The human β-globin locus control region confers an early embryonic erythroid-specific expression pattern to a basic promoter driving the bacterial lacZ gene

The locus control region (LCR) of the human β-globin multi-gene cluster is a cis-acting regulatory element present 5-25 kb upstream of the ε-globin gene. The cluster contains five developmentally regulated genes in the order 5′-embryonic (ε), fetal (γ;G and γ;A) and adult (δ and β)-3′. The regulatory function of the LCR became first apparent from the study of a Dutch thalassemia (Kioussis et al., 1983) and other naturally occurring deletions of the region where the intact globin gene is not activated (Driscoll et al., 1989; Taramelli et al., 1986). It has been shown that deletion of the LCR not only affects expression of the genes, but also the chromatin structure of the locus over a distance of at least 200 kb (Forrester et al., 1990).

Chromatin analysis of the LCR showed that it contains a set of developmentally stable DNaseI-hypersensitive sites (HS) termed 5′HS1-5 (Grosveld et al., 1987; Tuan et al., 1985). Various studies have been undertaken to understand the nature and function of this region. These studies revealed that the LCR, when coupled to globin genes in cis, confers high level, copy-number dependent- and position-independent expression to these genes in erythroid tissues in transgenic mice (Grosveld et al., 1987; Ryan et al., 1989; Talbot et al., 1989). 5′HS2-4 also have these properties when linked as single HS to the β-globin gene (Fraser et al., 1990). Transfection studies stated that only 5′HS2 acts as an enhancer in transient assays (Tuan et al., 1989; Moon and Ley, 1991).

A fully active 6.5 kb version of the LCR containing 5′HS1-4, termed the μLCR, has been described (Talbot et al., 1989). Further dissection of the μLCR revealed that the activity of each HS can be localized to 200-300 bp core fragments coin-ciding with the position of hypersensitivity in chromatin (Philipsen et al., 1990; Pruzina et al., 1991; Talbot et al., 1990). The minimal element required for LCR activity of 5′HS3 was assigned to a 125 bp fragment (Philipsen et al., 1993). Inter-estingly, recent observations from our laboratory suggest that 5′HS3 is the only site that possesses significant chromatin opening activity in single copy transgenic mice (Ellis et al., 1996).

The LCR core fragments all contain binding sites for the erythroid-specific protein GATA-1, and ‘GGTGG’ or ‘CACC’ box motifs binding members of the Sp1 family of transcription factors (Ellis et al., 1993; Philipsen et al., 1990, 1993; Pruzina et al., 1991, 1994; Talbot et al., 1990). This combination of binding sites is also found in the promoters and enhancers of the β-globin locus as well as many other erythroid-specific genes (Philipsen et al., 1990 and references therein). Hence, we have asked whether the μLCR and its individual HS can function in the absence of any other erythroid regulatory elements. To answer this, we have used the bacterial lacZ gene driven by a non-erythroid, 80 bp minimal promoter derived from the mouse heat shock protein 68 (Hsp68) gene (Kothary et al., 1987, 1988, 1989). The results show that only the μLCR and 5′HS3 are able to direct the expression of this gene in the erythroid lineage reproducibly. Interestingly, expression is limited to the embryonic stages of development, after which the chromatin of the transgenes is in an inactive conformation and transcription is repressed. Our data suggest that the presence of a β-globin gene is essential for LCR function in definitive erythroid cells, implying that there are more stringent requirements for erythroid-specific gene expression in later developmental stages than during the embryonic period.

Five different fragments, 5′HS4 (2.1 kb BamHI-XbaI), 5′HS3 (1.9 kb HindIII), 5′HS2 (1.5 kb Asp718-BglII), 5′HS1(1 kb SstI (partial)-Hind III) and μLCR (6.5 kb, containing 5′HS1-4) (Talbot et al., 1989) were linked to the mouse Hsp68 minimal promoter driving the bacterial lacZ reporter gene. Each of these fragments was isolated as described (Kollias et al., 1986). The fragments were dissolved in micro-injection buffer (10 mM Tris-Cl, pH 7.5 and 0.1 mM EDTA) at a concentration of 1-2 μg/ml and microinjected into the pronucleus of FVB fertilized mouse eggs. The injected eggs were transferred into the oviducts of pseudo-pregnant foster females as described (Kollias et al., 1986).

