Activation of β-Globin Promoter by Erythroid Krüppel-Like Factor

Abstract

Erythroid Krüppel-like factor (EKLF), an erythroid tissue-specific Krüppel-type zinc finger protein, binds to the β-globin gene CACCC box and is essential for β-globin gene expression. EKLF does not activate the γ gene, the CACCC sequence of which differs from that of the β gene. To test whether the CACCC box sequence difference is the primary determinant of the selective activation of the β gene by EKLF, the CACCC boxes of β and γ genes were swapped and the resulting promoter activities were assayed by transient transfections in CV-1 cells. EKLF activated the β promoter carrying a γ CACCC box at a level comparable to that at which it activated the wild-type β promoter, whereas EKLF failed to activate a γ promoter carrying the β CACCC box, despite the presence of the optimal EKLF binding site. Similar results were obtained in K562 cells. The possibility that overexpressed EKLF superactivated the β promoter carrying the γ CACCC box, or that EKLF activated the mutated β promoter through the intact distal CACCC box, was excluded. To test whether the position of the CACCC box in the β or γ promoter determined EKLF specificity, the proximal β CACCC box sequence was created at the position of the β promoter (−140) which corresponds to the position of the CACCC box on the γ promoter. Similarly, the β CACCC box was created in the position of the γ promoter (−90) corresponding to the position of the CACCC box in the β promoter. EKLF retained weak activation potential on the β^−140CAC promoter, whereas EKLF failed to activate the γ^−90βCAC promoter even though that promoter contained an optimal EKLF binding site at the optimal position. Taken together, our findings indicate that the specificity of the activation of the β promoter by EKLF is determined by the overall structure of the β promoter rather than solely by the sequence of the β gene CACCC box.

The programmed expression of globin genes is tissue and developmental stage specific. In humans, five β-like globin genes (ɛ, ^Aγ, ^Gγ, δ, and β) form a cluster on the short arm of chromosome 11, and their expression is characterized by two major switches initially from embryonic (ɛ) to fetal (^Aγ and ^Gγ) and subsequently to adult (δ and β) globin gene expression (28). Although a number of cis-acting elements of globin genes and corresponding trans-acting factors have been identified (7, 19), the precise molecular mechanisms of globin gene regulation are still unclear.

The CACCC (or GT) box is a cis-acting element found in a variety of genes expressed in erythroid and nonerythroid tissues. Each β-like globin gene (except the δ gene) has one or two CACCC boxes among the conserved promoter sequences. The importance of the CACCC box for β-globin gene transcription has been demonstrated by the existence of naturally occurring mutations in the proximal CACCC box which cause β⁺ thalassemias (13, 20, 21). The importance of the γ gene CACCC box is shown by the finding that γ gene transcription is reduced when the γ gene CACCC box is deleted (4, 14, 27, 33) and by in vivo footprinting studies showing significant protein binding in the γ CACCC sequence of γ gene-expressing cells (11).

Among the proteins binding to globin gene CACCC boxes, Sp1, a ubiquitous protein (12), and erythroid Krüppel-like factor (EKLF), an erythroid tissue-specific (16) Krüppel-like zinc finger protein, are well characterized. Sp1 is known to interact with the ɛ (37), γ (9), and β (10) gene CACCC boxes, but its in vivo role for globin gene transcription remains unknown. EKLF binds to the proximal β gene CACCC element (2, 16), which enables EKLF to increase the β gene promoter activity in vitro (6). Disruption of the EKLF gene results in a β-thalassemia-like phenotype characterized by lethality of the EKLF^−/− mouse embryos beyond embryonic day 15 due to the deficient β-globin production (18, 23). Similarly, EKLF-deficient mice carrying human β-globin loci cannot express the human β-globin gene but display no reduction in γ gene expression, indicating that β but not γ gene production is dependent on EKLF (22, 36). Thus, EKLF preferentially activates the β gene instead of the γ gene, despite the fact that in the mouse, EKLF is expressed in primitive erythroid cells as well as definitive erythroid cells (26).

