Erythroid Krüppel‐like factor (EKLF) is active in primitive and definitive erythroid cells and is required for the function of 5′HS3 of the β‐globin locus control region

A β‐globin gene linked directly to the LCR is expressed prematurely in embryonic erythroid cells (Hanscombe et al., 1991; Dillon et al., 1997). The EKLF protein is known to be present in these cells (Southwood et al., 1996) but no function has been assigned to EKLF in primitive erythropoiesis (Nuez et al., 1995; Perkins et al., 1995); a Northern blot showing the absence of EKLF‐encoding mRNA in E10.5 −/− yolk sac RNA is shown in Figure 1A. Thus, we wondered whether embryonic expression of the β‐globin gene is also dependent on EKLF. To test this directly, we used a transgenic line with a single copy of the μLCR‐β‐globin construct (Figure 1B). In E13.5 fetal liver, this human β‐globin transgene is expressed at the same level as the endogenous β‐major gene (line μD14 in Ellis et al., 1996). First, we determined transgene expression in primitive cells. We isolated RNA from E10.5 embryos and quantitated the expression level of the human β‐globin gene relative to that of the endogenous mouse α‐globin gene, which is expressed in both primitive and definitive cells and is not affected by the EKLF knock‐out. The results are shown in Figure 1C. As could be predicted, the β‐globin gene is expressed prematurely due to the absence of the ϵ‐ and γ‐globin genes in the construct (Hanscombe et al., 1991; Dillon et al., 1997). We then analysed expression of the transgene in EKLF^−/− embryos. Interestingly, there is only a modest (2‐ to 3‐fold), reduction in the level of transgene‐derived mRNA in E10.5 EKLF^−/− embryos, while expression of the transgene is more drastically affected in E13.5 EKLF^−/− fetal liver (>6–fold reduction). This reduction is smaller than reported previously for adult β‐globin expression (Perkins et al., 1996; Wijgerde et al., 1996) which could be the result of the presence of human β‐globin mRNA in circulating primitive cells. We therefore performed primary transcript in situ hybridization to determine the percentage of cells actively transcribing the transgene (Wijgerde et al., 1995). Probes detecting transcription of the mouse α‐globin genes were used as a control to establish the number of erythroid cells in each preparation. Figure 1D shows that the β‐globin transgene is actively transcribed in >95% of the cells in EKLF^+/+ E10.5 blood (Table I). In the absence of EKLF, active transcription of the human β‐globin gene is observed in ∼55% of the cells, in good agreement with the S1 nuclease analysis. Interestingly, human β‐globin gene transcription is detected in <1% of the cells in EKLF^−/− fetal liver, while 95% of E13.5 EKLF^+/+ fetal liver cells actively transcribe the transgene (Figure 1D and Table I). The extremely low number of cells expressing the transgene in EKLF^−/− mice closely resembles previous observations on adult β‐globin gene transcription (Wijgerde et al., 1996). Thus, the relatively high level of human β‐globin mRNA in E13.5 fetal liver of EKLF^−/− mice observed by S1 nuclease analysis is due to earlier expression of the transgene in primitive erythroid cells still present in the circulation at E13.5.

We conclude that EKLF has only a modest effect on the expression of the human β‐globin transgene in primitive cells, implying that expression of the β‐globin gene is not intrinsically dependent on EKLF in this lineage. However, EKLF is essential for β‐globin expression in definitive cells, regardless of the presence of embryonic and fetal globin genes.

The above results gave the first indication that EKLF could function as a transcriptional activator in the primitive lineage, since we observed a reduced expression of the β‐globin transgene in EKLF^−/− primitive cells. To further study the potential role of EKLF in LCR‐mediated gene activation in embryonic erythroid cells, we used lacZ reporter mice in which μLCR constructs drive expression of the lacZ gene from a basic, non‐erythroid promoter (Tewari et al., 1996). Three different constructs were used (Figure 2A). The first contained the μLCR, as used in the μLCR–β‐globin construct described above. The second construct contained the 1.9 kb 5′HS3 fragment of the μLCR, and the third construct contained only the 225 bp core fragment of 5′HS3 with six potential EKLF binding sites (Philipsen et al., 1993; Philipsen et al., 1990). LacZ expression of the μLCR and 5′HS3 transgenes is confined to primitive cells, while the core 5′HS3 transgene is also expressed in definitive erythroid cells (not shown; Tewari et al., 1996).

