Context-dependent EKLF responsiveness defines the developmental specificity of the human epsilon-globin gene in erythroid cells of YAC transgenic mice

Abstract

We explored the mechanism of definitive-stage epsilon-globin transcriptional inactivity within a human beta-globin YAC expressed in transgenic mice. We focused on the globin CAC and CAAT promoter motifs, as previous laboratory and clinical studies indicated a pivotal role for these elements in globin gene activation. A high-affinity CAC-binding site for the erythroid krüppel-like factor (EKLF) was placed in the epsilon-globin promoter at a position corresponding to that in the adult beta-globin promoter, thereby simultaneously ablating a direct repeat (DR) element. This mutation led to EKLF-independent epsilon-globin transcription during definitive erythropoiesis. A second 4-bp substitution in the epsilon-globin CAAT sequence, which simultaneously disrupts a second DR element, further enhanced ectopic definitive erythroid activation of epsilon-globin transcription, which surprisingly became EKLF dependent. We finally examined factors in nuclear extracts prepared from embryonic or adult erythroid cells that bound these elements in vitro, and we identified a novel DR-binding protein (DRED) whose properties are consistent with those expected for a definitive-stage epsilon-globin repressor. We conclude that the suppression of epsilon-globin transcription during definitive erythropoiesis is mediated by the binding of a repressor that prevents EKLF from activating the epsilon-globin gene.

Figure 1

Evaluation of EKLF contribution to the activity of human globin genes in YAC transgenic mice.(A) All of the lines examined are single copy transgenes; lines 31-wild (C,D) and 264-wild (E) bear wild-type (top) and line 31-inv. (B) Bears inverted (bottom) human β-globin YAC transgenes (Tanimoto et al. 1999a). (B–E) Total RNA was prepared from embryos derived from the intercross of male (EKLF^+/−::YAC Tg^+/−) and female (EKLF^+/−) animals. The EKLF genotype of each embryo was determined by allele-specific PCR analysis. Samples of the various EKLF genotypes were analyzed individually at least three times. The expression levels of the β-like globin genes, normalized to that of the endogenous mouse α-globin gene, were internally compared and the relative level of expression (with expression of EKLF^+/+ mice set at 100) was statistically analyzed. Data were collected from more than one litter. The average and standard deviation (S.D.) is graphically depicted for the genotypic group of more than two animals and the average is shown for two samples (at least two samples were analyzed for each group). Bars 1 and 4 in each histogram represent the expression level of the β-like globin genes in EKLF^+/+, 2 and 5 in EKLF^+/−, and 3 and 6 in EKLF^−/− mice. Representative results of RT–PCR for human ɛ (hɛ), γ (hγ), β (hβ), and mouse α (mα) in different EKLF backgrounds (+/+, +/−, or −/−) are shown below each panel. (B) β- (open) and γ (shaded)-globin gene expression in the embryonic yolk sac; (C) β- and γ-globin gene expression in the 14.5-dpc fetal liver; (D,E) ɛ- (black) and γ (shaded)-globin gene expression in the embryonic yolk sac.

Figure 2

Comparison of the proximal promoter regions of β-type globin genes. (A) Positions of distal (open box) or proximal (solid box) CAC boxes are shown relative to the CAAT box in human ɛ, γ, and β, and murine ɛy, b1, and b2 genes. For the human γ gene, the ^Aγ distal CAAT box and its upstream region was used for alignment. (B) Sequence alignment of human embryonic ɛ- and adult β-globin gene promoter regions as well as the mutant ɛ-globin promoters examined in this study (E + CAC and Bepsi). Mutated nucleotides in the two promoters are underlined. Positions corresponding to the distal CAC, proximal CAC, and CAAT boxes are bracketed. Putative binding sites for COUP–TFII in K562 and DRED in MEL cells are underlined with arrows.

