Loading of DNA-Binding Factors to an Erythroid Enhancer

Abstract

The HS-40 enhancer is the major cis-acting regulatory element responsible for the developmental stage- and erythroid lineage-specific expression of the human α-like globin genes, the embryonic ζ and the adult α2/α/1. A model has been proposed in which competitive factor binding at one of the HS-40 motifs, 3′-NA, modulates the capability of HS-40 to activate the embryonic ζ-globin promoter. Furthermore, this modulation was thought to be mediated through configurational changes of the HS-40 enhanceosome during development. In this study, we have further investigated the molecular basis of this model. First, human erythroid K562 cells stably integrated with various HS-40 mutants cis linked to a human α-globin promoter-growth hormone hybrid gene were analyzed by genomic footprinting and expression analysis. By the assay, we demonstrate that factors bound at different motifs of HS-40 indeed act in concert to build a fully functional enhanceosome. Thus, modification of factor binding at a single motif could drastically change the configuration and function of the HS-40 enhanceosome. Second, a specific 1-bp, GC→TA mutation in the 3′-NA motif of HS-40, 3′-NA(II), has been shown previously to cause significant derepression of the embryonic ζ-globin promoter activity in erythroid cells. This derepression was hypothesized to be regulated through competitive binding of different nuclear factors, in particular AP1 and NF-E2, to the 3′-NA motif. By gel mobility shift and transient cotransfection assays, we now show that 3′-NA(II) mutation completely abolishes the binding of small MafK homodimer. Surprisingly, NF-E2 as well as AP1 can still bind to the 3′-NA(II) sequence. The association constants of both NF-E2 and AP1 are similar to their interactions with the wild-type 3′-NA motif. However, the 3′-NA(II) mutation causes an approximately twofold reduction of the binding affinity of NF-E2 factor to the 3′-NA motif. This reduction of affinity could be accounted for by a twofold-higher rate of dissociation of the NF-E2–3′-NA(II) complex. Finally, we show by chromatin immunoprecipitation experiments that only binding of NF-E2, not AP1, could be detected in vivo in K562 cells around the HS-40 region. These data exclude a role for AP1 in the developmental regulation of the human α-globin locus via the 3′-NA motif of HS-40 in embryonic/fetal erythroid cells. Furthermore, extrapolation of the in vitro binding studies suggests that factors other than NF-E2, such as the small Maf homodimers, are likely involved in the regulation of the HS-40 function in vivo.

The interplay among the multiple nuclear factor-DNA complexes formed at the enhancers (enhanceosomes) and their cis-linked promoters (polymerase II [pol II] preinitiation complexes) is essential for the regulation of many eukaryotic genes (reviewed in references 6 and 9). Several modes of actions have been proposed for the enhanceosome function. Some enhanceosomes, such as the locus control region (LCR) of the human β-globin locus (reviewed in references 16, 20, and 22) may set up and/or maintain an active chromatin state of a gene or locus domain, thus allowing the formation of active pol II preinitiation complex. On the other hand, enhanceosomes could also facilitate the assembly of active pol II preinitiation complex by direct interaction with, or recruitment of, coactivators and different basal transcription factors (reference 6 and references therein).

HS-40 is an element located at 40 kb upstream of the human α-globin locus (Fig. 1). Genetic and molecular data have indicated that it is essential for transcriptional regulation of the human embryonic ζ- and adult α-globin promoters during erythroid development (21). Most of the α-globin gene cluster maintains an open chromatin structure in both erythroid and nonerythroid cells, possibly due to the transcription of certain ubiquitous genes (54). Further remodeling and modification of the chromatin structure must occur in erythroid cells, however, since several new HS sites appear, including those at the HS-40 element and the promoter regions of the ζ- and α-globin genes (54, 59). These latter sites apparently result from erythroid lineage-specific binding of nuclear factors to the above transcriptional regulatory elements, as demonstrated by previous genomic footprint analysis of HS-40 (50, 60) and of the α-globin upstream promoters (47).

FIG. 1.

(A) Physical map of the human α-like globin gene locus. A 110-bp region containing the HS-40 element is blown up below the cluster map. The different factor-binding motifs, as mapped previously (for references, see text), are indicated. (B) Nucleotide sequences of wild-type and mutant motifs of HS-40. The wild-type sequences are shown in full, with the central binding sites for nuclear factors indicated with thick letters. The mutated bases of the lower strands are indicated by the downward arrows.

HS-40 acts as a classical enhancer in transient transfection assays (44, 46, 60), and it confers appropriate developmental control of the human ζ-globin promoter activity in transgenic mice (19, 23, 45, 56). In vitro and in vivo binding studies have shown that the HS-40 enhanceosome consists mainly of six DNA sequence motifs that are bound with nuclear factors in an erythroid lineage-specific manner: two NF-E2/AP1 motifs (5′-NA and 3′-NA), three GATA-1 motifs (b, c, and d), and a GT motif (26, 50, 60) (Fig. 1A). Of these, the NF-E2/AP1 motifs could be recognized by the erythroid-enriched factor NF-E2, the ubiquitous AP1, the homodimers of small Maf family, or other proteins such as Nrf1, Nrf2, Nrf3, Bach1, and Bach2 (reviewed in references 7, 30, and 39). The GATA-1 motif is recognized by the erythroid-enriched factor GATA-1 (26, 52). Sp1 is the predominant protein binding to the GT motif in vitro (12, 26). Relatively little is known regarding proteins or factors bound in the outer layer of the HS-40 enhanceosome. It has been shown though that NF-E2 can associate with one of the transcription coactivators (CREB-binding protein) (11), the general transcription factors (1), a class of WW domain-containing proteins including the ubiquitin ligase Rsp5 (17, 38), and a novel transcription corepressor (N. R. Gavva, P. Daftari, L.-P. Yang, and C.-K. J. Shen, unpublished data). Also, GATA-1 factor can interact with Sp1 (36) as well as with an erythroid-specific coactivator, FOG (53).