Currently, the preferential activation of the β gene by EKLF is attributed to its binding affinity to the target DNA sequences. EKLF binds to an extended 9-bp CACCC box sequence (CCA CAC CCT), which can be recognized by the three zinc fingers of EKLF (2, 16). The analogous CACCC box sequence of the γ gene promoter is CTC CAC CCA. The CACCC box sequence of the β gene shows an affinity to EKLF that is eightfold higher than that of the CACCC box of the γ gene (6). However, there is no evidence that the binding affinity of EKLF to the γ CACCC box is low enough to ablate the γ gene activation by EKLF. β CACCC box sequences carrying a point mutation known to produce a β⁺ thalassemia show a binding affinity for EKLF that is 40- to 100-fold lower than that of the wild-type β CACCC box sequence (8), i.e., much lower than the binding affinity of EKLF for the 9-bp sequence of the γ CACCC box. Other reasons, in addition to the decreased affinity, may account for the lack of activation of the γ gene promoter by EKLF.

The purpose of this study was to test whether the difference in the 9-bp CACCC box sequence between the β and γ gene promoters is the sole determinant of the preferential β gene activation by EKLF. Our results show that the selective activation of the β gene by EKLF is dependent on the whole promoter context of the β-globin gene rather than exclusively on the sequence of the β gene CACCC box.

MATERIALS AND METHODS

Plasmid constructions.

pHS2γ^βCACLuc, containing the β gene CACCC box in the γ promoter, and pHS2β^γCACCAT, containing the γ gene CACCC box in the β promoter, were generated by using the Altered Sites II in vitro mutagenesis system (Promega). Briefly, a KpnI-BamHI fragment of pHS2γLuc (a generous gift from Tim M. Townes) and a PstI-BamHI fragment of pHS2βCAT (also a gift from Tim M. Townes) were subcloned into KpnI- and BamHI-digested and PstI- and BamHI-digested pALTER-1 vectors, respectively. γ and β CACCC boxes were substituted by β and γ CACCC boxes, respectively, by using 5′ phosphorylated oligonucleotides 5′-pGATTGGCCAACCCATGGGTGGAGTTCCACAGGGTGA-3′ and 5′-pGTCCCTGGCTAAGCCACACCCTTGGGTTGGCCAG-3′, respectively. After a mutagenesis reaction, a KpnI-BamHI fragment with a γ promoter containing a β CACCC box and a PstI-BamHI fragment with a β promoter containing a γ CACCC box were put back into KpnI- and BamHI-digested pHS2γLuc and PstI- and BamHI-digested pHS2βCAT, respectively. Similarly, plasmids containing a β promoter with point mutations which cause β⁺ thalassemia (pHS2β^−88mutCAT, pHS2β^−87mutCAT, and pHS2β^−86mutCAT) were constructed from pHS2βCAT. pHS2β^γCACΔdCACCAT, in which the distal CACCC box sequence of the β promoter was disrupted and the proximal CACCC box sequence was substituted by the γ CACCC sequence, was derived from pHS2β^γCACCAT. pHS2β^−140CACCAT, in which the proximal β CACCC box sequence was moved to the γ CACCC box position, was prepared from γ CACCC sequence-disrupted pHS2β^γCACΔdCACCAT (pHS2β^ΔdpCACCAT). pHS2γ^−90βCACLuc, in which the proximal β CACCC box sequence was generated at the position where it is located in the β promoter, was prepared from γ CACCC sequence-deleted pHS2γLuc (pHS2γ^ΔCACLuc). 5′ phosphorylated oligonucleotides used for these plasmid constructions are 5′-pGCC AACCCTAGGATGTGGCTCCACA-3′, 5′-pGGCCAACCCTAGCGTGTGGCTCCAC-3′, 5′-pTGGCCAACCCTACGGTGTGGCTCCA-3′, 5′-pGTGGAGTTCCACACTAGTAGGTCTAAGTGAT-3′, 5′-pATTGGCCAACCCTTCATATGAGTTCCACACTAGT-3′, 5′-pCGTACCTGTCCTTGAGGGTGTGGAGCTCTTCTGGCACT-3′, 5′-pGTCCCTGGCTAAATGGGTTGGCCAG-3′ and 5′-pCAAACTTGACCAATACCACACCCTAGGTCTTAGAGTATCCA-3′ for pHS2β^−88mutCAT, pHS2β^−87mutCAT, pHS2β^−86mutCAT, pHS2β^γCACΔdCAC CAT, pHS2β^ΔdpCACCAT, pHS2β^−140CACCAT, pHS2γ^ΔCACLuc, and pHS2γ^−90βCACLuc, respectively. Plasmids with mutations were verified by DNA sequencing by using a kit (Cyclist; Stratagene).

Transactivation analysis.

