INTRODUCTION

In the human -globin locus, the five functional genes are arranged in the order in which they are expressed during ontogeny of erythroid cells: 5--^G-^A---3. The expression of the human -like globin genes is restricted to erythroid tissues and regulated in a developmental stage-specific manner (see Refs. 1 and 2 for review). A cluster of DNase I hypersensitive sites (HSs)¹ has been mapped between 6 and 18 kb upstream of the -globin gene (3, 4, 5). This region has been designated the locus control region (LCR) (6). The LCR activates the -globin locus chromatin; it serves as a powerful enhancer of the -globin genes, insulates the -globin locus from the effects of the surrounding chromatin, and confers copy number-dependent, integration site-independent, and high level expression of linked globin genes in erythroid cells of transgenic mice (7).

Genetic and molecular studies have established that transcription of globin genes is regulated through protein-DNA and protein-protein interactions that take place in proximal and distant regulatory sequences of the -globin locus (1, 2, 8). A variety of cis-acting elements lying in the LCR or globin gene promoters or other flanking sequences have been identified by a great number of studies using in vitro DNA binding assays or DNA transfection assays and have been implicated in the regulation of transcription of -globin genes (9, 10, 11, 12). In vivo dimethyl sulfate (DMS) footprinting analysis of the human  globin gene locus provides an alternative method to identify those interactions between cis-acting DNA sequences and trans-acting factors that take place in living cells and is perhaps more relevant to the in vivo regulation of globin genes. Through such studies, several in vivo footprints that are erythroid-specific and change with the induction of globin gene expression have been identified in the LCR, as well as in the promoter regions of globin genes (13, 14, 15).

The study described in this paper has been undertaken to examine whether in vivo footprints of the human -globin locus differ between erythroid cells expressing fetal globin and those expressing adult globin. As a source of erythroid cells expressing the fetal or the adult globin gene program, we utilized human-murine erythroleukemia (MEL) hybrid lines, which were generated by fusion of human fetal liver erythroid cells with MEL cells (16). These hybrids initially express a typical fetal hemoglobin program consisting of exclusive (or very high) -globin gene expression and absent or low -globin gene expression. As time advances, a - to -globin gene switch occurs and is completed in 20 to 40 weeks. These hybrids, therefore, provide an ex vivo model for studying hemoglobin switching. The hybrids retain the unlimited proliferative capacity of MEL cells, thus allowing the accumulation of large numbers of cells required for molecular analysis. Therefore, this system provides excellent opportunities for studies of in vivo footprinting. We analyzed the sequences of the human -globin locus including four DNase I hypersensitive sites of the LCR (5HS-1 to 5HS-4), the promoter regions of the - and -globin genes, and the enhancers located 3 to the ^A- and -globin genes (17, 18). We find that several patterns of in vivo footprints within the human -globin gene locus change depending on which globin genes are expressed in the erythroid cells.

EXPERIMENTAL PROCEDURES

Human-MEL hybrid lines, A181 and A181, were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (Hyclone) and 100 units/ml penicillin and 100 µg/ml streptomycin. For induction, cells were grown for 4 days in the medium containing 4 mM hexamethylenebisacetamide (HMBA) as described previously (16).

Total RNA was extracted from cells by the method described previously (19). Ten micrograms of total RNA were subjected to primer extension analysis as described (20). Oligonucleotides used as primers were 5-end-labeled with T4 polynucleotide kinase and purified through a Push column (Stratagene). RNAs were annealed with 5 × 10⁵ cpm of labeled oligonucleotides in 30 µl of hybridization buffer (80% formamide/40 mM Pipes, pH 6.4, 1 mM EDTA, pH 8.0, 0.4 M NaCl) at 30 °C. Extension reaction was performed with 100 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) at 42 °C for 2 h in 20 µl of reverse transcriptase buffer (50 mM Tris-HCl, pH 7.6, 60 mM KCl, 10 mM MgCl₂, 1 mM dNTP, 1 mM dithiothreitol, 1 unit ml¹ RNasin, 50 µg ml¹ actinomycin D, 0.1% Triton X-100). The extension products were analyzed on 8% urea-polyacrylamide gels, and the gels were exposed to Kodak X-AR film overnight with an intensifying screen. The DNA sequences of oligonucleotides used as primers were as follows: 5-GCCTCTTCCACATTCATCTTGCTCCACAGGCTAGTGACG-3 (39-mer) for -globin mRNA; 5-TCACCTTGCCCCACAGGCTTGTGATAGTAGCCTTGTTCC-3 (38-mer) for -globin mRNA; 5-GCCCCACAGGGCAGTAACGGCAGACTTCTCCTC AGG-3 (36-mer) for -globin mRNA.