With the use of site-directed mutagenesis and transient transfection, we have previously shown that HS-40 motifs 5′-NA, 3′-NA, GT, and GATA-1(c), but not GATA-1(b) or GATA-1(d), positively regulate the erythroid-specific enhancer function of HS-40 on the expression of both ζ- and α-globin promoters in human embryonic/fetal erythroid cell line K562 (47, 61). Furthermore, competitive factor binding at the 3′-NA motif appears to modulate the function of HS-40 as a negative regulatory element for human ζ-globin promoter expression during the embryonic-to-fetal development (23, 47, 61). In particular, a 1-bp mutation in the 3′-NA motif, which was expected to abolish NF-E2 binding but not that of AP1 (see Discussion), converts HS-40 to a potent erythroid-specific enhancer for the expression of human ζ-globin promoter in fetal and adult erythroid cells (23, 61).

To further understand the molecular basis of the regulatory function of HS-40 during erythroid development, we have analyzed the interplay of four factor-binding motifs in the assembly of the HS-40 enhanceosome. In particular, we examined whether individual factor-DNA complexes within HS-40 interact with each other to form the molecular switch. That is, is there a hierarchy so that pivotal binding at one motif is prerequisite for factor-DNA complex formation at the other motifs? Because of its involvement in the developmental stage specificity of the ζ-globin gene expression, the 3′-NA motif is the main focus of this study. In particular, we designed experiments to further clarify the possible involvement of several nuclear factors in the competitive binding at the 3′-NA motif and consequently the developmental function of HS-40 in erythroid cells.

MATERIALS AND METHODS

Plasmids.

All recombinant DNA work was done according to standard procedures (48). Construction of the plasmids for stable DNA transfection of K562 cells has been described previously (60, 61). Briefly, a promoterless human growth hormone (AGH)-containing plasmid (p0GH; Nichols Institute) was used to generate plasmid pB-α590GH, which contains a DNA fragment extending from −574 to +21 of the human α1-globin gene cloned at immediately 5′ of the hGH gene in p0GH. The wild type HS-40 and different mutant HS-40 fragments generated by site-directed mutagenesis (Fig. 1B) were individually cloned upstream of the α1-globin promoter in pB-α590GH. Only plasmids containing HS-40 in the genomic orientation relative to the α1-globin promoter were used for stable transfection. Plasmid pRSV-MafK for transient transfection of COS-7 cells was constructed by first release of the MafK cDNA insert from p18 containing pMT2 (57) with EcoRI digestion. The fragment was then cloned into a HindIII/NotI-cut and blunt-ended pRSV-c-jun vector (3) to generate pRSV-MafK. To express recombinant MafK in Escherichia coli, the MafK cDNA was cloned into the EcoRI site of pGEX-4-T2 (Pharmacia Biotech) to create pGEX-MafK. For in vitro transcription and translation, the MafK cDNA was cloned into the EcoRI site of pBluescript II SK (Stratagene).

Stable DNA transfection of K562 cells.

Human erythroid K562 cells were cultured under 5% CO₂ in RPMI 1640 medium containing 10% fetal bovine serum, 50 μg of streptomycin per ml, and 5 U of penicillin per μl (GIBCO). For stable integration experiments, cells were harvested at densities of approximately 10⁶ cells/ml. A total of 0.5 × 10⁷ cells were pelleted and resuspended in 0.4 ml of RPMI 1640 medium containing 15 μg of test plasmids, 1 μg of pUC9γ neo^k (a gift from T. Ley), and 35 μg of carrier salmon sperm DNA. DNA samples were electroporated into the cells with Bio-Rad Gene Pulser. Pools of stable integrants were obtained after growth for 2 weeks under selection with the drug G418 (600 μg/ml).

To select for low-copy-number transfectants, exponentially growing cells of each selected pool were diluted to a final density of 1 cell per 100 μl. Each 100-μl aliquot was then put into one well of a 96-well microplate. The wells containing a single cell were identified under the microscope and observed for at least 1 week to ensure that the entire cell population indeed propagated from the single cell. The cells were then transferred to 100-mm-diameter dishes once they reached confluency. The copy numbers of integrated plasmids in each cell colony were estimated by Southern blot analysis after digestion of the genomic DNA with PvuII; then the blot was hybridized with the 1.5-kb PstI fragment encompassing positions −574 to +929 of the human α1-globin gene. The intensities of the exogenous and the endogenous gene fragments were compared in a PhosphorImager (Molecular Dynamics). Expression of transfected hGH gene was quantitated with an Allegro hGH radioimmunoassay kit as specified by the manufacturer (Nichols Institute).

In vivo footprinting.

The status of nuclear factor binding in the living K562 cells was investigated by dimethyl sulfate (DMS) cleavage in vivo and ligation-mediated PCR (LMPCR) as described previously (40, 43, 60). The endogenous HS-40 was analyzed with primer set E (413 [5′-TCAGGCTTTGCCCCTGAAGC-3′], 295 [5′-AGGCTTTGCCCTGAAGCCTGGCTGT-3′], and 403 [5′-GGCTGTGAACACTTTGGGCATGG-3′]); the genomic footprints of the transfected HS-40 were analyzed with the primer set EA (α528 [5′-AAGAATTTCTGCGCAGAGCC-3′], 295 [5′-AGGTTTGCCCTGAAGCCTGGCTGT-3′], and 403 [5′-GGCTGTGAACACTTTGGGCATGG-3′]).

Different batches of DMS-treated cells were analyzed several times to check for consistency of the protection patterns. The relative intensities of the bands on the autoradiographs from the above assays were estimated in a PhosphorImager.

Transient DNA transfection.

