∼72% of fractionated erythroid and myeloid E12.5 fetal liver progenitors, respectively, substantially higher percentages than were obtained using the 3.7-kb enhancer (Table 2). There was significant enrichment for erythroid (∼20-fold) and myeloid (∼10-fold) progenitors in the β-Gal⁺ fraction compared to the sorted β-Gal⁻ population. Similarly, in adult bone marrow, erythroid (∼10-fold) and myeloid (∼7-fold) progenitors were also enriched in the β-Gal⁺ fraction, with ∼12% and ∼8% of sorted erythroid and myeloid progenitors found in the β-Gal⁺ fraction (Table 2).

DISCUSSION

SCL +40 enhancer activity centers on a 400-bp core and is mediated by GATA/E-box motifs.

In this paper, we demonstrate that a 0.7-kb SCL +40 enhancer fragment is sufficient for activity in embryonic blood and midbrain. A smaller 0.4-kb core region and the two conserved GATA/E-box motifs it contains were shown to be essential for enhancer function in both transfection studies and transgenic mice. Moreover, both GATA-1 and SCL bind this region in definitive erythroid cells. In the context of the 3.7-kb enhancer, mutation of both GATA/E-box motifs resulted in loss of enhancer function, whereas either motif was able to independently support activity in blood. By contrast, the 3′ GATA/E-box motif was particularly important for midbrain enhancer function.

GATA/E-box motifs are not present in the SCL promoter or in five other SCL enhancers characterized so far (4, 11-13). However, GATA/E-box motifs have been described in a number of hematopoietically expressed genes, including EKLF (1, 2), GATA-1 (25, 41), α-globin (3), glycophorin A (20), EPCR (22), Cdc6 (40), and protein 4.2 (43). Many of the motifs present in these genes are in agreement with, but do not necessarily exactly match, the consensus sequence described by Wadman and colleagues, of an E-box (CAGGTG) and a GATA binding site [(A/T)GATA(A/G)] separated by 8 to 10 bases (42). In erythroid cells, GATA/E-box motifs mediate the recruitment of a complex containing SCL, E2A, LMO2, LDB1, and GATA-1 (32, 42). Within the SCL +40 enhancer, the structure of the 5′ GATA/E-box motif is generally in accord with that of the consensus sequence, with the GATA and E-box binding sites separated by 9 bases. By contrast, the 3′ GATA/E-box motif exhibits 6-bp spacing. Such spacing was previously shown to preclude recruitment of the SCL complex in erythroid cell lines (42). This is consistent with our observation that the 3′ GATA/E-box motif was particularly important for +40 enhancer activity in the midbrain but contributed little to luciferase activity in erythroid cells in vitro. However, some flexibility in the nature and arrangement of sites within GATA/E-box motifs exists. Indeed, partially overlapping GATA and E-box binding sites are present in motifs found in both the erythroid lineage marker gene glycophorin A (20) and in the human Cdc6 gene (40), which is expressed in megakaryocytic cells. Interestingly, in primary hematopoietic cells, the glycophorin A promoter is bound in vivo by a complex containing SCL and GATA-1, and the assembly of this complex is necessary for promoter activation (20). This implies that the spacing described by Wadman and colleagues (42) is not an absolute requirement for function. With reference to the +40 enhancer, complexes containing different components presumably exhibit differential affinity for the 5′ and 3′ GATA/E-box motifs. It may be particularly relevant that erythroid cells contain high levels of GATA-1 (10), whereas the midbrain regions in which the +40 enhancer is active express GATA-2 and GATA-3 (18, 24), but not GATA-1 (38). However, deletion of the 5′ motif did not completely abolish hematopoietic activity in transgenic mice, suggesting that, despite its distinct structure, the 3′ motif can compensate to some extent for loss of the 5′ motif in vivo. Similarly, the reverse is also true.

The +40 enhancer is active in both primitive and definitive hematopoietic cells in vivo.