For in vivo methylation of cells, the human-MEL hybrid cells were treated with 0.1%(v/v) DMS for 3 min at room temperature as described previously (14). The methylation reaction was quenched by adding 20 ml of 1 × phosphate-buffered saline containing 1 M -mercaptoethanol. Extraction of genomic DNAs from in vivo methylated cells was performed as described (20). In vitro methylation of naked DNA was carried out by the method of Maxam and Gilbert (21). In vivo and in vitro methylated DNAs were cleaved by piperidine as described (21). Residual piperidine remaining in DNA preparations was removed as described (14).

LM-PCR was performed by the method described by Mueller and Wold (22) with modifications. Briefly, 3 to 7 µg of DNA were used for first-strand synthesis. The reaction was performed in a solution of 20 mM MgCl₂, 20 mM dithiothreitol, 60 µM dNTP by adding 1.5 µl of 1:4 diluted Sequenase version 2.0 (United States Biochemical Corp.) with 1 × TE (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and incubating at 50 °C for 10 min. Ligation of the PCR common linker was carried out as described (22). DNA sequences were amplified by performing 16 to 18 cycles of PCR as described (14). The amplified products were then digested with mung bean nuclease (New England Biolabs) as described (14). Footprint ladders were visualized by performing primer extension using 5-end-labeled primer 3 (see below) as described (22). All of the footprints were analyzed two or more times to ensure reproducibility. The DNA sequences of primers used in LM-PCR of this study are shown in our previous study (14), and, in addition, the following primers were used (primers 1, 2, and 3 were used for first-strand synthesis, PCR, and primer extension, respectively). 5HS-1 (coding) primer 1, 5-GCCTCATTCAGACACTAGTGTC-3; primer 2, 5-AGTGTCACCAGTCTCCTCATATACC-3; primer 3, 5-TTTAGCCATGTTTTCTGGGAGCTTAGGGGC-3; 5HS-1 (noncoding) primer 1, 5-GCAACTAACTCATGCAGGACT-3; primer 2, 5-CTCTCAAACACTAACCTATGACCTTTTCTATG-3; primer 3, 5-GCCTTTTCTATGTATCTACTTGTGTAGAAACC-3; 5HS-4 (coding) primer 1, 5-GCCAGGGCTGGGCAGTCTCC-3; primer 2, 5-TCCTGTTATTTCTTTTAAATAAATATATCATTTAAATGC-3; primer 3, 5-AAATGCATAATAAGCAAACCCTGCTCGGG-3; 5HS-4 (noncoding) primer 1, 5-AAGAATAGCCATGACCTGAG-3; primer 2, 5-GAGTTTATAGACAATGAGCCCTTTTCTCTC-3; primer 3, 5-ACAATGAGCCCTTTTCTCTCTCCCACTC-3.

In order to verify the reproducibility of in vivo footprints and to determine the percent protection on G residues, the gels were scanned by PhosphorImager (Molecular Dynamics). The percent protection on each G residue was determined relative to the radioactivity of the corresponding G residue of the naked DNA lane (see Fig. 1). Using a G residue as a control that has no protection (the G residue B in Fig. 1), the percent protection on the G residue A was calculated as follows: % protection = [1  a × d/b × c] × 100 (a, b, c, and d represent the radioactivities of the G residues A, B, C, and D, respectively; see Fig. 1). G residues that were protected 40% or more, which were clearly appreciable on autoradiograms, were classified as those that were strongly footprinted. G residues having 20 to 40% protection were considered as those that were weakly footprinted. Protection less than 20% was not noticeable on autoradiograms. Hyper-reactive G residues, which are often seen around the margin of protein-DNA interactions as described (23), were defined as those that showed 20% or more stronger signals than the corresponding G residues of the naked DNA lane. Throughout this analysis, hyper-reactivity of A residues was noted in some regions, as found in other studies (24, 25).