For transfection of the COS-7 cell line, the cells were grown in Dulbecco modified Eagle medium–10% fetal calf serum and transfected at 60% confluency in 10-cm-diameter dishes. The DNA constructs were introduced as a calcium phosphate coprecipitate consisting of 30 μg of pRSV-MafK. pRSV was used for control transfections. Sixteen hours after addition of the calcium phosphate DNA precipitate, the medium was changed and cells were incubated for an additional 48 h before whole cell extract preparation.

For transfection of K562 cells, maintenance and electroporation were as described previously (61). Eight micrograms of one of the GH reporter plasmids (see Table 2), 10 μg of salmon sperm DNA, 1 μg of pCMV-β-gal expression plasmid, and 30 μg of pRSV-MafK or pRSV (control) were used for transfection. Cells were grown for 48 h. GH and β-galactosidase assays were performed as described elsewhere (5).

TABLE 2.

Effect of MafK overexpression on the α1 globin promoter activities in K562 cells^a

Reporter plasmid	HS-40	Cotransfected MafK	Mean relative promoter activityb ± SD (n)
pHS-40-α590GH	Wild type	−	100 (8)
pHS-40-α590GH	Wild type	+	32 ± 1 (8)
pHS-40(3′-NA-II)-α590GH	3′-NA(II)	−	75 ± 12 (9)
pHS-40(3′-NA-II)-α590GH	3′-NA(II)	+	83 ± 10 (9)

^aK562 cells were first transiently cotransfected with pHS40-α590GH plus 1, 5, 10, 20, and 30 μg of pRSV-MafK, which resulted in promoter activities, as percentages of the activity of pHS-40-α590GH without cotransfected pRSV-MafK of 74 ± 7, 43 ± 11, 33 ± 11, 35 ± 10, and 26 ± 2, respectively. Two independent experiments were performed in duplicate. 

^bEffect of 30 μg of cotransfected pRSV-MafK on activities of the α1 globin promoter linked in cis to wild-type HS-40 or HS-40 with the 3′-NA(II) mutation. The promoter strengths are relative to that of pHS-40-α590GH without cotransfected pRSV-MafK. 

EMSA.

Proteins used for electrophoretic mobility shift assay (EMSA) were prepared in the following ways. Whole cell extracts from K562 and COS-7 were prepared as described elsewhere (27). K562 nuclear extracts were prepared as described by Dignam (14). Extract preparation was done in the presence of phenylmethylsulfonyl fluoride, leupeptin, aprotinin, and pepstatin. To prepare glutathione S-transferase (GST) fusion proteins, E. coli BL21 cells were transformed with pGEX-MafK or pGEX-4-T2 (control) and grown to an optical density at 600 nm of 0.7; fusion protein was induced for 6 h with isopropyl-β-d-thiogalactopyranoside (0.1 mM). Crude lysates were prepared by mild sonication of the bacteria on ice, using short bursts. Extracts were cleared by centrifugation (12,000 × g), and the supernatants were stored at −80°C and used directly for EMSA. Proteins were also generated by in vitro transcription and translation using a TNT kit (Promega). Samples were incubated at 30°C for 90 min before DNA-binding activity was measured by EMSA.

The oligonucleotides used in EMSA included 3′-NA (5′-GATCCAGGACTGCTGAGTCATCCTG-3′/3′-GTCCTGACGACTCAGTAGGACCTAG-5′), 3′-NA(II) (5′-GATCCAGGACTTCTGAGTCATCCTG-3′/3′-GTCCTGAAGACTCAGTAGGACCTAG-5′), and 5′-NA (5′-GATCCGCCAACCATGACTCAGTGCGG-3′/3′-GCGGTTGGTACTGA-GTCACGCCTAG-5′).

Gel shifts were performed with 20 μg of whole cell extract or 5 μg of nuclear extract preincubated at room temperature for 5 min (20 mM HEPES [pH 7.9], 100 mM KC1, 3.2 mM MgCl₂, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 12% glycerol) containing 1 μg of poly(dI-dC) · poly(dI-dC) in 20 μl. Radioactively labeled oligonucleotides (20,000 cpm, approximately 0.5 ng) were added, and the samples were incubated for 15 min at room temperature. To separate DNA-protein complexes, samples were loaded onto a running nondenaturing 4% polyacrylamide gel (prerun for 30 min at 4°C and 200 V) in 0.5× Tris-borate-EDTA buffer. Electrophoresis was carried out at 4°C and 200 V for 2.5 h. For comparison, the EMSA conditions described by Ney et al. (41) were used in lanes 13 and 14 of Fig. 4; after incubation of the binding reaction (20,000 cpm of DNA, approximately 0.5 ng; 20 mM HEPES [pH 7.8], 60 mM KC1, 6 mM MgCl₂, 0.2 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol) for 30 min at 25°C, samples were run on a nondenaturing 4% polyacrylamide gel in 0.25× Tris-borate-EDTA buffer at 10 V/cm for 90 min at room temperature.

FIG. 4.

EMSA of 3′-NA and 3′-NA(II) oligonucleotides in K562 extracts. For panels A and B, the whole cell extract was used either directly (lanes 1, 4, 7, and 9) or preincubated with preimmune (PI) serum (lanes 2 and 5), anti-p45 antiserum (lanes 3 and 6), and anti-c-Jun antiserum (lanes 8 and 10); for panel C, K562 nuclear extract was used instead. NF-E2 and AP1 indicate the NF-E2–DNA and AP1–DNA complexes, respectively; a and b are complexes of unknown identities.

To further identify the complexes, 1 μl of anti-MafK (Santa Cruz Biotechnology Inc.) or anti-p45 (17) antiserum was preincubated with the extract for 10 min at room temperature. The preincubated extracts were then used for analysis by standard EMSA. Preimmune rabbit immunoglobulin G antibody was used as the control. Gels were dried and exposed on X-ray film at −80°C.

Competitive binding EMSA.