Compared to the 3.7-kb construct, the 5.0-kb +40 construct gave rise to increased activity in erythroid cells in transfection assays and more robust reporter expression in transgenic embryos. Both were active in E7.5 extraembryonic mesoderm, primitive erythroblasts, and definitive fetal liver progenitors. Importantly and in contrast to the 3.7-kb construct, the 5.0-kb construct also directed expression to erythroid cells in adult blood, bone marrow, and spleen, as well as to erythroid and myeloid bone marrow progenitors. Several mechanisms might account for this more extensive expression pattern. The 1.3-kb region, which distinguishes the two constructs, lies directly 5′ of the 3.7-kb fragment and might conceivably contain a distinct, novel enhancer with activity in definitive erythroid cells. However, analysis of the aligned upstream region uncovered no further conserved GATA or E-box sites and few other conserved factor binding sites. The concept of a separate enhancer responsible for activity in definitive hematopoietic cells is rendered less likely by the fact that the 3.7-kb construct does target a minority of definitive fetal liver progenitors. We therefore favor the possibility that our observations reflect quantitative differences in the activity of the 3.7-kb and 5.0-kb constructs, with better maintenance of 5.0-kb construct function in adult hematopoiesis. Rather than representing an independent enhancer, sequences present in the 5.0-kb, but not 3.7-kb, fragment, may facilitate increased or continued expression by maintaining an open chromatin domain and/or counteracting nearby negative influences. The latter may include integration site effects (19) or a repressive effect of the β-galactosidase reporter itself (15, 16, 26, 37).

In addition to the +40 enhancer, two other SCL enhancers (−4 and +18/19) have been identified that are active in hematopoietic cells. These direct expression to HSCs, hematopoietic progenitors, and endothelium, but not to erythroid cells beyond early progenitor stages (12, 33). A reporter construct driven by the +18/19 enhancer is active in frog hemangioblasts, and it has been proposed that enhancer activity in both endothelial and hematopoietic progenitors is characteristic of genes, such as SCL, which exhibit hemangioblast expression (13, 28). The absence of +40 enhancer activity in endothelial cells indicates that it is unlikely to be active in hemangioblasts. Our results suggest a model in which the +18/19 and/or the −4 enhancers are responsible for initiation of SCL transcription in early mesodermal derivatives with the potential to generate both blood and endothelium. By contrast, the +40 enhancer may be recruited at a later stage of development, perhaps via an autoregulatory loop, thereafter playing a key role in directing SCL expression in the erythroid lineage.

Midbrain +40 enhancer activity closely mimics that of the endogenous SCL promoter.

Expression of SCL in the midbrain and spinal cord (9, 38, 39) is maintained throughout vertebrate evolution. In spinal cord, SCL directs neuronal progenitors to adopt particular cell fates (23). During midbrain development, SCL is expressed mainly in the posterior and inferior commissure, and in the adult, in the superior and inferior colliculi (9, 39). These CNS regions, where SCL plays a nonredundant role, are associated with the processing of visual and auditory reflexes, and of pain-related information (5). We have now shown that the +40 enhancer directs lacZ transgene expression to the midbrain in a pattern indistinguishable from that generated by SCL promoter 1a (38). Both target the same regions of embryonic midbrain, including the ventrolateral nuclei, the posterior and inferior commissure, and the diencephalic extensions. Similarly, in the adult, both direct expression to the superior and inferior colliculi. Their apparent redundant functions may relate to the biological importance of SCL for the development of these midbrain regions. Further, both elements appear to be regulated by GATA proteins (this study and reference 38). However, this likely reflects distinct mechanisms, since promoter 1a lacks a recognizable GATA/E-box motif. As suggested above for hematopoietic cells, binding of GATA factors to SCL promoter 1a in neuronal cells might initiate expression, thereby generating SCL protein subsequently involved in +40 midbrain enhancer activity.

Although the promoter 1a and the +40 enhancer elements have very similar activities in midbrain, their activity in blood differs. Unlike the +40 enhancer, promoter 1a does not direct lacZ expression to hematopoietic cells in transgenic mice (38). The GATA/E-box sites found within the +40 enhancer are necessary for both hematopoietic and midbrain activity; however, they are not sufficient, since the 0.4-kb +40 core fragment was inactive in both transfection studies and transgenic mice. Further analysis of the +40 enhancer, its constituent sites, and mode of action is likely to provide additional valuable insights into SCL transcriptional regulation.