For the ‘transient’ assay, the foster females were killed 13.5 days after egg transfer and the fetuses were collected from the uterus. For stable lines, transgenic founders were bred to non-transgenic mice. After the lines were established the F1 or F2 males were used for mating with non-transgenic females for the collection of embryos and fetuses at various stages of development.

Genomic mouse DNA was prepared from part of the body of the fetuses or tail clips from adult mice. The DNA was cut with EcoRI, Southern blotted and probed with lacZ to detect the transgene. A fragment from the mouse ThyI gene was used as a loading control (Talbot et al., 1990).

Nuclei were isolated from 8-10 transgenic fetal livers (embryonic day 13.5, E13.5) and DNaseI hypersensitive site mapping was performed as described (Ellis et al., 1996). After this treatment the DNA was isolated and analyzed by digesting with EcoRI. The blots were first hybridized with the Sau3AI-XbaI fragment, coordinates 39819-40269 (Shehee et al., 1989) detecting the hypersensitive site at the promoter of the mouse β-major gene, and then with the lacZ probe to detect hypersensitive sites at the transgene.

For the collection of mice at various developmental stage, transgenic males from each line were kept with wild-type FVB female mice. The day of the vaginal plug was taken as E0.5. The embryos were dissected out of the decidua and fixed on ice in a solution containing 1% formaldehyde, 0.5% glutaraldehyde in PBS for 30 minutes. They were then rinsed in embryo buffer (PBS, 0.02% NP40, 0.01% deoxy-cholate, 2 mM MgCl2) and incubated in staining solution (1 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galacto-pyranoside (X-gal), 5 mM K3Fe(CN)6, 5 mM K4Fe(CN)6 and 1 mM ethylene glycol-bis( β-aminoethyl ether) N,N,N′,N′-tetraacetic acid (EGTA) in embryo buffer). After 3-4 hours at 37°C, or overnight at room temperature in the dark, the staining reaction was stopped by removing this solution and rinsing the embryos in PBS. The same procedure was followed for E10.5 and E13.5 blood staining, except that the cells were fixed for only 5 minutes.

E10.5 embryonic blood from the heart or E13.5 liver cells from indi-vidual fetuses were disrupted in 50 and 150 μl of PBS respectively. 20 μl of this cell suspension was fixed onto a poly(lysine)-coated slide in 4% formaldehyde, 5% acetic acid for 20 minutes at room temper-ature. The cells were then washed 3× for 10 minutes in PBS and stored in 70% ethanol at −20°C. The remains after collection of blood or liver were stained for β-galactosidase to detect the transgenics.

Two probes were utilized for in situ hybridization. A nick-trans-lated biotinylated lacZ probe for the detection of lacZ transcription, and DIG-labelled mouse α intron-specific oligonucleotides to reveal the erythroid cells. As a control, some of the slides were RNase-treated and hybridized with these probes. Hybridizations were done as described by Wijgerde et al. (1995) with a few modifications. The lacZ hybridization mix contained 50% formamide and 10% dextran sulphate. For antibody detection, the Tyramid amplification system TSA of Dupont was used (Raap et al., 1995).

Murine erythroleukemia cells (MEL, strain C88) were maintained in DMEM plus 10% fetal calf serum. The construct μLCR-lacZ with neomycin as selectable marker was linearized at the ScaI site and transfected by electroporation as described (Antoniou et al., 1988; Antoniou and Grosveld, 1990). Stably transfected MEL populations were selected in G418 and induced to undergo erythroid differen-tiation by the addition of DMSO to a final concentration of 2% (v/v) for 2 days. The cells were collected and used for hypersensitive-site mapping. The DNaseI-treated samples were cut with EcoRI and SacII. The Southern blots were hybridized with the 1.9 kb fragment of 5′HS3 (Talbot et al., 1989) or the lacZ probe to detect the hypersensitive sites.