To analyze complex formation at 15°C as a function of time (Fig. 7A and B), whole cell extract (35 μg) was incubated with labeled probe (1.4 μCi, approximately 35 ng) at 15°C in a total volume of 130 μl. Aliquots (20 μl) of the binding reaction were loaded onto an already running polyacrylamide gel after 0, 1, 2, 4, 8, and 16 min of reaction. The binding conditions were modified so that as the binding reaction approaches equilibrium, the concentrations of the free probe were still approximately 90% of the total probe initially added, as required for a valid kinetic analysis (35). The complexes were quantified with a PhosphorImager. To evaluate the relative affinities (Fig. 7C and D), unlabeled competitor DNA oligonucleotides (5- to 80-fold, approximately 2.5 to 40 ng) were mixed with the radiolabeled oligomer prior to addition of the extract. To assess the stabilities of the complexes (Fig. 7E and F), whole cell extract (140 μg) was first incubated with labeled DNA (140,000 cpm, approximately 3.5 ng) for 15 min at room temperature. A 1,000-fold molar excess of 3′-NA oligonucleotide as the competitor (approximately 3.5 μg) was added, and aliquots of the binding reaction were loaded onto a continuously running gel after further incubation at 15°C for different time periods.

FIG. 7.

(A and B) Time courses for the binding reactions of NF-E2 (A) or AP1 (B) with DNA oligonucleotides containing the 3′-NA or 3′-NA(II) motif; (C and D) relative binding affinities of NF-E2 (C) or AP1 (D) with oligonucleotides containing either of the two motifs; (E and F) dissociation rates of complexes formed between NF-E2 (E) or AP1 (F) and the oligonucleotides containing 3′-NA or 3′-NA(II). The time point at which 50% of the displaceable complexes were removed is defined as the apparent t_½. Under the experimental conditions used, the amounts of the displaceable NF-E2–DNA complexes were approximately 20 and 80% for the 3′-NA and 3′-NA (II) motifs, respectively. Longer incubation for up to 128 min gave similar results (data not shown). Each data point is the average of two to five independent experiments; standard deviations are indicated by error bars.

Chromatin immunoprecipitation.

The procedures and conditions for formaldehyde fixation in vivo, sonication, and immunoprecipitation of chromatin were as described by Orlando and Paro (42). The preparation and use of anti-NF-E2 and anti-c-Jun antibodies for chromatin precipitation and PCR analysis were as described in detail elsewhere (13). Fixed and sonicated chromatin fragments from K562 and HeLa cells were immunoprecipitated with anti-NF-E2, anti-c-Jun, or preimmune sera. DNA samples in the precipitates and the supernatants were purified, and the cross-links were removed. PCR analysis was then carried out with DNA primers specific for different genomic regions. The primers for the HS-40 and β-actin regions were 5′-AGCACATCTGCCCAAGCCA-3′/5′-TCAGGCTTTGCCCCTGAAGC-3′ and 5′-ATCTGGCACCACACCTTCTACAATGAGCTGCG - 3′ / 5′ - CGTCATACTCCTGCTTGCTGATCCACATCTGC - 3′, respectively.

RESULTS

Positive regulatory roles of HS-40 enhancer motifs for human α-globin promoter activity in stably integrated K562 cells.

Among the six DNA motifs of HS-40 that are bound with nuclear factors in a developmental stage- and erythroid lineage-specific manner (50, 60) (Fig. 1), at least four [5′-NA, 3′-NA, GT, and GATA-1(c)] contribute positively to the erythroid-specific activation of the human α-globin promoter in K562 cells, as suggested by transient transfection analysis (47, 61). Five plasmid constructs, each containing a different mutant HS-40 linked to the human α-globin–GH hybrid gene, have been stably integrated into K562 cells. As shown by the GH assay (Table 1), all mutations caused significant decrease of the α-globin promoter activity, by approximately 50 to 90%, compared to the wild-type constructs. The similarity between the present stable integration assay and the previous transient transfection analysis suggests that after integration into a chromosomal environment, the above four DNA motifs still contribute in concert to the erythroid-specific enhancer function of HS-40.

TABLE 1.

Relative promoter activities and copy numbers of individual HS-40 mutant pools and colonies

HS-40 mutant	Relative expression level (%)
	Pool (copy no.)	Clone (copy no.)
5′-NA	58 (11)	24 (2)
		27 (2)
		33 (6)
3′-NA(I)	29 (11)	23 (2)
		25 (2)
		32 (6)
3′-NA(II)	47 (8)	38 (2)
		41 (2)
GT	15 (9)	9 (2)
		12 (2)
		22 (8)
GATA-1(c)	25 (11)	11 (2)
		16 (2)

Functional roles of individual factor-binding motifs in the assembly of the HS-40 enhanceosome.

The genomic footprinting techniques were applied to investigate the contribution of the individual factor-binding motifs to the assembly of HS-40 enhanceosome. The transfected pools contain cells with a wide range in copy numbers of the integrated transgenes. Since multiple copies of integrated plasmids could titrate out limited amounts of certain nuclear factors, we first selected for cells with low copy numbers of the transgenes. A few clones for each of the five transfected mutant plasmids were isolated. Their expression levels of the transgenes relative to the wild type are also listed in Table 1. For each mutant construct, cells from two different clones with two copies of integrated plasmids (Table 1) were treated with DMS and subjected to LMPCR analysis. The choice of the copy number is somewhat arbitrary, because that is the lowest number of plasmid integration we could obtain for each of the five different mutant constructs used for transfection. It should be noted here that a similar approach has been used to investigate nuclear factor binding of other transcriptional regulatory elements, for example, the HLA-DRA promoter (58).