CV-1 cells and K562 cells were cultured in Eagle’s minimal essential medium and RPMI 1640, respectively, supplemented with 10% fetal calf serum. Transient transfections of CV-1 cells were performed by the calcium phosphate coprecipitation method (1). Briefly, 1.8 × 10⁵ CV-1 cells were plated in a 6-cm-diameter culture dish 24 h prior to transfection; 3.6 ml of fresh complete medium was added 2 to 4 h before transfection. A DNA mixture containing 3.6 μg (except where indicated otherwise) of each activator, reporter, and pSVβ-Gal control vector (Promega) was ethanol precipitated, rinsed with 80% ethanol, air dried, dissolved in 180 μl of distilled H₂O, and mixed with 20 μl of 2.5 M CaCl₂. The DNA solution was mixed with 200 μl of 2× HEPES-buffered saline and added to the culture medium. After overnight incubation, cells were glycerol shocked. The cells were completely washed, further incubated for 24 h in the complete medium, and then lysed in 325 μl of reporter lysis buffer (Promega).

Transient transfection of K562 cells was performed with reporter, expression (10 times greater molar amount of the reporter plasmid), and pSG5 vectors (to total 40 μg) and 10 μg of pSVβ-Gal. Log-phase cells (3 × 10⁷ to 4 × 10⁷) in RPMI 1640 medium were electroporated at 960 μF and 320 mV (Bio-Rad Gene Pulser). After standing at room temperature for 10 min, the cells were plated in 10 ml of the complete medium, incubated at 37°C for 24 h, and then harvested. The cell extracts were prepared in 400 μl of reporter lysis buffer. Aliquots (100 μl) of the extracts, which had been diluted 1:10 in the EKLF samples, were heat inactivated and assayed for chloramphenicol acetyltransferase (CAT) activities by the phase extraction method (25). For luciferase assays, the cell extracts were diluted 1:10, and 100-μl aliquots were analyzed by using the (Promega) luciferase assay system.

All transfection assays were performed multiple times and with different preparations of the same plasmid. CAT and luciferase activities obtained were corrected for transfection efficiencies by β-galactosidase (β-Gal) A₄₀₅.

RESULTS

EKLF activates a β-globin gene promoter which contains the γ CACCC box sequence.

EKLF interacts with the β-globin gene CACCC box by recognizing the 9-bp sequence CCA CAC CCT (2, 16). The sequence of the CACCC box of the γ-globin promoter is CTC CAC CCA. To test whether the 9-bp CACCC box sequence difference between β- and γ-globin gene promoters is the sole determinant of the selective function of EKLF on the β gene promoter, we substituted the original β gene CACCC box sequence of pHS2βCAT with the γ gene CACCC box sequence. The resultant construct was designated pHS2β^γCACCAT (Fig. 1).

FIG. 1.

(A) Structure of pHS2β^γCACCAT, containing a β gene promoter with a γ CACCC box. The CAT gene is driven by a 1.5-kb KpnI-BglII fragment of HS2 and the β gene promoter (a fragment extending from bp −265 to +48 relative to the cap site) carrying a γ CACCC box. Uppercase letters denote the 9-bp γ CACCC sequence analogous to the β CACCC sequence recognized by EKLF. Numbers above the γ CACCC box sequence show base pair distances from the cap site. (B) Transactivation of a β promoter containing a γ CACCC box by EKLF. CAT activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of CAT activity of pHS2βCAT in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are expressed as mean (columns) ± SD (error bars) derived from four independent transfections using two different plasmid sets. Notice that EKLF transactivates the β gene promoter despite the fact that the promoter contains the γ CACCC box.

The reporter constructs pHS2βCAT and pHS2β^γCACCAT and the activator plasmid pSG5/EKLF were transiently cotransfected into CV-1 cells with plasmid pSVβ-Gal as an internal control of transfection efficiency. CV-1 is an established cell line derived from monkey kidney and has been previously used by Bieker and Southwood (3) to evaluate the activity of EKLF on globin gene promoters. As shown in Fig. 1, EKLF increased the activity of the β gene promoter of pHS2βCAT (i.e., the construct containing the normal CACCC box) by roughly 800% of the control value. Notice that EKLF activated the β^γCAC gene promoter as effectively as the wild-type β promoter, even though the β^γCAC gene promoter contains the γ CACCC box instead of the β CACCC box. The average CAT activities driven by the β promoter carrying the γ CACCC box sequence were 97% (without EKLF) and 960% (with EKLF) relative to that driven by pHS2βCAT in the absence of EKLF stimulation (taken as 100%).