RESULTS

Analysis of -Globin Gene Expression in Human-MEL Hybrids

A hybrid produced by fusing erythroblasts from the fetal liver of a 60-day-old fetus with MEL cells (hybrid A181) was used. Live cells from this hybrid have been stored at the time when the hybrid was expressing exclusively -globin soon after fusion, and at the time when the hybrid had totally switched to -globin expression several months following fusion. The pre-switch cells were expanded, which provided the population of -globin-expressing cells, and were used for our analysis (hybrid A181). Similarly, expansion of the post-switch cells provided the population of only -globin-expressing cells (hybrid A181). Globin gene expression was analyzed in the expanded cells by primer extension. As shown in Fig. 2, the A181 cells expressed 100% human -globin (Fig. 2, lane 3); conversely, the A181 cells expressed 100% -globin (Fig. 2, lane 4). These results indicate that the A181 and A181 cells correctly express the fetal or adult globin gene transcription program and can be used as surrogates for the primary erythroid cells of the fetal and the adult human developmental stages.

Expression of the human -globin genes in erythroid cells.

5HS-1 of the LCR

A DNA sequence corresponding to 5HS-1 was originally mapped to a region 6.1 kb upstream of the -globin gene (3, 4, 5). In this study, the 707-bp NcoI/HindIII DNA fragment from 5HS-1 (26) was subjected to in vivo DMS footprinting analysis.

In vivo footprints were observed in several regions whose sequences are compatible with the GATA-1, GT, and AP-1 consensus motifs (27, 28, 29) (Fig. 3C). There are two GATA-1 motifs in this DNA fragment, which are designated GATA-1A (5) and GATA-1B (3). Although the GATA-1A motif was footprinted in both A181 and A181 cells, the GATA-1B motif, which is located 38 bp downstream of GATA-1A, was not footprinted in either of the hybrids. The GT-1 motif, located 5 to GATA-1A, was footprinted in both hybrids, but it is interesting that the A181 cells revealed patterns of footprints on this motif different from those seen for the A181 cells; the analysis of gels by PhosphorImager confirmed these differences (see Fig. 3, B and C). Additional footprints lying in the most 5 and 3 regions of 5HS-1 were observed over DNA sequences compatible with an AP-1 consensus motif. Both of these AP-1 motifs were footprinted in both the A181 and A181 cells.

footprinting analysis of 5HS-1.

5HS-2 of the LCR

We analyzed the 732-bp HindIII/BglII fragment to investigate the in vivo protein-DNA interactions of 5HS-2, as performed previously (14).

As shown in Fig. 4, eight major footprints were found upon analysis of 5HS-2 in both hybrids and designated in a 5 to 3 orientation, GT-1, GT-2, NF-E2/AP-1, GT-3, GATA-1, GT-4, GT-5, and AP-1 motifs. Both hybrids, A181 and A181, have demonstrated extensive protection on the NF-E2/AP-1 motif. The GATA-1 motif, located approximately 60 bp downstream of the NF-E2/AP-1 motif, was not footprinted in either of the hybrids (Fig. 4, A and C).

footprinting analysis of 5HS-2.

Differences in in vivo footprints of - versus -globin-expressing hybrids (A181 versus A181) were seen at virtually all of the GT motifs of 5HS-2. At the GT-1 motif, both A181 and A181 cells showed similar protection patterns, but an additional G residue of the noncoding strand was footprinted in the -globin-expressing A181 cells (see Fig. 4, Ab, B, and C). The A181 cells displayed substantial footprints at the GT-2 motif, while only limited footprints were seen in the A181 cells (see Fig. 4B for PhosphorImager analysis). Moreover, both the GT-3 and GT-4 motifs were not footprinted in the A181 cells but were in the A181 cells. This is consistent with a recent study showing that human adult erythroblasts isolated from bone marrow of patients with sickle cell anemia exhibited in vivo footprints over the GT-3 motif (30). A prominent hyper-reactive G residue accompanied the footprints of the GT-5 motif in the A181 cells, but the corresponding GT-5 footprints in the A181 cells were completely devoid of any such hyper-reactive G residue. This study has demonstrated that the A181 cells have several minor footprints, for instance, between the GT-2 and NF-E2/AP-1 motifs, which were not observed in the A181 cells (see Fig. 4C).