Figure 3

Structural analysis of human β-globin YACs in transgenic mice. (A) End-fragment analysis of β-globin YAC transgenic mice. Genomic DNA from transgenic thymus was digested with PstI and blotted to nylon membranes. A 723-bp PstI–AlwNI fragment (L-end, derived from the left YAC vector arm), and a 639-bp PvuII–AvaI fragment (R-end, from the right YAC arm) were used as probes. These fragments are located outside of the PstI restriction enzyme sites within the YAC vector arms (C), thus enabling analysis of the ends of the transgene at the genomic integration site. Using this strategy, it is possible to determine the arrangement of multiply integrated copies of the transgene (e.g., whether they integrate in a head-to-head; head-to-tail, or tail-to-tail configuration; Liu et al. 1998). Head-to-tail junction fragments are indicated by open triangles (in line 418). DNA size markers are shown at right (in kb). The identity of the transgenic lines are indicated by the number on top. (B) Structural organization of the human β-globin YACs in transgenic mice. Each of the four lines, 402, 585, 588, and 590 (bottom), yield end fragments with different sizes (A) for the L- and R-end probes, indicating that each of these lines harbor single copies of the transgene. Line 418 (middle) shows two bands with both L- and R-end probes and one of these is the size of a H-T junction fragment (open triangle in A). The intensity of this junction band is stronger than that of others, indicating that this line carries more than two copies of the transgene. Further internal comparison, using a fragment of the GATA-2 locus as a probe, confirmed that this line has three intact copies of the transgene (data not shown). Line 408 (top) has three fragments for both probes and none of them hybridized to both L and R probes. In subsequent generations, all of these fragments segregated together, indicating that three copies of the transgene are integrated close to one another but not directly linked. The results were confirmed by internal fragment assay using the GATA-2 probe (data not shown). (C) Schematic representation of the human β-globin YAC (A201F4.3) indicating the positions of SfiI and PstI restriction enzyme sites. The whole β-globin locus is contained within two SfiI restriction enzyme fragments (10 kb and 100 kb, as indicated). The positions of the probes used for Southern blot analysis are depicted by the solid boxes. Mutations shown in Fig. 2B (5 and 9 nucleotides each for E + CAC and Bepsi, respectively) were introduced into the YAC by homologous recombination in yeast. (D) Integrity of the human β-globin YAC transgenes. Thymus cells from transgenic mice were embedded in agarose plugs and digested with SfiI at 50°C (Tanimoto et al. 1999b). The DNA was then separated by PFGE, blotted to nylon membranes, and hybridized separately to probes (indicated at left) from the β-globin locus or from the right YAC vector arm (see C). The sizes of the expected bands are shown at right.

Figure 4

Expression of the human β-type globin genes in the E + CAC mutant YAC. (A) ɛ- and γ-globin gene expression in the embryonic yolk sac. RNA was isolated from the yolk sacs of two (except for wild-type lines 42 and 31; Tanimoto et al. 1999a) independent litters for each line at 9.5 dpc. Expression of human ɛ (hɛ)- and human γ (hγ)/mouseα (mα)-globin genes was analyzed separately by semiquantitative RT–PCR because of the difficulty in comparing low levels of ɛ with the more abundant γ/α expression at the same cycle number (see Materials and Methods for more details). The signals for ɛ-globin at 18 cycles and γ/α-globin at 12 cycles were quantitated by PhosphorImager, and the ratios of ɛ/α (top) and γ/α (bottom) were calculated (the mouse α signal at 12 cycles was set at 100 and the values are normalized by transgene copy numbers) and illustrated as histograms. For each sample, the mean ± standard deviation (SD) from at least three independent experiments was statistically analyzed; the SD is shown for each. Representative data are shown beneath each panel. (B) ɛ (top)-, γ(middle)-, and β (bottom)-globin gene expression in the fetal liver. RNA was isolated from the fetal liver of 14.5-dpc embryos. Expression of ɛ-, γ-, and human β (hβ)/α-globin genes was analyzed separately by RT–PCR. The signal for ɛ- and γ-globin at 18 cycles and β/α-globin at 12 cycles was quantified and the ratio of ɛ/α, γ/α, and β/α was calculated (α signal at 12 cycles was set at 100). (C) ɛ (top)- and β (bottom)-globin gene expression in the anemic adult spleen. Two animals (∼4 wk old) representing each transgenic line were made anemic and RNA was isolated from the spleen. Expression of ɛ- and β/α-globin genes was analyzed separately by RT–PCR. No γ-globin expression was observed after 18 cycles of amplification. The signals for ɛ-globin at 18 cycles and β/α-globin at 12 cycles were quantified, and the ratios of ɛ/α and β/α were calculated (α signal at 12 cycles was set at 100).

Figure 5

Enhanced activities of the Bepsi, but not E + CAC, promoters require EKLF. Two lines of mutant animals from each construct (E + CAC and Bepsi) were used to analyze the expression of human β-like globin genes in EKLF-null background. Total RNA was prepared and analyzed as described in the legend to Fig. 1. The expression levels of the β-like globin genes in EKLF^+/+ background was set at 100. The average and standard deviation (SD) is shown graphically for the genotypic group of more than two animals and only an average is shown for that of two (at least two samples were analyzed for each group). Expression of β-like globin genes in EKLF^+/+ (bars 1 and 4); in EKLF^+/− (bars 2 and 5), and in EKLF^−/− (bars 3 and 6) mice is shown. Representative results of RT–PCR for human ɛ (hɛ), γ (hγ), β (hβ), and mouse α (mα) in different EKLF mutant backgrounds (+/+, +/−, or −/−) are shown below each panel. (A) Embryonic (9.5-dpc) expression of ɛ- and γ-globin genes in the E + CAC transgenic mice. (B) Fetal liver (14.5-dpc) expression of ɛ-, β-, and γ-globin genes in the E + CAC transgenic mice. (C) Embryonic (9.5-dpc) expression of ɛ- and γ-globin genes in the Bepsi transgenic mice. (D) Fetal liver (14.5-dpc) expression of ɛ-, β-, and γ-globin genes in the Bepsi transgenic mice.