These findings suggested that EKLF functions in the context of the whole β gene promoter rather than exclusively through its affinity to the 9-bp sequence of the β CACCC box.

EKLF fails to activate a γ-globin gene promoter which contains the β CACCC box sequence.

If the 9-bp β CACCC box sequence has a critical role for EKLF function, we would expect EKLF to activate a γ-globin gene promoter containing the β CACCC box sequence instead of the γ CACCC box sequence. To test this possibility, we substituted the γ gene CACCC box sequence of pHS2γLuc with the β gene CACCC box sequence. The resultant construct was designated pHS2γ^βCACLuc (Fig. 2). The reporter constructs pHS2γLuc and pHS2γ^βCACLuc, plus pSG5/EKLF and pSVβ-Gal, were transfected into CV-1 cells.

FIG. 2.

(A) Structure of the construct pHS2γ^βCACLuc containing a γ gene promoter with a β CACCC box. The luciferase gene is driven by the 1.5-kb fragment of HS2 and the γ gene promoter (bp −299 to +37 relative to the cap site) carrying a β CACCC box. Uppercase letters denote the 9-bp β sequence recognized by EKLF. Numbers above the β CACCC box sequence show base pair distances from the cap site. (B) Transactivation of a γ promoter containing a β CACCC box by EKLF. Luciferase activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of luciferase activity of pHS2γLuc in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are derived from four independent transfections using two different plasmid sets. Notice that EKLF cannot transactivate the γ gene promoter despite the fact that the promoter contains the β CACCC box.

As shown in Fig. 2, in the presence of EKLF, the average luciferase activity derived from pHS2γLuc was 120% of the activity obtained in the absence of EKLF (100%). Thus, EKLF did not activate the γ gene promoter. The average luciferase activities from pHS2γ^βCACLuc were 58% (without EKLF) and 92% (with EKLF) relative to the activity obtained from pHS2γLuc without EKLF (100%). Therefore, EKLF failed to activate the γ^βCAC gene promoter, although the γ^βCAC gene promoter carried the β CACCC box sequence, an optimal EKLF binding site. These results provided further evidence that the 9-bp β CACCC box sequence is not enough to mediate β-globin gene-specific EKLF function.

The effects of EKLF on the β^γCAC and γ^βCAC promoters are reproduced in the erythroid environment.

The results described above were obtained in transactivation assays using a nonerythroid line, CV-1. It was possible that these results reflected the lack of other transcriptional factors which are present in erythroid cells. For these reasons, we repeated the transient transfection assays with K562 cells, a human erythroleukemia line which exhibits an embryonic/fetal globin phenotype. There is no endogenous β gene expression in K562 cells, but β gene constructs linked to locus control region cassettes or to DNase I-hypersensitive site 2 (HS2) display efficient β gene transcription (34, 38). The reporter constructs pHS2βCAT, pHS2γLuc, pHS2β^γCACCAT, and pHS2γ^βCACLuc, plus pSG5/EKLF and pSVβ-Gal, were transiently transfected into the K562 cells.

As shown in Fig. 3, EKLF activated pHS2βCAT and pHS2β^γCACCAT to similar degrees. The average CAT activities driven by β and β^γCAC in the presence of EKLF were 1,062 and 936%, respectively, relative to that of pHS2βCAT in the absence of EKLF (100%). As in the experiments with CV-1 cells, substitution of the CACCC box of the γ promoter by the β CACCC box did not increase the effect of EKLF on the γ promoter. The average luciferase activities stimulated by EKLF were 292% (pHS2γLuc) and 235% (pHS2γ^βCACLuc); thus, the β CACCC box-containing γ promoter was not activated by EKLF more than the wild-type γ gene promoter. These results provide further evidence that CACCC box recognition is not the sole determinant of EKLF activity.

FIG. 3.

Studies using K562 cells. Results of transactivation by EKLF of a β promoter carrying a γ CACCC box and a γ promoter carrying a β CACCC box are depicted. CAT and luciferase activities were normalized to β-Gal activity and expressed as relative percentages of CAT and luciferase activities of pHS2βCAT and pHS2γLuc in K562 cells which were not transfected by a transactivator plasmid (100%). Data are derived from four independent transfections using two different plasmid sets. Notice that the CACCC box substitutions do not influence the level of activation of the β or γ gene promoter by EKLF.

The pattern of promoter activation by EKLF in K562 cells (Fig. 3) is very similar to that in CV-1 cells (Fig. 1 and 2), indicating that the results observed in CV-1 cells are not an artifact caused by the nonerythroid environment. Hence, we used CV-1 cells in all subsequent experiments.