5HS-3 of the LCR

5HS-3 is characterized by an array of cis-acting elements that are apparently distinct from those constituting 5HS-2 (10, 31). The putative cis-acting elements that were examined in this study include one NF-E2, four GATA-1 (GATA-1A to GATA-1D), and seven GT motifs (GT-1 to GT-7) (see Fig. 5C).

footprinting analysis of 5HS-3.

The NF-E2 motif, located at the 5 end of 5HS-3, showed virtually no protein-DNA interactions in either of the hybrids (Fig. 5, A and C), suggesting a minor role, if any, of the NF-E2 motif in the 5HS-3 function. This result is consistent with prior functional studies (10) and with our preliminary data that erythroid cells isolated from both fetal livers and bone marrow of patients with sickle cell anemia failed to exhibit visible footprints over the NF-E2 motif.² Among the four GATA-1 motifs in 5HS-3, we detected discernible footprints in both hybrid lines at two GATA-1 sites, GATA-1C and -D, which are located at the 3 half of 5HS-3.

Extensive footprints were observed over several GT motifs of 5HS-3. Tandemly reiterated GT motifs, GT-1 and GT-2, which may account for the activity of 5HS-3 conferring position-independent expression on a linked globin gene (31), were footprinted in both the - and -globin expressing hybrids; however, there were distinct differences in the details of the footprint patterns between these two lines (see Fig. 5, A and C). The GT-6 motif, which is located in the middle of 5HS-3, demonstrated more extensive footprints in the A181 cells than in the A181 cells (see Fig. 5B for PhosphorImager analysis). Similar patterns of footprints were obtained with the most 3 GT motif GT-7. Thus, the A181 cells consistently showed more extensive footprints over these GT motifs (GT-1, GT-2, GT-6, and GT-7) than did the A181 cells.

5HS-4 of the LCR

The enhancer sequence that directs integration site-independent and copy number-dependent expression in transgenic mice carrying 5HS-4 has been localized to a 280-bp SstI/AvaI fragment of 5HS-4 (32, 33). We analyzed this DNA fragment to examine the in vivo footprints of 5HS-4.

The NF-E2 motif, located at the 5 end of 5HS-4, showed detectable footprints in both hybrids. The most distinctive cis-acting elements comprising 5HS-4 are two adjacent GATA-1 motifs, GATA-1A and GATA-1B, which are present in the middle of 5HS-4 (Fig. 6, A and C). These two GATA-1 motifs were footprinted in vivo in both hybrid lines. These results indicate that binding of the GATA-1 and NF-E2 proteins to their cognate sites is not regulated in a manner associated with the developmental changes in globin gene expression, an observation that appears to be common to all 5HSs of the LCR. Although the GT-1 motif revealed identical footprints for both the A181 and A181 cells, footprints whose pattern varied with the globin genes expressed were again observed over the GT-2 and GT-3 motifs; the A181 cells exhibited more extensive footprints over the GT-2 and GT-3 motifs than did the A181 cells (see Fig. 6B for PhosphorImager analysis). These protein-DNA interactions correlating with adult globin expression may be involved in the preferential activation of the adult -globin gene by 5HS-4 that has recently been demonstrated in studies of transgenic mice (34). In addition, at the AP-1 motif and its downstream sequences, which are located 3 of 5HS-4, extensive footprints were obtained in the A181 cells, while the A181 cells showed a hyper-reactive G residue only (see Fig. 6B for PhosphorImager analysis). The protein-DNA interactions at this region might also be involved in the adult stage-specific enhancer activity of 5HS-4.

footprinting analysis of 5HS-4.