In order to analyze the expression status of the LCR and its individual hypersensitive sites (HS) with a ubiquitous promoter, each of these sites (Talbot et al., 1989) was linked to the minimal mouse Hsp68 promoter. This 80 bp promoter contains a CCAAT box, an Sp1 binding site and a TATA box. All these constructs used the bacterial β-galactosidase (lacZ) gene as a reporter, to allow visualization of expression by X-gal staining (Bonnerot and Nicolas, 1993). These constructs were first tested in transgenic founder animals. Fetuses and yolk sacs were collected at 13.5 days of development (hence-forth referred to as E13.5) and analyzed by X-gal staining All fetuses were typed for the transgene by Southern blot hybrid-ization. The results are summarized in Table 1.

LacZ expression was observed in blood, yolk sac and fetal liver of a few founder animals containing constructs with 5′HS1 or 5′HS4. However, no hematopoietic-specific staining was observed in founder fetuses with 5′HS2. In most of the 5′HS1, 5′HS2 and 5′HS4 founders, X-gal staining was observed in various other parts of the fetuses. This type of staining corresponds to classical position effects, to which the minimal Hsp68 promoter is females to collect embryos and fetuses at different develop-mental time points. very sensitive (Kothary et al., 1987, 1988, 1989). These position effects show that the transgenes are functional and have not integrated in an inher-ently closed chromatin domain of the mouse genome. In sharp contrast, X-gal staining could be easily detected in the erythroid cells of the yolk sac, fetal liver and blood vessels in the majority of founder fetuses with either the μLCR or the 5′HS3 transgene. As expected, ectopic expression was also observed, since the LCR or parts of it do not neutralize position effects in non-erythroid tissues (Grosveld et al., 1987).

Previously, it has been shown that each of the hypersensitive sites of the LCR may have some developmental specificity (Fraser et al., 1993), and hence it was possible that the absence of erythroid staining in the 5′HS1, -2 and -4 founder fetuses is due to the stage specificity of these individual sites. Therefore, we studied the expression of the constructs in stable transgenic lines at different developmental stages. The data for each of the estab-lished stable transgenic lines are summarized in Table 2. F1 or F2 males from these lines were mated with non-transgeni.These results revealed that, at E13.5, X-gal staining of the erythroid cells could only be detected in the transgenic lines containing the μLCR or 5′HS3 constructs (Fig. 1D,F). This is in agreement with the data obtained from the founder fetuses (Table 1). Staining of the erythroid system was not seen for the other constructs (Fig. 1B,C,E). Again, expression outside the erythroid system, due to position effects, was observed in many of the lines with all of the constructs (Fig. 1 and Tables 1, 2).

The expression status of the transgenes at earlier stages in mouse development was examined by X-gal staining of embryos at various time points starting from the onset of hematopoiesis at E7.5-8.Embryos containing the 5′HS1, 5′HS2 and 5′HS4 transgenes showed no staining in the yolk sac or blood cells at any of the stages examined (Table 2). However, the μLCR and 5′HS3 lines showed the first erythroid-specific staining at E8-8.5, with a ring of blue cells observed at the base of the yolk sac (Fig. 2B, arrow). Later, X-gal staining cells are found over the entire yolk sac (Fig. 2C,D). In E9-E10.5 embryos, the staining is seen in the cells within the blood capillaries (Fig. 2E-G), the heart (Fig. 2G, arrows) and the dorsal aorta (Fig. 2G, arrowheads). In addition, staining of the fetal liver is observed in E13.5 fetuses (Fig. 1D,F, arrows). Surpris-ingly, very few blue cells are found in adult blood from the same transgenic lines (Fig. 3A). This prompted us to study the blood from E10.5 and E13.5 transgenics in more detail. While >98% of the cells are lacZ-positive in E10.5 blood (Fig. 3F), this number decreases to around 50% in E13.5 blood (Fig. 3E). Most of the cells staining at the early stages are presum-ably embryonic primitive erythroid cells, since they are large and nucleated (Fig. 3E,F). Strikingly, less than 10% lacZ-positive cells are found in the fetal liver (Fig. 3C), the site of erythropoiesis at E13.5 (see Fig. 1D,F, arrows). These data therefore suggest that the μLCR and 5′HS3 transgenes are active in primitive cells of embryonic origin, but not in fetal- and adult-type definitive erythroid cells.