Activation of the β^γCAC promoter by EKLF cannot be attributed to EKLF overexpression.

Since the transactivator gene used in a transient expression system is generally overexpressed for full activation of the reporter gene, the results described above, especially those for β^γCAC gene promoter activation by EKLF, could be attributed to overexpression of EKLF. We examined the relationship between the transfected amount of pSG5/EKLF and the CAT activity from pHS2βCAT to (i) determine whether EKLF is overexpressed under the experimental condition that we used and if so (ii) find transfection conditions which are not associated with pSG5/EKLF overexpression.

Amounts of pSG5/EKLF ranging from 0 to 3.6 μg were cotransfected with pHS2βCAT and pSVβ-Gal into CV-1 cells. Results are shown in Fig. 4. The highest CAT activity obtained from cells transfected with 3.6 μg of pSG5/EKLF was taken as 100%, and CAT activities obtained with the lower concentrations of pSG5/EKLF were expressed as percentages of this highest activity. CAT activity exhibited a plateau between 0.7 and 3.6 μg of pSG5/EKLF (Fig. 4), indicating that EKLF was overexpressed in our previous CV-1 transfection studies in which 3.6 μg of pSG5/EKLF was used. Between 0 and 0.7 μg, CAT activity increased toward a plateau level along with the increase in the amount of pSG5/EKLF. Thus, transfection using less than 0.7 μg of pSG5/EKLF does not produce full activation of the β gene promoter and is considered to give rise to unsaturated EKLF expression in CV-1 cells.

FIG. 4.

Relationship between amounts of transfected activator plasmid and degree of activation of the reporter gene in CV-1 cells. The reporter construct, pHS2βCAT, was cotransfected with various amounts of pSG5/EKLF. CAT activities were normalized to β-Gal activity and expressed as relative percentages of CAT activity of pHS2βCAT transfected with 3.6 μg of pSG5/EKLF (100%). Average values ± SD (error bars) were derived from three independent transfections using two different plasmid sets. Notice that CAT activities show two phases, ascending (0 to 0.7 μg of EKLF plasmid) and plateau (0.7 to 3.6 μg of EKLF plasmid).

To test whether pSG5/EKLF overexpression enabled EKLF to activate the β gene promoter carrying the γ CACCC box instead of the β CACCC box, we repeated the transient transfection experiments with CV-1 cells, with 0.5 μg of pSG5/EKLF as a transactivator. The amount of the EKLF plasmid used should create unsaturated EKLF expression in the cells.

As shown in Fig. 5, transfection of 0.5 μg of pSG5/EKLF activated the β and β^γCAC gene promoters to similar degrees. The average CAT activities of pHS2βCAT and pHS2β^γCACCAT stimulated by EKLF were 639 and 601%, respectively, relative to that of pHS2βCAT lacking EKLF stimulation (100%). Thus, the activation by EKLF of a β promoter carrying the γ CACCC box sequence was consistently comparable to that of the β gene promoter containing the β CACCC box. These data suggest that pSG5/EKLF overexpression is not the cause of activation of the β^γCAC promoter by EKLF.

FIG. 5.

Transactivation of the β promoter containing a γ CACCC box by a small amount of pSG5/EKLF (0.5 μg per transfection). CAT activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of CAT activity of pHS2βCAT in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are derived from three independent transfections using two different plasmid sets. Notice that EKLF transactivates the β gene promoter containing the γ. CACCC box, a result which is similar to the results of assays using a standard amount (3.6 μg) of pSG5/EKLF (Fig. 1).

The distal CACCC box of the β-globin promoter is not the cause of activation of the β promoter containing a γ CACCC box sequence.

The β gene promoter has two CACCC boxes, one proximal and one distal, at bp −90 and −105 relative to the cap site; EKLF recognizes the proximal (−90) CACCC box sequence (16). A possible interpretation of our findings that EKLF can activate a β promoter carrying a γ CACCC box is that in the absence of the wild-type β gene CACCC box at −90, EKLF interacts with the distal CACCC box, resulting in β promoter activation. This interpretation is unlikely because naturally occurring mutations of the proximal CACCC box of the β gene cause β⁺ thalassemia (35), and one of these promoter mutations has been shown by Donze et al. (6) to significantly decrease β gene promoter activation by EKLF in transient expression assays; therefore there is evidence that the distal CACCC box contributes minimally, if at all, to β gene activation.