^A3 Enhancer and 3 Enhancer

The ^A3 enhancer has originally been mapped to a 750-bp DNA fragment which is located 400 bp downstream of the poly(A) site of the ^A-globin gene on the basis of transient expression studies (17). Protein-DNA interactions in this enhancer were previously studied in vitro, and eight in vitro footprints (FP-I to FP-VIII) were demonstrated (36). However, with the exception of one region upstream of FP-V, we were unable to detect substantial in vivo footprints at any of the regions for these in vitro footprints; for example, there are no in vivo footprints over the sequence of FP-IV (see Fig. 7, A and C). Over several G residues upstream of the GATA-1 motif within FP-V, we detected limited footprints consisting of weak protection with hyper-reactive G residues in the A181 cells, but not in the A181 cells.

footprinting analysis of the 3 enhancers of the

- and -globin genes.

A well-documented -globin gene enhancer is located 2.2 kb downstream of the -globin gene promoter (37). Conventional in vitro DNA binding assays demonstrated four putative binding motifs for GATA-1 (38). The human-MEL hybrids showed visible footprints on two GATA-1 motifs, GATA-1A and GATA1-B, but did not exhibit detectable footprints over the two GATA-1 sites further 3 in the enhancer, GATA-1C and GATA-1D (Fig. 7B; data not shown for GATA-1D; see Fig. 7D for summary). Interestingly, the A181 cells exhibited more extensive footprints on the GATA-1A site than did the A181 cells. Minor footprints were also observed at a region upstream of GATA-1C (Fig. 7, B and D).

Promoter Regions of the - and -Globin Genes

Previous studies have shown that in promoter regions of both the - and -globin genes, a DNA fragment extending up to 200 bp relative to the transcription start site is sufficient to direct efficient transcription of globin genes (39, 40, 41, 42). We therefore focused on this region of the respective promoter for in vivo footprinting analysis.

In the -globin promoter, the fetal globin-expressing cells (A181) demonstrated significant footprints over the CACCC and two CCAAT boxes, as well as in the 3-flanking region of the TATA box (Fig. 8), the patterns of which were similar to those observed with K562 cells (14). In contrast, the adult globin-expressing cells (A181) did not show substantial footprints to the CACCC, CCAAT-1, and CCAAT-2 boxes of the -promoter, but revealed limited footprints over a region downstream of the CCAAT-2 box. This might suggest the binding of transcriptional silencers to that region. Thus, protein binding to both the CACCC and CCAAT boxes is controlled in a manner dependent on developmental expression of the globin genes. In both the A181 and the A181 cells, extensive footprints and hyper-reactive G residues were seen at a region around position 175 , where the GATA-1A, the octamer, and GATA-1B sequences are located (43) (Fig. 8). This region is of interest because base pair substitutions at positions 175, 192, 196, and 198 result in continued fetal hemoglobin production in the adult (2), suggesting that transcription factors binding to this region contribute positively or negatively to the regulation for -globin gene expression.

footprinting analysis of the -globin gene promoter.

In the -globin promoter, a DNA fragment which includes the TATA (30), CCAAT (70), and proximal CACCC-2 (90) boxes, but not the distal CACCC-1 box (110), proved to be capable of directing efficient and inducible transcription of the -globin gene (39). Consistent with this observation, we detected significant in vivo footprints over the CCAAT, CACCC-2, and TATA boxes in the A181 cells, while the A181 cells showed no footprints on either the CCAAT or CACCC-2 box (see Fig. 9B for PhosphorImager analysis). It is intriguing that not only the A181 but also A181 cells showed strong protection on a G residue immediately 3 to the TATA box (Fig. 9A). The functional importance of the CACCC-2 box is suggested by a clinical observation that mutations at positions 87 and 88 within the CACCC-2 box cause -thalassemia (44) and the recent finding that EKLF, a Krüppel-like transcription factor, binds specifically to the -globin gene CACCC box (45, 46, 47). The -globin promoter contains two tandem GATA-1 motifs located in the region around 200, but no footprints were obtained on these motifs. Detectable footprints were also observed at a region from 114 to 133 in both the A181 and A181 cells.

footprinting analysis of the -globin gene promoter.