The LCR is characterized by the presence of developmentally stable, very strong, erythroid-specific DNaseI hypersensitive sites (Grosveld et al., 1987; Jimenez et al., 1992b; Tuan et al., 1985). To determine whether these hypersensitive sites are correctly established in our transgenic mice, we examined the chromatin structure in E13.5 liver of the lacZ transgenic lines described above. Nuclei from the livers of E13.5 fetuses were treated with DNaseI (Ellis et al., 1996), after which the purified DNA was digested with EcoRI and Southern-blotted. As a control, the Southern blots were first hybridized with a probe detecting the hypersensitive site at the mouse β-major promoter to assess DNaseI sensitivity. The blots were then hybridized with a lacZ probe to detect hypersensitive sites at the transgenes. As a reference, tail DNA from each transgenic mouse line was cut with EcoRI and NcoI to show the position of the Hsp68 promoter. To mark the positions of the core elements of the LCR fragments, EcoRI-digested tail DNA was recut with SpeI for 5′HS1, HindIII for 5′HS2, ApaI for 5′HS3 and SphI for 5′HS4 (Talbot et al., 1989).

The results obtained from this analysis on E13.5 liver show that hypersensitive sites are not detectable at the correct positions within any of the transgenes, while the hypersensi-tive site at the mouse β-major promoter could be detected con-sistently (Fig. 4). For example, the weak sensitivity seen in DNA from the 5′HS2-lacZ line b (grey arrow in Fig. 4D) does not map to the core fragment of 5′HS2, nor to the Hsp68 promoter and most likely reflects the general, non-erythroid, DNaseI sensitiv-ity associated with 5′HS2 (e.g. Ellis et al., 1996). These results are not surpris-ing for the lines with the 5′HS1, -2 and-4 constructs, since we could not detect erythroid-specific expression of the lacZ gene in these mice. However, no DNaseI hypersensitivity is observed in the DNA from the lines with the μLCR and 5′HS3 constructs (Fig. 4B,E), even though lacZ protein is easily detectable at this stage (Fig. 1D,F; Fig. 3C). Since it is possible that the lacZ gene was transcribed only at earlier stages in fetal liver develop-ment, we repeated the DNaseI fade-out analysis on E12.5 livers of the μLCR and 5′HS3 lines. These results were also negative (data not shown) and thus, we conclude that the transgenes are in a closed chromatin conformation in the fetal liver.

Since the numbers of erythroid cells that can be obtained from E8.5-E10.5 embryos are too small for a DNaseI sen-sitivity analysis (N. Gillemans, unpub-lished results), we used in situ hybridiz-ation as an alternative to monitor the activity of the μLCR and 5′HS3 trans-genes. To more closely examine when and at what frequency the μLCR- and 5′HS3-lacZ transgenes are actively tran-scribed, we analyzed specific transcrip-tion of the lacZ gene at the single cell level (Wijgerde et al., 1995). In situ hybridization was performed on erythroid cells from E10.5 blood and E13.5 liver with a biotin-labelled lacZ probe to detect transcription of the transgenes. Oligonucleotide mouse α-globin probes were used to establish the number of erythroid cells in each preparation. The slides were analyzed by confocal- or standard fluorescence microscopy.