As previously shown for native K562 cells (50, 60), all four motifs of HS-40 on the stably integrated plasmids are also well protected (Fig. 2B). Substitution of 3 bp in the factor-binding cores of 5′-NA and GATA-1(c) motifs, as well as the 1-bp mutation of 3′-NA, had no apparent effects on the presence of the HS-40 genomic footprints (Fig. 2C, E, and G, respectively). On the other hand, 3-bp mutation of either 3′-NA or the GT motif greatly affected the genomic footprints. In particular, the HS-40 enhancer carrying the GT mutation becomes empty (Fig. 2F), while factor binding at both the GT and GATA-1(c) motifs of HS-40 with the 3′-NA(I) mutation was abolished (Fig. 2D).

FIG. 2.

In vivo DMS footprints of wild-type and mutant HS-40 in stably integrated K562 cells. Representative autoradiographs of the analysis of the upper strand of HS-40 by DMS protection assay and LMPCR are shown. Locations of the motifs are indicated by brackets on the left. Numbers on the right correlate with those used in Fig. 1A and reference 61. Protected and hyperreactive residues are denoted by open and closed circles, respectively, on the right of each panel. The large open and closed circles are used when the extents of protection and the relative hypersensitivities, respectively, in vivo (K lanes) are greater than 50% compared to the purified DNA controls (N lanes). The small circle indicate those that are 25% greater than the controls. Also, only those residues consistently showing differences from the controls are indicated. Certain nucleotides, such as residue 175, occasionally did not exhibit an obvious difference (e.g., lanes 1 and 2 in panel C) but did in other sets of the experiments. (A) Footprint of the endogenous (END) HS-40 region; (B) wild type; (C to G) plasmids with HS-40 mutations corresponding to those listed in Fig. 1B. Cell clones with the same copy number of integrated plasmids (Table 1) exhibit similar footprints (data not shown).

It should be noted that there are still footprints present at the mutant motifs 5′-NA (Fig. 2C), 3′-NA(I) (Fig. 2D), and GATA-1(c) (Fig. 2G). However, a close look at the motif sequences (Fig. 1B) indicated that each of the motifs still contains the sequence 5′-CTGA-3′/3′-GACT-5′ after the mutagenesis. This would have allowed, at each of the three mutant motifs in the stably transfected K562 cells, their binding with other factors, such as CREB (29), capable of recognizing the above tetranucleotide sequence. However, it is not practical to identify the factor bound at 3′-NA(II) since too many factors, such as the CREB isoforms and ATF family, are capable of recognizing the CREB site-like sequence in 3′-NA(II). The genomic footprint data are summarized schematically in Fig. 3.

FIG. 3.

Summary of the genomic footprint analysis of HS-40. The top map depicts the factor-binding patterns in vivo of the endogenous HS-40 and the transfected wild-type HS-40, which are indistinguishable. Positions of the mutated base pairs in the different mutant HS-40 motifs are indicated by “x.” Motifs occupied in vivo, as reflected by the genomic footprint analysis of Fig. 2, are covered with different symbols representing the factors bound at the motifs. The random-shaped symbols at the mutated 5′-NA and 3′-NA motifs indicate altered factor binding at these motifs (see text for more details).

NF-E2 recognizes the 3′-NA(II) motif but with a lower binding activity.

EMSA was first used to analyze the effect of the 1-bp mutation NA(II) on nuclear factor-DNA interaction at the 3′-NA motif. The binding reactions were carried out with K562 whole cell extract (Fig. 4A), using oligonucleotides containing the 3′-NA and 3′-NA(II) sequences. Binding of NF-E2 to both of these sequences could be detected, although the intensity of the complex formed at the mutant 3′-NA(II) sequence appears to be weaker (compare lanes 1 and 4 in Fig. 4A). Both NF-E2–DNA complexes could be removed by anti-p45 (lanes 3 and 6) but not by the preimmune serum (lanes 2 and 5). The two motifs could also bind AP1 (Fig. 4A). The authenticity of the AP1-DNA complex was also confirmed with anti-c-Jun (Fig. 4B).

The data in Fig. 4A are somewhat unexpected. Several groups have previously shown that the same GC→TA mutation introduced into the NF-E2/AP1 binding sites of the HS2 element of human β-globin LCR (Fig. 5) completely abolished their NF-E2-binding activity (41, 51). We have thus repeated EMSA using conditions identical to those used by Ney et al. (41). As shown, our EMSA conditions (Fig. 4C, lanes 11 and 12) and theirs (lanes 13 and 14) gave nearly identical results. The difference in NF-E2 binding to the 3′-NA motif of HS-40 and the NF-E2/AP1 motif of HS-2, as affected by the 1-bp mutation, remains to be resolved. It may be due to certain intrinsic sequence characteristics flanking the two motifs or due to minor differences in EMSA conditions not considered (age and quality of the nuclear extracts used, time and temperature of incubation before addition of the labeled probes, loading of binding mixtures onto a running versus a nonrunning gel, time of preelectrophoresis, etc.).

FIG. 5.

NF-E2/AP1 binding sites and their 5′-G/T mutations. Oligonucleotides are shown double stranded, and the restriction sites at their ends are boxed. The 5′-G/T mutations in HS-2mut (41) PDBG-mut (37), and 3′-NA(II) (26) as well as the one nucleotide in 5′-NA not corresponding to the NF-E2 consensus sequence are shown in bold. Sequences corresponding to the NF-E2 consensus are boxed, while the AP1 consensus motif in each oligonucleotide is underlined. For comparison, the upper strands of the consensus motifs for NF-E2 (2), Maf (28), and AP1 (28) binding are shown.

We also carried out EMSA using in vitro-translated polypeptides. MafK cDNA was transcribed and translated in rabbit reticulocyte lysate which already contained p45 (Western blot data not shown). As expected, the translated MafK interacted with the endogenous rabbit p45, and the resulting heterodimers formed specific complexes with either the 3′-NA or 5′-NA motif (Fig. 6, lanes 2 and 8). The specificities of these complexes could be demonstrated by the use of anti-p18 (lanes 3 and 9) or anti-p45 (for 5′-NA, lane 11; for 3′-NA, data not shown). More importantly, the same complexes also formed on the oligonucleotides containing the mutant 3′-NA(II) motif (lanes 5 and 6).