DISCUSSION

The developmental switches of human globin gene expression presumably accompany alterations in the interplay of multiple cis-acting elements with several different transcription factors (1, 2, 8). To obtain insights into the sequences involved in these interactions, we performed a comprehensive analysis of the in vivo interactions of sequence-specific transcription factors of the human -globin gene locus. Our study provides several insights into the type of interactions that occurs within the human -globin locus during the development of erythroid cells.

Since it is generally thought that the individual 5HSs of the LCR appear to function synergistically, rather than independently, by interacting with each other to achieve full enhancement of globin gene expression (48, 49, 50), we examined the in vivo footprints of all 5HSs of the LCR; in vivo footprints of 5HS-2 and 5HS-3 have also been reported before (13, 14, 15). As described earlier (8), we also found that binding of transcription factors at every 5HS of the LCR has occurred at three kinds of cis-acting DNA elements, NF-E2 (or AP-1), GATA-1, and GT.

With the exception of AP-1 of 5HS-4, in vivo footprints at one or two NF-E2 or AP-1 motif(s) of the 5HSs of the LCR did not differ between cells expressing fetal globin and those expressing adult globin. This may suggest that transcription factors are stably bound to NF-E2 or AP-1 motifs of the LCR during erythroid development, regardless of which globin genes are expressed. If this is the case, transcription factors binding to NF-E2 or AP-1 motif may not play a crucial role in the switch of globin gene expression, but rather serve in regulating the function of the LCR or the interactions between the individual 5HSs or between the holocomplex of the LCR and the individual globin genes.

Although the most extensive footprints were detected on the NF-E2/AP-1 motif of 5HS-2, DNA sequences compatible with the NF-E2 or AP-1 consensus motif have also been demonstrated in vivo in 5HS-1, 5HS-3, and 5HS-4 (15) (see Figs. 3, 5, and 6). However, none of these sequences showed in vivo footprints as extensively as did the NF-E2/AP-1 motif of 5HS-2. These observations suggest that, in view of a new consensus motif described for murine NF-E2 (51), DNA sequences of the NF-E2 or AP-1 motifs that are present in 5HSs other than 5HS-2 may be more compatible with the consensus for AP-1 than for NF-E2. Binding of the NF-E2 protein complex to the NF-E2/AP-1 motif of 5HS-2 might give rise to extensive footprints, whereas less conspicuous footprints on other NF-E2 (or AP-1) motifs might result from binding of the heterodimer of transcription factors other than NF-E2, such as AP-1, Nrf1 (52), or Nrf2 (53), which are ubiquitously expressed and highly homologous to the NF-E2 protein. The possibility that various heterodimers between NF-E2, Maf, and AP-1 are bound to NF-E2 motifs has also been postulated by a recent study (54).

The extensive footprints over the NF-E2 motif of 5HS-2 in both hybrid lines may suggest that NF-E2 plays an important role in globin gene expression in the context of human-MEL hybrids. A recent study with MEL cells has reached the same conclusion (55). However, globin gene expression is not affected in mice lacking the NF-E2 gene (56). Fiering et al. (57) have recently shown that deletion of 5HS-2 from the murine -globin LCR has a subtle effect on -globin gene expression. Thus, whether NF-E2 is required for globin gene expression or not might differ depending on erythroid environments, erythroleukemic, or physiologic. Further investigation is necessary to define the role of NF-E2 in globin gene expression.

In vivo footprints over GATA-1 motifs of the LCR have also turned out to be stable, in the sense that the footprints on this type of element do not appear to be influenced by the type of globin gene expressed. This study has disclosed interesting findings regarding footprints on GATA-1 motifs of the LCR. First, although detectable footprints have been shown previously on the GATA-1 motif of 5HS-2 in K562 (13, 14), adult erythroblasts (30), and Hu11 (15), this motif was not footprinted in either of the fetal or the adult globin-expressing hybrids used in the present study. These results may suggest that protein occupancy on the GATA-1 motif of 5HS-2 varies depending on the cellular characteristics of erythroid cells. Alternatively, binding of the GATA-1 protein may not be essential for the enhancer function of 5HS-2. The latter possibility has been suggested in a recent study showing that mutations introduced into the GATA-1 motif of 5HS-2 does not alter the levels of expression of a linked globin gene in K562 cells (58). Second, when multiple GATA-1 motifs are included in a regulatory sequence, as in the case of 5HS-1, 5HS-3, and  3-enhancer, only some, not all, of the GATA-1 sites showed detectable footprints. This observation suggests that only a limited number of GATA-1 motifs bind the GATA-1 protein and contribute to the regulation of the globin expression. Functional heterogeneity of the redundant GATA-1 motifs has also been demonstrated in the  gene promoter (59); binding of GATA-1 to the 208  GATA motif results in suppression of -globin gene expression, while binding at the 163  GATA motif contributes to -globin gene activation.