When lacZ was used as probe, we observed a single nuclear signal in a significant number (around 15%) of cells in E10.5 blood (Table 3; Fig. 5A, arrow). The much higher percentage of X-gal staining cells (Fig. 3F) suggests that the lacZ gene may not be transcribed continuously, as has been observed in transgenic mice carrying the human β-globin locus with small deletions in the LCR (Milot et al., 1996). Hence, we conclude that the hemizygous lacZ gene is actively transcribed at this stage. Contrasting observations were made with E13.5 liver cells where only a few cells (0.5-1%) had a nuclear signal (Table 3 and Fig. 5B). The control hybridization for expression of the homozygous mouse α-globin genes detected two nuclear signals in >90% of the cells (Table 3; Fig. 5C, arrowheads), confirming that these cells are of the erythroid lineage. No nuclear signal was detected in RNase-treated or lacZ-negative cells at either stage when hybridized with the lacZ probe (Fig. 5D and data not shown). Thus, we conclude that at E10.5, many of the erythroid cells from the μLCR and 5′HS3 lines are actively transcribing the lacZ gene. As development advances, the percentage of actively transcribing cells becomes considerably lower. This is in agreement.. with our DNaseI fade-out analysis, in which <1% of cells with an active gene in E13.5 liver would be too low to detect hypersensitive sites at the transgene.

The absence of hypersensitive sites in the μLCR-lacZ and 5′HS3-lacZ trans-genes raised the question whether these constructs are intrinsically inactive in the adult erythroid environment. To investigate this, we transfected the μLCR-lacZ construct into murine ery-throleukemia (MEL) cells. These cells represent the pro-erythroblast stage of adult cell differentiation. Stable clones were obtained by selection with G418, area of the chromatin. The cell popula-tions were expanded and induced with DMSO to undergo erythroid differen-tiation for 2 days. These cells were then analyzed for DNaseI sensitivity, as described earlier (Philipsen et al., 1990).The blots were hybridized with a 5′HS3 probe, because the lacZ probe hybridizes to vector sequences preclud-ing a straightforward interpretation of the results (not shown). Interestingly, we find that all four DNaseI hypersensi-tive sites are strongly represented, as shown in Fig. 6. The sub-band corre-sponding to the hypersensitive site at 5′HS4 runs just below the parent restric-tion band and is not well resolved. However, the hypersensitive sites at 5′HS1 and the promoter (which comigrate), 5′HS2 and 5′HS3 are found. Thus, we conclude that the μLCR-lacZ construct is not intrinsically inactive in the adult erythroid environment.

Previous studies involving the human β-globin LCR and its individual hypersensitive sites 5′HS1-4 have shown that these regulatory elements confer position-independent, copy-number dependent expression to the β-globin gene in trans-genic mice (Caterina et al., 1991, 1994; Ellis et al., 1993; Fraser et al., 1990; Philipsen et al., 1990, 1993; Pruzina et al., 1991, 1994; Ryan et al., 1989; Talbot et al., 1990). It has not been established whether the LCR can direct erythroid-specific expression in the absence of any globin gene sequences. We linked the μLCR or individual HS elements to a simple 80 bp promoter derived from the mouse Hsp68 gene that does not have any developmental or tissue specificity (Kothary et al., 1987, 1988, 1989).

We analyzed the ability of the μLCR and 5′HS1, 5′HS2,5′HS3 and 5′HS4 to activate this reporter construct in the erythroid cells of transgenic mice. The results show that only the complete μLCR and 5′HS3 can direct expression of the lacZ reporter in the erythroid system of the developing mouse. While erythroid-specific staining was observed in some of the E13.5 founder fetuses analyzed for 5′HS1 and 5′HS4, none of the established transgenic lines with these constructs showed such a staining at any stage of development. This indicates that 5′HS1 and 5′HS4 need other elements to activate erythroidspecific transcription.