The three reporter lines were bred into the EKLF^−/− background, and expression of the lacZ reporter gene in primitive cells was assayed by X‐gal staining of E10.5 embryos (Bonnerot and Nicolas, 1993). At E10.5, erythroid expression of the μLCR‐lacZ transgene is easily detectable in EKLF^−/− mice (Figure 2B). X‐gal staining of erythroid cells is greatly reduced in the (multi‐copy) 5′HS3‐lacZ line (not shown) and undetectable in the (single copy) core 5′HS3‐lacZ line (Figure 2C); this is particularly evident when the yolk sacs are stained (compare with Figure 2D and E). Note that the position effect, i.e. lacZ expression outside the erythroid system (Tewari et al., 1996), is unaffected in the core 5′HS3‐lacZ/EKLF^−/− embryos (Figure 2C). To demonstrate that the observed effects on lacZ activity are due to reduced transcription of the transgenes, we performed in situ hybridizations to determine the percentage of cells actively transcribing the lacZ reporter gene (Wijgerde et al., 1995; Tewari et al., 1996). These experiments are summarized in Table II and show that there is a small reduction (from 14 to 10%) in the number of cells with lacZ transcription spots in μLCR‐lacZ/EKLF^−/− transgenics. In 5′HS3‐lacZ/EKLF^−/− and core 5′HS3/EKLF^−/− mice, there is a >10‐fold reduction in the number of cells with lacZ transcription spots; a >10‐fold reduction was also found in E13.5 fetal liver cells of core 5′HS3/EKLF^−/− mice, as expected. Together, these data imply that EKLF is not essential for transcriptional activation by the μLCR, in agreement with observations on the μLCR‐β‐globin transgene (this paper) and the complete human β‐globin locus (Perkins et al., 1996; Wijgerde et al., 1996). However, the activity of 5′HS3 and the core of 5′HS3 is dependent on the presence of EKLF. Furthermore, our data show unambiguously that EKLF is a functional transcriptional activator in primitive cells. This activity of EKLF has previously gone unnoticed (Nuez et al., 1995; Perkins et al., 1995; Wijgerde et al., 1996).

It has been proposed that globin switching is mediated through a gradual change in the transcription factor environment (Grosveld et al., 1993), and previous experiments carried out with transgenic mice harbouring the complete human β‐globin locus showed that the switch from γ‐ to β–globin expression, which occurs in the fetal liver between E12 and E14, is delayed in EKLF^+/− mice (Wijgerde et al., 1996). This suggests that EKLF is present at critical levels in switching cells, thus we asked if increased EKLF levels would also influence switching. To investigate this, we cloned a cDNA encoding the mouse EKLF protein in the expression vector pEV3 (Needham et al., 1992). This vector contains the μLCR, the first 400 bp of the β‐globin promoter, a modified β‐globin gene and the 3′ flanking region of the β‐globin gene (Figure 3A). Two transgenic mouse lines (EKLF‐pEV) were generated with this construct. Western blot analysis of E13.5 fetal liver protein shows modest overexpression (2‐ to 3‐fold) of EKLF in one of these lines (Figure 3B). This line was crossed with mice containing the human β‐globin locus (Strouboulis et al., 1992) to analyse the expression of γ‐ and β‐globin mRNA in E12.5 and E13.5 fetal liver. The results of the S1 nuclease analysis are shown in Figure 3C. The level of endogenous mouse α‐globin mRNA was used to control for the amount of RNA used in each sample. Quantitation of these data reveals that the ratio of γ/β‐globin expression decreases earlier in mice carrying the EKLF‐pEV transgene, i.e. with a higher EKLF level. Thus, we conclude that the rate of γ‐ to β‐globin gene switching correlates with the level of EKLF (Wijgerde et al., 1996).