To test whether there is functional interaction between EKLF and the distal CACCC box under the conditions of our experiments, we generated three types of point mutations in the proximal β gene CACCC box: −88 (relative to the cap site) C→T (pHS2β^−88mutCAT), −87 C→G (pHS2β^−87mutCAT), and −86 C→G (pHS2β^−86mutCAT). Reporter plasmids carrying these mutations, 3.6 μg of EKLF plasmid, and pSVβ-Gal were transiently cotransfected into CV-1 cells. Absolute promoter activities were decreased to similar degrees in all three mutated β promoters. The CAT activities with and without EKLF were 151% ± 23% (mean ± standard deviation [SD]) and 20% ± 7%, respectively, for pHS2β^−88mutCAT, 109% ± 15% and 25% ± 5% for pHS2β^−87mutCAT, and 148% ± 10% and 18% ± 7% for pHS2β^−86mutCAT. In the same experiments, EKLF activated the wild-type β promoter of pHS2βCAT to 634% ± 81% of the control without EKLF (100%). These results indicated that the lack of functional interaction between EKLF and the distal CACCC box is also observed under the experimental condition that we have used.

To test the role of the distal CACCC element more directly, we disrupted the distal CACCC box of the pHS2β^γCACCAT construct by producing the nucleotide substitutions CCT CAC CCT→CCT ACT AGT; the resulting construct was designated as pHS2β^γCACΔdCACCAT (Fig. 6A). When they are introduced in the comparable residues in the proximal CACCC box, the point mutations denoted by the underlined three C residues cause thalassemias and reduce the interaction of the CACCC box with EKLF. The reporter constructs pHS2β^γCACCAT and pHS2β^γCACΔdCACCAT, plus pSG5/EKLF and pSVβ-Gal, were transfected into CV-1 cells. The transfected amount of the EKLF plasmid was 0.5 μg in this experiment. CAT activity was increased about sixfold by the addition of EKLF in both pHS2β^γCACCAT and pHS2β^γCACΔdCACCAT compared to that of pHS2β^γCACCAT without EKLF (100%), although the variation was relatively large (Fig. 6B). Therefore, EKLF activated the β^γCAC promoter with a disrupted distal CACCC box, even though this promoter had no original β CACCC box sequences. This finding provides direct evidence that the presence of an intact distal CACCC box is not the cause of activation of the γ CACCC box-containing β promoter by EKLF.

FIG. 6.

(A) Disruption of the distal CACCC box sequence (β^γCACΔdCAC). The distal CACCC box sequence (CCT CAC CCT) of the β promoter carrying a γ CACCC box sequence (CTC CAC CCA) at the proximal CACCC site is altered to CCT ACT AGT. Numbers above the promoter sequences show base pair distances from the cap site. (B) Transactivation of a distal CACCC box-disrupted β^γCAC promoter by EKLF. A small amount of EKLF plasmid (0.5 μg) was used for transfection. CAT activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of CAT activity of pHS2β^γCACCAT in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are derived from three independent transfections using two different plasmid sets. Notice that in the presence of EKLF, similar levels of CAT activities were obtained from β^γCAC and β^γCACΔdCAC.

The position of the CACCC box is not critical for activation of the β gene promoter by EKLF.

The results described above demonstrate that the difference in the 9-bp CACCC box sequence between β and γ gene promoters is not a critical determinant of the specificity of the β gene activation by EKLF. As shown in Fig. 7, there are two major differences between the β and γ CACCC boxes: one is the position of the CACCC box relative to the cap site, and the other is the configuration of the surrounding cis elements. To test whether the position of the CACCC box confers the β gene specificity on EKLF, we generated a γ promoter which contained a β CACCC box sequence placed in the position in the β promoter (i.e., 90 bp upstream from the cap site). We also generated a β promoter which contained a β CACCC box in the position where the γ CACCC box is normally located in the γ promoter (i.e., 140 bp upstream from the cap site). The normal CACCC box sequence of each promoter was deleted or disrupted.

FIG. 7.

Comparison of the locations of cis elements of the β and γ gene promoters. The positions of the functional CACCC box are shown by solid rectangles.