5′HS2 is believed to be a crucial component of the LCR.This part of the LCR shows the highest degree of evolution-ary conservation (Hardison et al., 1993; Jimenez et al., 1992a). Its function has been studied in great detail in transfection (Hardison et al., 1993; Moon and Ley, 1991; Tuan et al., 1989), transgenic (Caterina et al., 1991, 1994; Ellis et al., 1993, 1996; Ryan et al., 1989; Talbot et al., 1990) and knock-out experiments (Fiering et al., 1995). To our surprise, we could not detect any erythroid-specific expression with the construct containing 5′ HS2, neither in E13.5 day founders, not at any stage of development in four established lines. Thus we conclude that 5′ HS2 is also unable to activate the non-erythroid Hsp68 promoter in the erythroid lineage. The fre-quently observed ectopic expression (i.e. expression outside the erythroid cells) in 5′ HS2 transgenics shows that this construct is not intrinsically inactive. Rather, these results suggest that 5′ HS2 is dependent on local elements provided by theβ-globin gene for erythroid-specific activation. Prelim-inary data show that these elements might reside within the first 800 bp of the β-globin promoter, since we have one E13.5 founder with 5′HS2 and the 5′-800 β-globin promoter driving the lacZ gene that shows extensive staining of the erythroid system (not shown).

It should be noted that all the transgenes were present as multi-copy concatamers. Previous studies from our laboratory have shown that only 5′HS3 of the LCR is capable of activat-ing the β-globin gene in single-copy transgenic mice. 5′HS2 was only active when present in a multi-copy configuration (Ellis et al., 1996). In contrast, it appears from the present study that 5′HS2 is not able to activate the minimal Hsp68 promoter, even in multi-copy transgenics. Our data corroborate those of Ellis et al. (1996) that 5′HS3 is the most important hypersen-sitive site of the β-globin LCR.

In the lines with the μLCR or 5′HS3, expression of the lacZ transgene could first be detected at E8-8.5 as a ring of cells cells were stably transfected with the μLCR-lacZ construct, and hypersensitive site analysis was performed as described (Philipsen et al., 1990). The DNA was digested with EcoRI and SacII, Southern blotted and around the base of the yolk sac. This is the first site of ery-thropoiesis in the developing mammalian embryo (Russell and Bernstein, 1966). Later, as vascularization occurs, lacZ-positive cells are found in the developing circulatory system of the embryo. In E10.5-11 blood, the vast majority of the circulating primitive erythrocytes are lacZ-positive. A paradox occurs in E13.5 liver. Although the erythroid fetal liver can be easily stained for the presence of the lacZ protein, large numbers of erythrocytes begin to appear in the circulation that do not with X-gal. We believe that these data can be explained by the model for the development of the hematopoietic system proposed by Dzierzak and Medvinsky (1995). According to this model, the fetal liver is first colonized by a wave of primitive hematopoietic precursor cells derived from the yolk sac. A second wave of colonization occurs when definitive stem cells originating from the AGM (Aorta, Gonads, Mesonephros) region migrate to the fetal liver. In our transgenic mice, the lacZ-positive cells in the fetal liver would be derived from the yolk sac precursor cells. These cells actively expressed the transgene at the embryonic stages of development. Although they have switched off the lacZ gene in the E13.5 fetal liver, as we have shown by in situ hybridization, they still contain lacZ protein and mRNA. Hence, these cells may stain lacZ-positive even after a number of cell divisions. The lacZ-negative cells in the circulation would be derived from fetal liver cells origi-nating from the AGM region, which have never expressed the lacZ gene at detectable levels. This model implies that the lacZ gene has a repressive effect on the LCR in definitive but not in primitive erythropoiesis. Recent observations of Guy et al. (1996) and Robertson et al. (1995) are consistent with this hypothesis, and preclude the simple explanation that our obser-vations are an artefact of the particular transgene construct that we have used. These authors found that the presence of the lacZ gene, even in the presence of a complete β-globin gene (Guy et al., 1996), interferes with LCR function (Robertson et al., 1995). Unfortunately, these groups did not present data on early embryonic time points. The absence of DNaseI hypersensitive sites in E13.5 fetal liver suggests that the repression is due to heterochromatization of the transgenes (Adams and Workman, 1993; Elgin, 1990; Hebbes et al., 1992; Patterton and Wolffe, 1996), and not caused by the inability of the LCR to interact with the model promoter we employed. Recent data of knock-out experiments in the mouse β-globin locus show that non-erythroid promoters can be activated very efficiently by the LCR (Fiering et al., 1995; Hug et al., 1996; Kim et al., 1992; Nandi et al., 1988; Shehee et al., 1993). Consistent with these observations, we can detect hypersensitive sites at the transgene when transfected into MEL cells. Mechanistically, the lacZ gene could serve as a nucleation point for heterochromatization in definitive erythroid cells, resulting in repression of the trans-genes in mice. In tissue culture, these events are eliminated because the selection procedure only leaves cells in which inte-gration has occurred in open areas of chromatin, to allow expression of the resistance marker. The results suggest that the μLCR-lacZ construct is active when integrated in such chromatin in adult erythroid cells, but that it is suppressed when it integrates in closed chromatin domains, as is probably the case in transgenic mice. The reason for the inactivation in mice is presently unknown, but we have ruled out the possibility that it is caused by progressive methylation of the transgenes (results not shown).