The EKLF‐pEV mice should express transgene‐derived EKLF in primitive and definitive cells (Hanscombe et al., 1991). Thus, this transgene might rescue the lethality of the EKLF null mutation. To test this, we bred the EKLF‐pEV transgene into the EKLF^−/− background. In the F2 generation, we obtained mice with an EKLF‐pEV/EKLF^−/− genotype with both transgenic lines at the expected Mendelian ratio (data not shown). As could be expected, the EKLF‐pEV transgene also rescues expression in the primitive cells of the 5′HS3‐lacZ mice discussed above (data not shown).

Haematological analysis of adult mice shows that EKLF‐pEV/EKLF^−/− mice have lower red blood cell counts, haemoglobin content and haematocrit in comparison with wild‐type mice, while the mean cell volume of the erythrocytes is increased (Table III). Thus, erythropoiesis is still slightly impeded in the rescue mice, which might be caused by inadequate expression of transgene‐derived EKLF. In addition, we observe that platelet counts are lower in EKLF‐pEV/EKLF^−/−, and even further reduced in EKLF‐pEV/EKLF^+/+ mice. Finally, in agreement with the analysis of fetal erythropoiesis (Nuez et al., 1995; Perkins et al., 1995), the EKLF^+/− mice have haematological values that are close to those found for EKLF^+/+ mice, although the slightly reduced haemoglobin content might indicate a very mild phenotype.

In conclusion, we show that the expression of transgene‐derived EKLF rescues the lethal phenotype of the EKLF null mutation. This formally proves that the phenotype of the EKLF^−/− fetuses was indeed the result of the mutation in the EKLF gene, and not some other defect in the ES cells or an unknown other gene that might be present at the EKLF locus.

EKLF is essential for high level expression of the adult β‐globin genes, and EKLF^−/− mice die of anaemia at E14–15 due to defective erythropoiesis caused by severely reduced β‐globin levels (Nuez et al., 1995; Perkins et al., 1995). Transgenic mice with just a β‐globin gene driven by the μLCR express the transgene prematurely in primitive cells (Hanscombe et al., 1991); we have used a single copy transgenic line with such a construct (Ellis et al., 1996) to assess the potential role of EKLF in embryonic β‐globin expression. Surprisingly, there is only a modest reduction in human β‐globin mRNA levels in E10.5 EKLF^−/− embryos. This reduction is reflected in the number of cells actively transcribing the transgene, as analysed by primary transcript in situ hybridization (Wijgerde et al., 1995). In EKLF^+/+ cells, transcription foci of the transgene are detected in the nuclei of >95% of the erythroid cells, and in EKLF^−/− cells in ∼55% of the nuclei. This indicates that the β‐globin gene is transcribed only part of the time in −/− primitive cells, which could be attributed to a less stable interaction between the LCR and the β‐globin promoter (Milot et al., 1996). However, the gene is transcriptionally competent and still transcribed at high levels. This is in sharp contrast with the observations in definitive cells of E13.5 EKLF^−/− fetal liver. Primary transcript in situ hybridization shows a dramatic reduction in the number of cells actively transcribing the β‐globin gene. In agreement with data on the expression of the β‐globin gene in the intact globin locus (Wijgerde et al., 1996), fewer than 1% of the cells contain a nuclear transcription focus of the transgene. In control littermates, a transcription signal is detectable in >95% of the cells. Thus, expression of the β‐globin gene is strictly dependent on the presence of EKLF in definitive cells. However, the expression in primitive cells shows that the β‐globin gene is not intrinsically dependent on EKLF. Rather, this dependency is acquired in definitive cells through an as yet unidentified mechanism. A more restrictive chromatin environment in definitive cells, as has been proposed to explain the embryonic expression pattern of many μLCR‐lacZ transgenes, could be such a mechanism (Tewari et al., 1996). Consequently, the autonomous silencing of the ϵ‐ and γ‐globin genes (Grosveld et al., 1993) might at least in part be an intrinsic property of these genes mediated by repressive chromatin interactions.