Figure 6

Expression of the human β-globin genes in the Bepsi mutant YAC. Samples were analyzed in the same way as described in the legend to Fig. 4. (A) ɛ- and γ-globin gene expression in the embryonic yolk sac. Samples from wild-type (lines 31 and 264; Tanimoto et al. 1999a) and E + CAC (line 402) transgenic mice were analyzed at the same time and typical data are shown. All of the reference lines as well as three of the Bepsi lines carry single copy transgenes, hence the data can be directly compared without normalization. The signals for ɛ-globin at 18 cycles and γ/α-globin at 12 cycles were quantified, and the ratios of ɛ/α (top) and γ/α (bottom) were calculated (the mouse α signal at 12 cycles was set at 100) and illustrated as histograms. For each sample, the mean ± standard deviation (SD) from at least three independent experiments was statistically analyzed; the SD is shown for each. Representative data are shown beneath each panel. (B) ɛ-, γ-, and β-globin gene expression in the fetal liver. The signal for ɛ-globin and γ-globin at 18 cycles and β/α-globin at 12 cycles was quantified and the ratio of ɛ/α, γ/α, and β/α was calculated (α signal at 12 cycles was set at 100). (C) ɛ- and β-globin gene expression in the anemic adult spleen. The signals for ɛ-globin at 18 cycles and β/α-globin at 12 cycles were quantified, and then the ratios of ɛ/α and β/α were calculated (α signal at 12 cycles was set at 100).

Figure 7

Analysis of nuclear factor binding to the β-globin gene promoter. (A) Schematic representation of the proximal promoter region of ɛ, E + CAC, Bepsi, and β genes and their fragments used as competitors in the EMSA experiments (thick lines). Positions of distal and proximal CAC as well as CAAT boxes are shown as open or black boxes (see Materials and Methods for detailed sequence information). (B) EMSA using cell extracts from K562 cells and a probe containing the proximal CAC and CAAT boxes from the adult β-globin promoter (Beta in A). A 200-fold molar excess of unlabeled DNA fragments was used as competitor unless stated otherwise. The supershift appearing in the presence of Sp1 antibody is indicated by an arrowhead (lane 8). (C) The same probe was used in EMSA with MEL cell extracts. The antibody/Sp1 supershift is depicted by an arrowhead (lane 15). (D) EMSA with affinity-purified GST–EKLF. An 80-fold molar excess of unlabeled competitor was used in this experiment. The protein/DNA complexes in lanes 11 and 12 disappeared when a 400-fold molar excess of competitor was added (Epsi distal and Beta distal; data not shown). The antibody/EKLF supershift (JB) is indicated by an arrowhead (lane 13).

Figure 8

Analysis of nuclear factor binding to the ɛ-globin gene promoter. (A) Schematic representation of the promoter region of the ɛ, E + CAC, Bepsi, and β genes, and fragments used as competitors in the EMSA experiments (thick lines). Positions of proximal CAC (black box) as well as ɛ (in an open arrow) or β (black box) CAAT boxes are shown. Putative binding sites for both COUP–TFII and DRED are indicated by the open arrow (each arrow represents a direct repeat sequence). (B) EMSA with K562 cell extracts and a probe containing proximal sequences and CAAT box of the ɛ-globin promoter (Epsi in A). A 200-fold molar excess of unlabeled DNA was used as competitor unless otherwise stated. The shifted band generated after including an antibody recognizing COUP–TFII is indicated with an arrowhead (lane 14). (C) Comparison of proteins from MEL and K562 cells that interact with the Epsi fragment. A comparable amount of protein extract was used for analysis. A 200-fold molar excess of competitor DNA or antibodies were used to test binding specificity. The supershifted COUP–TFII/antibody complex is detectable only in K562 cells and indicated by an arrowhead. (D) An EMSA similar to that shown in B was performed with MEL cell extract.

Figure 9

Schematic model of β-globin gene switching. At the primitive stage, the direct repeat (DR) sequences (shown as green inverted arrows) on the ɛ-globin promoter are inactive as silencing elements, and therefore EKLF can interact with the distal CAC site and contribute to promoter activation (top left). The β-globin promoter is also potentially active at this stage and this activity is completely dependent on EKLF (bottom left). In a situation where multiple genes are potentially active, the genes that are closest to the LCR would be activated because of gene competition (Tanimoto et al. 1999). In definitive erythroid cells, DRED repressor protein binding to the DR sites prevents EKLF (and other activating proteins) from binding to the distal CAC site of the ɛ-globin promoter (top right) and this gene is autonomously silenced. However, the β-globin gene is active because it does not have DRED-binding sites.