Previous studies have shown that the μLCR and 5′HS3 direct high level expression of the β-globin gene in transgenic mice (Ellis et al., 1996; Philipsen et al., 1990, 1993; Talbot et al., 1989); the data in this paper indicate that the presence of this gene is a requirement for the activity of the μLCR and 5′HS3 in fetal and adult erythroid cells.

Our data are in apparent contradiction with a number of published observations regarding hypersensitive site formation in the LCR. Two independent laboratories have studied this for 5′HS2 and 5′HS4. In the case of 5′HS2, Pawlik et al. (1995) have shown that 5′HS2, in the absence of a gene, has a DNaseI-hypersensitive site in transgenic mice. We also observe weak hypersensitivity at 5′HS2 in our transgenic lines, and suggest that this corresponds to the ubiquitous hypersensitive site asso-ciated with the human 5′HS2 (Ellis et al., 1996). This hyper-sensitive site is located outside the core fragment of 5′HS2, where the erythroid-specific activity resides.

Transgenic mice experiments also suggest that 5′HS4 can autonomously form a hypersensitive site (Lowrey et al., 1992). However, since only one transgenic line was analyzed in this study, the transgene may have integrated in an area of open chromatin. Stamatoyannopoulos et al. (1995) have reported on the formation of 5′HS4 in the presence of 5′HS3 in transfec-tion experiments in MEL cells. As we show here, we can detect hypersensitive sites when we transfect the μLCR-lacZ construct into MEL cells, but not in the transgenic lines.

Our data are compatible with those obtained for the chicken β-globin locus in transgenic mice. These data suggest that hypersensitive site formation in the chicken 3′ enhancer is dependent on the presence of an active promoter (Reitman et al., 1993). We have shown that in E13.5 fetal liver the μLCR and 5′HS3 transgenes are in a closed chromatin conformation, and that transcription of the lacZ gene is not detectable in these cells. Hence, as in the chicken case (Reitman et al., 1993), a closed promoter correlates with the absence of hypersensitive sites in the LCR.

The data in this paper show that the μLCR and 5′HS3 can activate the minimal Hsp68 promoter in transgenic mice. The other HS elements of the LCR are inactive and therefore require additional elements of the β-globin locus for their activity. The expression of the μLCR and 5′HS3 transgenes is specific for embryonic, primitive erythropoiesis. Hence, these mice are useful tools for the study of these first erythroid cells. Recently, gene-targeting experiments have implicated a number of transcription factors in the development of the erythroid cell (reviewed in Shivdasani and Orkin, 1996). The role of these transcription factors in the activating function of the μLCR and 5′HS3 in primitive erythropoiesis can now be addressed directly by breeding our transgenes into these knock-out backgrounds.

We would like to thank Drs Elaine Dzierzak and Dies Meijer for their thoughtful comments on the manuscript, Ms An Langeveld for help with the in situ hybridizations, Mirko Kuit for his expert help with photography, and Adriaan Houtsmuller for confocal microscopy. This work was supported by an NWO grant to F.G.