FIG. 6.

EMSA of 3′-NA-, 3′-NA(II)-, and 5′-NA-containing oligonucleotides with in vitro-translated MafK. The assay was carried out with unprogrammed reticulocyte lysate (u.p.; lanes 1, 4, and 7) or with in vitro-translated MafK (lanes 2, 3, 5, 6, and 8 to 11). EMSA was also carried out with the use of preimmune (PI) serum (lane 10), anti-MafK (lanes 3, 6, and 9), or anti-p45 (lane 11). S₁ and S₂ indicate the gel positions of the putative supershifted complexes. The identity of band a is not known.

The lower binding activity of NF-E2 to 3′-NA(II) is due to decreased complex stability.

First, we performed a series of EMSA to analyze the time required for the binding of NF-E2 or AP1 to 3′-NA or 3′-NA(II) (Fig. 7A and B). Formation of the NF-E2/3′-NA complex closely paralleled that of NF-E2/3′-NA(II), in that binding of NF-E2 to either oligonucleotide was rapid and reached a steady state within 10 min (Fig. 7A). As expected, the binding of AP1 to the two oligonucleotides also appeared to be similar (Fig. 7B). The molecular basis of the apparently weaker affinity of NF-E2 to bind 3′-NA(II) than 3′-NA (Fig. 4A) was further exploited by competitive EMSA. A labeled 3′-NA oligonucleotide was mixed with increasing concentrations of unlabeled 3′-NA or 3′-NA(II) oligonucleotide prior to the addition of K562 whole cell extract. Protein binding to the labeled 3′-NA oligonucleotide was then quantified by EMSA (Fig. 7C and D). As shown in Fig. 7C, it takes less of the 3′-NA oligonucleotide to compete for binding of NF-E2 to the labeled 3′-NA oligonucleotide. Based on the molar excess of unlabeled oligonucleotides required to reduce protein binding to the labeled probe by 50%, the relative affinity of NF-E2 binding was estimated to be approximately twofold higher for 3′-NA than for 3′-NA(II) (Fig. 7C). With the same analysis, no difference in the relative binding affinities of AP1 to the two sequences could be detected (Fig. 7D). Varying the time (10 to 30 min) and temperature (room temperature and 15°C) of preincubation did not significantly change the twofold ratio of relative NF-E2 binding affinities to 3′-NA and 3′-NA(II) (data not shown).

Finally, we assessed the stabilities of NF-E2 or AP1 complexes formed on the 3′-NA or 3′-NA(II)-containing oligonucleotide, respectively (Fig. 7E and F). In these experiments, the relative off rates for the complexes were determined by measuring their apparent half-lives, i.e., the time at which 50% of the displaceable complex remained in the competition experiments. As shown in Fig. 7E, the apparent half-life of the NF-E2–3′-NA complex (12 min) was approximately twofold longer than that of the NF-E2–3′-NA(II) complex (6.5 min). This finding is consistent with the conclusion drawn from Fig. 7C. Again, we observed only a small difference for the apparent half-lives of the two different AP1-DNA complexes (Fig. 7F).

The GC→TA mutation disrupts MafK binding in vitro and in vivo.

The effect of the GC→TA mutation on the binding of MafK to 3′-NA motif was also analyzed by EMSA (Fig. 8). Since little MafK complex forms in K562 extract (Fig. 4A), we first used crude extract from E. coli cells expressing a GST-MafK fusion protein. As shown in Fig. 8A, GST-MafK, presumably a homodimer (25), binds 3′-NA but not 3′-NA(II). A similar conclusion was reached with the use of whole cell extract from COS-7 cells transiently transfected with MafK expression plasmid. In Fig. 8B, only the 3′-NA oligonucleotide exhibited an intense band in EMSA (lane 6), which could be removed by anti-MafK (lane 7) but not by preimmune serum (data not shown).

FIG. 8.

MafK binding in vitro to 3′-NA and 3′-NA(II). (A) EMSA of 3′-NA- or 3′-NA(II)-containing oligonucleotides in crude extracts of E. coli expressing GST (lanes 1 and 3) or GST-MafK fusion protein (lanes 2 and 4). (B) EMSA of oligonucleotides in cell extracts prepared from COS-7 cells transfected with pRSV vector (lanes 5 and 8) or with pRSV-MafK (lanes 6, 7, 9, and 10). For samples loaded in lanes 7 and 10, the extract was incubated with anti-p18 prior to use.

We also studied the functional consequences of the inhibition of MafK-binding to HS-40 with mutated 3′-NA(II) motif. For this, K562 cells were first transiently cotransfected with pHS40-α590GH plus pRSV-MafK. Coexpression of MafK apparently inhibited the HS-40 enhancer function on α-globin promoter activity (Table 2, footnote a). MafK-dependent repression of promoter activity via the NF-E2-binding site in the promoter has been shown in previous studies of NIH 3T3 (24) and QT6 (25) cells. Interestingly, however, MafK no longer repressed the α-globin promoter activity when the latter was linked in cis with HS-40 containing the mutant 3′-NA(II) motif (Table 2). The data in Fig. 8 and Table 2 together clearly demonstrate that only the wild-type 3′-NA site allows MafK binding and consequently its repression function in vivo. Furthermore the specific GC→TA mutation in 3′-NA(II) completely inhibits the binding of MafK.

NF-E2, but not AP1, is bound to HS-40 in vivo on HS-40 chromatin.

The in vitro binding studies of Fig. 4 and 6 to 8 indicate that 3′-NA(II) is refractory to MafK binding but still can interact with NF-E2 or AP1, although the NF-E2–DNA complex is destabilized. To further examine whether AP1 or NF-E2 is involved in the regulatory function of HS-40, we performed chromatin immunoprecipitation reactions (13, 42). The notion is that stable binding in vivo on chromatin at specific DNA motifs appears to be a prerequisite for the functioning of sequence-specific DNA-binding transcription factors.