The construct pHS2γ^−90βCACLuc, in which the original γ CACCC box was deleted and the proximal β CACCC box was created at exactly the same position as in the β promoter, is shown in Fig. 8A. The reporter constructs pHS2γLuc and pHS2γ^−90βCACLuc, plus pSG5/EKLF and pSVβ-Gal, were transiently transfected into CV-1 cells. If the position of the CACCC box sequence is an important determinant for selective activation of the β gene promoter by EKLF, we would expect activation of the γ^−90βCAC gene promoter by EKLF. As shown in Fig. 8B, the average luciferase activity of pHS2γ^−90βCACLuc was less than 10% of that of pHS2γLuc without EKLF stimulation (100%), and the addition of EKLF did not alter the low activity. Thus, EKLF failed to activate the γ gene promoter even though its optimal CACCC box sequence is placed at a distance from the transcription start site which is optimal for functioning in the β gene promoter. These results suggest that the β-globin gene specificity of EKLF is not determined solely by the position of the CACCC box.

FIG. 8.

Effect of the position of the CACCC box on activation of the γ gene promoter by EKLF. (A) Generation of a γ promoter containing a β CACCC box at bp −90 (γ^−90βCAC). The original γ CACCC box sequence (CTC CAC CCA) was deleted, and the proximal β CACCC box sequence was inserted into position −90, i.e., the position of the β CACCC box in the β promoter. Numbers above the promoter sequences are base pair distances from the cap site. (B) Transactivation of a γ promoter containing a β CACCC box at position −90 (β^−90βCAC) by EKLF. Luciferase activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of luciferase activity of pHS2γLuc in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are derived from four independent transfections using two different plasmid sets. Notice that the promoter activity is almost ablated by the CACCC box movement. In addition, EKLF cannot activate the mutant promoter even though this promoter contains an optimal binding sequence at its optimal site.

Figure 9A shows the reporter construct pHS2β^−140CACCAT, in which both proximal and distal CACCC boxes were disrupted and the proximal β CACCC box was created at exactly the same position as that where the γ CACCC box is located in the γ promoter. This CACCC box relocation decreased the basal CAT activity without EKLF stimulation to about 20% of that of pHS2βCAT (100%). In contrast to the relocation of the CACCC box in the γ gene promoter, EKLF activated the β^−140CAC promoter to about 165% relative to the original β promoter without EKLF (100%) (Fig. 9B). Thus, the β promoter containing a relocated CACCC box (β^−140CAC) retained mild reactivity to stimulation by EKLF.

FIG. 9.

Effect of the position of the CACCC box on the activation of the β-globin gene promoter by EKLF. (A) Generation of a β promoter containing a CACCC box at bp −140 (β^−140CAC), i.e., the position of the CACCC box in the γ promoter. The proximal CACCC box sequence (CCA CAC CCT) and the distal CACCC box sequence (CCT CAC CCT) were disrupted by base substitutions (shown in underlined italics). Subsequently, the proximal CACCC box sequence was inserted into the γ CACCC position in the γ promoter. Numbers above the promoter sequences correspond to base pair distances from the cap site. (B) Transactivation of a β promoter containing a CACCC box at position −140 (β^−140CAC) by EKLF. CAT activities in CV-1 cells were normalized to β-Gal activity and expressed as relative percentages of CAT activity of pHS2βCAT in CV-1 cells which were not transfected by a transactivator plasmid (100%). Data are derived from four independent transfections using two different plasmid sets. Notice that the promoter activity is remarkably decreased by the relocation of the CACCC box to a position equivalent to that of the CACCC box of the γ gene and that EKLF is still capable of weakly activating this mutant promoter.

These results suggest that the position of the CACCC box in the β or γ gene promoter is not the critical determinant of EKLF function. The relationship of the CACCC box with the surrounding cis elements and the interactions of EKLF with some other protein(s) binding to these elements, i.e., the overall context of the promoter, may determine the specificity of the activation of the β-globin gene promoter by EKLF.

DISCUSSION

The purpose of this study was to investigate whether the 9-bp CACCC box sequence underlies the specificity of activation of the β-globin gene by EKLF. To address this issue, we swapped the CACCC boxes between the β and γ gene promoters and analyzed the activities of EKLF on these mutated promoters by transient transfection assays. The results indicate that the CACCC box sequence of the β gene promoter is not the only determinant of the specific activation of the β gene by EKLF. We have further shown that factors such as the lack of an erythroid environment in the initial transactivation studies, the overexpression of EKLF in the transactivation assays, or the activation of the γ CACCC box-containing β gene promoter through its intact distal CACCC box cannot account for our results. Thus, the selective transcriptional activation of the β-globin gene (compared to that of the γ-globin gene) by EKLF (22, 36) is not due exclusively to the higher affinity of EKLF to the β gene CACCC box sequence. Rather, the specificity of the activation of the β-globin gene by EKLF is dependent on the whole promoter context of the β-globin gene. Thus, just the protein-DNA interaction between the DNA-binding domain of EKLF and the β CACCC box is not sufficient to activate the β-globin gene. Most likely, protein-protein interactions between the EKLF transactivator domain and the transcriptional complex are necessary to bring about the specific activation of the β gene promoter by EKLF.