The promoter of the human β‐globin gene contains a canonical binding site for EKLF (Feng et al., 1994). Thus, it is likely that direct binding of EKLF to this element is mandatory for high level expression of the gene in definitive cells (Perkins et al., 1996; Wijgerde et al., 1996). In addition, EKLF binding sites are found in the hypersensitive sites of the LCR (Grosveld et al., 1993). The best example is provided by 5′HS3, the only element of the LCR that can reproducibly direct expression of single copy globin transgenes in mice and, therefore, has been postulated to provide a dominant chromatin opening activity (Ellis et al., 1996). The 225 bp core fragment of 5′HS3 contains six potential EKLF binding sites (Philipsen et al., 1990). Intactness of these binding sites is essential for the activity of 5′HS3 (Philipsen et al., 1993). Due to the overlapping binding specificity of proteins present in erythroid nuclei, most notably those belonging to the Sp1 family of transcription factors, it is not known which proteins bind to these sequences in vivo (Philipsen et al., 1993). To separate the LCR binding sites from those of the promoter, we used mice with a lacZ reporter gene and a non‐erythroid minimal promoter linked to LCR derivatives (Tewari et al., 1996) and analysed reporter gene expression in EKLF null mutants. Consistent with the analyses of the β‐globin loci (Perkins et al., 1996; Wijgerde et al., 1996), the μLCR‐lacZ construct is still active in EKLF^−/− embryos. In contrast, expression of the 5′HS3‐lacZ and core 5′HS3‐lacZ transgenes is greatly reduced in E10.5 blood, showing that EKLF is required for transcriptional activation by 5′HS3. These results suggest that EKLF binds directly to the core fragment of 5′HS3. Our observations are consistent with the reduced hypersensitivity of 5′HS3 in EKLF^−/− fetal livers (Wijgerde et al., 1996) and the phenotype of the 5′HS3 deletion in mice (Hug et al., 1996). This deletion has a very small effect on the expression of the linked embryonic ϵy and βh1 genes, while adult β‐globin expression is reduced by 30%.

The results with the lacZ transgenics show that EKLF is a functional transcriptional activator in primitive cells, in addition to its previously described activity in definitive cells (Nuez et al., 1995; Perkins et al., 1995; Wijgerde et al., 1996). Thus, EKLF may activate specific genes in the primitive lineage, but reduced expression of these genes in EKLF^−/− embryos would not result in an obvious phenotype. Experiments aimed at the identification of such EKLF target genes in the primitive and definitive lineage are in progress.

We have assessed the effect of increased EKLF levels on γ‐ to β‐globin switching in transgenic mice with the complete human β‐globin locus (Strouboulis et al., 1992). Globin switching is a gradual process in which the cells change from γ‐ to β expression over a period of ∼2 days (Wijgerde et al., 1995). It has been proposed that switching is effectuated by subtle changes in transcription factor environment (Grosveld et al., 1993). In accord with this notion, our data show that higher EKLF levels result in an earlier switch from γ‐ to β‐globin expression in the fetal liver. Thus, the timing of switching is influenced by EKLF levels, and higher than normal EKLF levels expedite the switch. These data are compatible with the prevailing model that the LCR acts as a holocomplex activating only one gene at a time (Wijgerde et al., 1995). EKLF might increase the stability of interactions between the LCR and the β‐globin promoter. Hence, we propose that higher EKLF levels will reduce the time the LCR spends activating the γ‐globin genes, thus resulting in an earlier switch (Wijgerde et al., 1996). After switching has completed, increased EKLF levels no longer affect β‐globin gene expression, presumably because EKLF is present at saturating levels in adult erythroid cells. This is consistent with the observation that the mouse and human β‐globin genes are expressed at wild‐type levels in EKLF^+/− mice (Wijgerde et al., 1996).

We show that the lethal phenotype of EKLF^−/− mice can be rescued by expressing EKLF under the control of the β‐globin LCR. The haematological analysis shows that, compared with wild‐type mice, adult EKLF rescue mice have lower red blood cell numbers, haemoglobin concentration and haematocrit, and a slightly increased mean cell volume. Thus, a subtle phenotype remains in these rescue mice which indicates that appropriate developmental expression of EKLF is important for normal erythropoiesis. This notion is emphasized by the observation that pEV‐EKLF/EKLF^+/+ mice have significantly reduced numbers of platelets, suggesting that EKLF may affect the balance between the megakaryocytic and erythroid lineages. Hence, there might be an unexpected link between EKLF and megakaryopoiesis, as has recently been described for another erythroid‐specific transcription factor, GATA‐1 (Shivdasani et al., 1997).