To directly address whether NF-E2 and/or AP1 is bound in vivo at HS-40, we fixed the K562 and HeLa cells and cross-linked their chromatin before immunoprecipitation with anti-p45, anti-c-Jun, or preimmune serum. The precipitated chromatin DNAs were then purified and subjected to PCR analysis with specific DNA primers bracketing different genomic regions. The anti-c-Jun antibody specifically enriched DNA sequences of the c-Jun promoter in the chromatin immunoprecipitate from K562 cells (Fig. 9A, lane 4) and HeLa cells (data not shown). This result supports the scenario that transcription of c-Jun is autoregulated through stable binding of AP1 to one or both AP1-binding sites in the c-jun promoter (13). However, neither the actin promoter nor HS-40 was enriched (Fig. 9B and C, lanes 4). In contrast, anti-p45 (or anti-NF-E2) specifically precipitated chromatin fragments containing the HS-40 sequence (Fig. 9B, lane 6) but not the c-jun promoter (Fig. 9A, lane 6), the actin promoter (Fig. 9C, lane 6), the keratin gene, and a β-globin intergene sequence (13). Use of preimmune serum enriched neither c-Jun nor the HS-40 region in the chromatin precipitates (lanes 5).

FIG. 9.

PCR analysis of the c-jun promoter, HS-40 region, and actin gene region in chromatin immunoprecipitates. PCRs were carried out with different DNA samples as the templates and analyzed by ethidium bromide-agarose gel electrophoresis. Lanes 1 to 3, serial dilutions of preimmune supernatant DNA from the immunoprecipitation. Numbers above the lanes indicate fold dilution. Lanes 4 to 6, anti-c-Jun, preimmune, and anti-NF-E2 precipitate (ppt).

DISCUSSION

In this study, we analyzed the contributions of four individual nuclear factor-binding motifs to the assembly of the erythroid-specific enhanceosome HS-40 and investigated the identities of nuclear factors that compete for binding at the 3′-NA motif and regulate the developmental function of HS-40. Since the same set of factor-binding motifs, NA, GT, and GATA-1, are also responsible for the functions of the four HS sites within the β-globin LCR, our data provide comprehensive information regarding of enhanceosome assembly of the human globin gene switch system. However, it should be noted that our discussion below of the regulatory model in Fig. 10 is based mainly on factor-DNA binding studies in vitro and in cell lines, and one should be cautious about extrapolating the findings to the in vivo situation.

FIG. 10.

Models of developmental regulation of the human α-globin locus by HS-40 enhanceosome. (A and B) The switch from ζ- to α-globin gene expression involves changes of the composition and conformation of the HS-40 enhanceosome. It is proposed that the specific GC→TA mutation in 3′-NA motif reverses the structure and function of HS-40-A to HS-40-E (see text for more details). (C) Three alternative schemes of competitive factor binding at the 3′-NA motif of HS-40.The data supporting and refuting each scheme are discussed in the text.

Functional and structural synergism of factor-binding motifs within HS-40.

Individual nuclear factor-DNA complexes within HS-40 together contribute to the architecture and functions of the enhanceosome. Consistent with previous transient transfection studies (47, 61), mutations in each of the four HS-40 motifs all significantly lowered the activity of the cis-linked α-globin promoter, by 50 to 90%, in stably integrated K562 cells (Table 1). This functional synergism most likely results from the cooperative contributions of these motifs in the assembly of a wild-type HS-40 enhanceosome. However, this cooperativity is not completely reflected by the genomic footprint analysis (Fig. 3). For instance, the HS-40 footprints of the constructs with the 5′-NA, 3′-NA(II), or GATA-1(c) mutation are not significantly different from the wild type. It should be noted, however, that different factors could bind to the same sequence motif and yet exhibit similar genomic footprints (see below). Furthermore, other architectural alterations, as the result of a mutation, could occur in the outer layer of the multiple protein-DNA complex and not affect the genomic footprints.

Two mutations, 3′-NA(I) and GT (Fig. 2D and F), nearly abolished all footprints in the HS-40 element. The HS-40 region with the GT mutation even becomes free of factor occupancy. While the data could be explained by the stabilization of factor binding on the other HS-40 motifs through their physical interaction with the factor(s) bound at the 3′-NA or GT motif, the possibility of involvement of chromatin structure is intriguing. It is particularly interesting that two factors capable of recognizing these two HS-40 motifs, NF-E2 and Sp1 (26), could remodel the chromatin structure near the vicinities of their binding sites (4, 18, 34). In summary, the data in Fig. 2 and Table 1 together indicate that perturbation or disruption of just a single nuclear factor-DNA complex could greatly alter the architecture of HS-40 and consequently its function.

Competitive factor binding at the 3′-NA motif of HS-40: a refined model of human ζ-globin gene regulation.

Among the three mutations that did not significantly alter the HS-40 footprints in vivo, 3′-NA(II) (Fig. 2E) is particularly noteworthy. As already mentioned, this mutation confers HS-40 the capability to efficiently reactivate the silenced human ζ-globin promoter at the adult stage. A model was then proposed for the developmental regulation of the human ζ-globin gene expression (23). It stated that competitive factor binding at the 3′-NA motif of the HS-40 modulates the stage specificity of the enhancer function. In particular, dominant binding of HS-40 in adult erythroid cells by NF-E2 at the 3′-NA motif allows the formation of a specific conformation of the HS-40 enhanceosome. This complex (HS-40-A [Fig. 10A and B) could efficiently interact with and activate the α-globin promoter but not that of ζ-globin (passive repression). In the embryonic erythroid environment, on the other hand, NF-E2 binding to 3′-NA is somehow limited or inhibited, and binding of another factor such as AP1 to this motif converts HS-40 to a conformationally different enhancer, HS-40-E, preferentially activating the ζ-globin promoter (Fig. 10A and B).