Insights on the nature of the promoter context-dependent β gene activation by EKLF were provided by the experiments using β and γ gene promoters in which the positions of the CACCC boxes were interchanged. It is reasonable to assume that EKLF, like other transcriptional activators (32), interacts directly or indirectly with the basal transcriptional machinery. Our findings can be explained by assuming that when EKLF is tethered onto the β gene promoter, it interacts with the basal transcriptional machinery formed on the TATA box and flanking regions of the β gene, whereas when it is tethered onto the γ gene promoter by the β CACCC box, it cannot interact with the basal transcriptional machinery on the γ gene. If the β and the γ genes use the same basal transcriptional machinery, a simple explanation of why EKLF, although bound on the β CACCC box of the γ^βCAC promoter, cannot activate γ gene transcription is that the position of the β CACCC box in the β^γCAC promoter does not allow EKLF to interact with the basal transcriptional machinery. However, our results of assays using a γ promoter carrying a β CACCC box at −90 and a β promoter carrying a β CACCC box at −140 (Fig. 8 and 9) do not agree with this hypothesis. Instead, our results argue that the location of the CACCC box relative to the transcription start site is not the critical determinant of the specificity of β-globin gene activation by EKLF.

An alternative explanation of our findings is based on the recruitment model of action of transcriptional activators (24, 29). This model proposes that a transcriptional activator functions by recruiting the transcriptional machinery to the DNA (the regulatory motifs of the promoter). EKLF may act similarly and recruit a subcomplex of the basal transcriptional machinery to the β-globin gene as illustrated in Fig. 10A. If this is so, the putative subcomplex recruited by EKLF must be unique and critical for the assembly of the basal transcriptional machinery on the β gene because disruption of the EKLF gene totally ablates β gene transcription but not γ gene transcription (22, 36). The nonresponsiveness of the γ gene to EKLF can be explained by assuming that the assembly of the basal transcriptional machinery of the γ gene does not utilize the putative subcomplex which is essential for the basal transcriptional machinery on the β gene (Fig. 10B). Instead, the basal transcriptional machinery on the γ gene requires a different subcomplex which can be recruited to the γ gene by a factor that interacts with the γ CACCC box (Fig. 10C).

FIG. 10.

Proposed mechanism of β-globin gene activation by EKLF. (A) EKLF bound to the β gene recruits a subcomplex of the transcriptional machinery and enables formation of a transcription initiation complex (IC) of the β gene together with TFIID containing TATA box-binding protein (TBP), RNA polymerase II (pol II) holoenzyme, and probably other subcomplexes recruited by other transcriptional activators. This initiation complex gives rise to high-level β gene transcription. (B) EKLF tethered to the γ gene by a β CACCC box also recruits the same subcomplex as described above. However, the EKLF-bound subcomplex is different from the subcomplex normally interacting with the γ gene transcriptional machinery and fails to assemble with other components, resulting in failure of initiation of γ gene transcription. (C) Putative γ CACCC box-binding factor recruits an appropriate subcomplex of the transcriptional machinery of the γ gene. The appropriate assembly of the initiation complex on the γ gene gives rise to high-level γ gene transcription.

The hypothesis that the β gene specificity of EKLF is dependent on two factors, protein-protein interaction(s) mediated by the transactivation domain and protein-DNA interaction mediated by the DNA-binding domain, was previously proposed by Bieker and Southwood (3). Although DNA-binding specificity is considered to be the main determinant of promoter specificity of transcriptional activators (17), there is also evidence, from studies of a limited number of transcription factors, that the transactivation domain may also critically influence the promoter specificity of a transcriptional activator (15, 31). Since the transactivator domain of EKLF is composed of multifunctional subdomains (5), it is possible that the DNA binding of EKLF is augmented by the activator domain as is the case of the Oct-2 POU DNA-binding domain (30). In that case, the activator domain of EKLF may play a dual role, i.e., increase the binding of EKLF on the CACCC box of the β promoter and recruit a component of the transcriptional machinery of the β-globin gene.