The rescue experiment sets a platform for the analysis of EKLF protein domains required for high level β‐globin expression. EKLF transactivation domains have been characterized in cell culture assays, and a cellular factor interacting with EKLF has been implicated in EKLF‐mediated transcativation (Chen and Bieker, 1996). These observations can now be stringently tested through the rescue of EKLF^−/− mice. Furthermore, co‐immunoprecipitation of EKLF and its interaction partners from primary erythroid cells could be attempted after the introduction of an epitope tag into the EKLF transgene.

The EKLF‐pEV construct was generated by inserting the EKLF cDNA into the BglII site of the pEV3 expression vector (Needham et al., 1992). The microinjection fragment was released by digestion with AatII and Asp718, gel purified and used for transgenesis as described (Kollias et al., 1986). The 225 bp core fragment of 5′HS3 (Philipsen et al., 1990) was cloned in a lacZ reporter vector (Kothary et al., 1988) and digested with HindIII and Asp718 to obtain the microinjection fragment. The other transgenic lines used in this study have been described before: μLCR‐β‐globin line μD 14 (Ellis et al., 1996); μLCR‐lacZ line b and 5′HS3‐lacZ line b (Tewari et al., 1996); 70 kb human β‐globin locus line 72 (Strouboulis et al., 1992) and EKLF knock‐out mice (Nuez et al., 1995).

Genomic mouse DNA was prepared from part of the body of the fetuses or tail clips from adult mice. The mice were genotyped by Southern blot analysis (Strouboulis et al., 1992; Nuez et al., 1995; Ellis et al., 1996; Tewari et al., 1996). RNA was isolated from erythroid tissues and globin gene expression was quantitated by S1 nuclease analysis as described (Strouboulis et al., 1992; Wijgerde et al., 1996). Northern blotting was done as described (Talbot et al., 1989). Probes used were the complete EKLF cDNA and a cDNA detecting the erythroid‐specific form of aminolevulinate synthase (ALA‐S), a kind gift of Dr Peter Curtis (Boston).

Protein was isolated from E13.5 liver, and 10 μg was used for Western blot analysis with a rabbit anti‐EKLF polyclonal antibody (Southwood et al., 1996), a kind gift of Dr J.J.Bieker (New York). A mouse monoclonal antibody recognizing the 62 kDa subunit of THII‐H was used as a loading control; this antibody was generously provided by Drs B.Winkler and G.Weeda (Rotterdam).

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 and 0.5% glutaraldehyde in PBS for 30 min. They were then rinsed in embryo buffer (PBS, 0.02% NP40, 0.01% deoxycholate, 2 mM MgCl₂) and incubated in staining solution [1 mg/ml X‐gal, 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆ and 1 mM EGTA in embryo buffer]. After 3–4 h 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.

E10.5 embryonic blood from the heart or E13.5 liver cells from individual fetuses were disrupted in 50 and 150 μl of PBS respectively. Twenty μl of this cell suspension was fixed onto a poly‐lysine coated slide in 4% formaldehyde, 5% acetic acid for 20 min at room temperature. The cells were then washed three times for 10 min in PBS and stored in 70% ethanol at −20°C.

Three probes were utilized for in situ hybridization. A nick‐translated biotinylated lacZ probe was used for the detection of lacZ transcription (Tewari et al., 1996). A set of four biotinylated oligonucleotides was used to detect human β‐globin transcription (Dillon et al., 1997). DIG‐labelled mouse α‐globin intron‐specific oligonucleotides were used to reveal the erythroid cells (Tewari et al., 1996). Hybridizations were done as described (Wijgerde et al., 1995; Tewari et al., 1996). The lacZ hybridization mix contained 50% formamide and 10% dextran sulfate. For antibody detection, the Tyramid amplification system (TSA) of Dupont was used (Raap et al., 1995). For quantitation, ∼300 cells were scored for each slide. The reproducibility of these data fell within a 10% error margin.

Adult mice (3–5 months of age) were bled by orbital bleeding, and blood samples were analysed on an F800 microcell counter. Statistical analysis was done by unpaired two‐sided t‐testing.