The new findings documented in Fig. 4 and 6 to 9) have indicated the need to modify the above model. First, disturbance of exchange of DNA-binding factors at a single motif could drastically modify the configuration of the HS-40 enhanceosome (Fig. 1 to 3). Second, the GC→TA substitution of the 3′-NA motif abolishes binding of small MafK homodimers but not that of NF-E2 or AP1 (Fig. 4 and 6 to 8). The same GC bp has been suggested to be essential for binding of other small Maf homodimers such as MafG (8). In fact, the same GC→TA mutation also abolished binding of MafG (unpublished results). However, the stability of the NF-E2–DNA complex is affected (Fig. 7), suggesting that its conformation, and very likely that of the HS-40 enhanceosome as well, is different from the wild type. Finally, in contrast to NF-E2, very little, if any, AP1 binding at the HS-40 could be detected in vivo (Fig. 9).

It should be noted that our observed effects of the GC→TA mutations on the binding of NF-E2 (Fig. 4, 6 and 7) are in contrast to several previous studies. For example, Ney et al. (41) showed that the inducibility of HS-2 function by hemin depends on binding of NF-E2. In contrast, a mutant HS-2 enhancer containing the GC→TA mutations, one each in the tandem NF-E2/AP1 sites (HS-2mut [Fig. 5]), was not inducible by hemin. Furthermore, NF-E2 could not bind to HS-2mut, as demonstrated by EMSA or competitive EMSA in K562 or MEL extract. A similar observation was made for the NF-E2/AP1 site at −160 of the promoter of human porphobilinogen deaminase gene (PBDG) (37, 50) (Fig. 5). Also, by a series of competitive EMSA, Andrews et al. have derived a binding consensus of NF-E2 in which the GC base pair appears to be required for NF-E2 binding (2). On the other hand, competition studies using nuclear extracts demonstrated that the NF-E2/AP1 motif with the mutated GC bp could still compete for NF-E2 binding against the wild-type motif (33, 55). The molecular basis for the different findings regarding the effect of the GC→TA mutation on NF-E2 factor binding is not clear, but likely nucleotide sequences flanking the NF-E2-binding consensus (boxed in Fig. 5) could affect the stability of the NF-E2–DNA complex.

In any case, based on the above, we have formulated a refined model in Fig. 10. In the model, there still exist two configurationally and functionally different complexes of HS-40, HS-40-E, and HS-40-A (Fig. 10A). Of the two, HS-40-E forms mainly in embryonic erythroid cells. It recognized and activated the ζ-globin promoter efficiently but could not do so with the α-globin promoter. On the other hand, HS-40-A dominates in adult erythroid cells, activates the α-globin promoter, and cannot overcome the negative silencing mechanisms (56, 62) operating on the ζ-globin gene (Fig. 10B). One of the key differences between HS-40-E and HS-40-A is the nuclear factor(s) bound at their 3′-NA motifs.

Three alternative scenarios, in which the binding of nuclear factor(s) at 3′-NA changes as development proceeds, are outlined in Fig. 10C. Thus far, the nuclear factors capable of recognizing the 3′-NA sequence include AP1, NF-E2, and small Maf homodimers, as already described, plus Nrf1, Nrf2, Nrf3, Bach1, and Bach2 (reviewed in references 7, 30, and 39). From data of this study, the direct involvement of AP1 at this motif, as hypothesized previously (23) (Fig. 10C, top) can be ruled out. In the second scheme (Fig. 10C, middle), binding by MafK, or one of the other small Maf homodimers, at the 3′-NA leads to the formation of HS-40-A, while it is replaced in the embryonic erythroid cells by either NF-E2 or one of the other NF-E2/AP1 motif-binding factors mentioned above. This scheme, while supported by the data in Fig. 8 and Table 2, is somewhat unexpected from two observations. One, in all studies of their function (24, 25, 28) (Table 2), the small Maf homodimers mainly act as transcriptional repressors of either a promoter or an enhancer. More importantly, coexpression of MafK inhibited but did not increase, as expected from the second scheme of Fig. 10C, the α-globin promoter activity in pHS-40-α590GH (Table 2, footnote a). It should be noted, though, that only MafK has been tested in this cotransfection study. The involvement of the other small Maf homodimers in the model cannot be ruled out at this time. Furthermore, the switch scheme may operate only through developmental programming under physiological conditions (23) but not in cell lines.

The third and final scheme (Fig. 10C, bottom) posits that the formation of HS-40-E and HS-40-A is regulated by binding of NF-E2 in competition against a nuclear factor other than AP1 or the small Maf homodimers. The NF-E2 binding is predominant in adult erythroid cells but is outcompeted by Nrf1, Nrf2, Nrf3, Bach1, Bach2, or a still unidentified factor (Fig. 10C, bottom panel). This scheme is especially supported by the fact that the GC→TA mutation renders the NF-E2–3′-NA complex relatively unstable and that NF-E2 has been shown to be required for globin gene expression in adult mouse erythroid cells (31). Some speculation could also be made regarding the nuclear factors competing against NF-E2 for binding at the 3′-NA motif in embryos. Mice with their Nrf1/LCRF1, Nrf2, or p45/NF-E2 gene knocked out all have normal globin expression patterns at the embryonic stage (10, 15, 32, 49), thus arguing against the involvement of Nrfl/LCRF1 or Nrf2 in this last competition scheme in Fig. 10C. Finally, the developmental signals leading to the exchange of nuclear factors bound at the 3′-NA motif of HS-40 are unspecified in our model. They could simply be changes of the relative concentrations of different nuclear factors, or they may be regulated by the differential binding of factor(s) at another DNA motif of HS-40 or at the outer layer of the HS-40 enhanceosome. All of these possibilities await further investigation.