The most extensive differential footprints between the fetal and the adult globin-expressing hybrids were observed on GT motifs of the -globin locus LCR. It is of particular interest that the DNA sequence of the GT-6 motif of 5HS-3, ACTCCACCCA, is also included in the -globin promoters as a CACCC box, and that this sequence is differentially footprinted between the fetal (A181) and the adult globin-expressing cells (A181) (see Figs. 5 and 8). Moreover, among the seven GT motifs of 5HS-3, only this decanucleotide sequence is completely conserved between human, mouse, and goat (60). These observations allow us to postulate that in the context of human fetal erythroid × MEL hybrids, the GT-6 motif is important for the 5HS-3 function. Perhaps some common trans-acting factors, not as yet identified, bind to the GT-6 motif of 5HS-3 as well as to the CACCC box of the -globin promoter, thereby modulating simultaneously the function of both 5HS-3 and -globin promoter in the fetal globin expressing cells (A181). Similarly, as suggested by Reddy et al. (30), in the adult stage of development, the LCR and the -promoter may communicate via the GT-3 motif of 5HS-2 and the CACCC-2 box of the -promoter, since the DNA sequences of both elements are identical and footprinted only in the adult globin-expressing cells (A181). Thus, the protein-DNA interactions on GT motifs might be tightly controlled in a manner associated with the developmental stages of erythroid cells, thereby playing an important role in regulating globin gene expression.

In Vivo Footprints of Globin Gene Promoters and 3 Regulatory Sequences

Our study demonstrated that the A181 cells, but not the A181 cells, exhibit detectable footprints in the ^A3 enhancer. A recent study with transgenic mice has shown that the ^A3 enhancer is necessary for providing position-independent expression to µLCR ^A transgene constructs, suggesting that the ^A3 enhancer as well as the -globin promoter are required for engaging the LCR.³ Moreover, this element was found to contain a matrix attachment region (61). The differential footprints of the ^A3 enhancer may indicate that protein binding in this 3 element in the adult stage of development inhibits the LCR/ gene engagement, thus contributing to -globin gene silencing. Support to this hypothesis is provided by the finding that a -thalassemia with increased expression of the -globin genes is due to a mutation in a DNA sequence that was found to be footprinted in our study (62).

The 3 enhancer of the -globin gene acts as a classical enhancer by conferring tissue- and adult stage-specific expression on linked globin genes in transgenic mice (37). Consistently, we detected more extensive footprints on one of the GATA-1 motifs of the 3 enhancer in the A181 cells than in the A181 cells. Thus, the regulatory sequences located 3 to the ^A- and -globin genes showed significant protein-DNA interactions that vary with the globin genes expressed.

We have shown that, when either the - or -globin gene is actively transcribed, there is virtually complete protein occupancy over the canonical cis-acting elements in the respective promoter region, including the CCAAT and CACCC boxes. Another striking finding on the in vivo footprints in promoter regions is that a G residue immediately 3 to the TATA box of the -globin gene promoter in the A181 cells was footprinted extensively, even though that gene is not expressed at all. This is consistent with our recent result that erythroblasts isolated from human fetal livers exhibit extensive footprints on the same G residue, but not on either the CCAAT or CACCC box, in the -globin gene promoter.⁴ These results may suggest that the -globin gene is rendered nonfunctional in the fetal stage by the absence of binding of transcription factors to the CCAAT and CACCC-2 boxes, even though the transcription initiation complex is bound to the TATA box; in other words, binding of the transcription initiation complex to the TATA box may not be sufficient to initiate transcription of the -globin gene. Alternatively, the TATA box of the A181 cells is bound by the transcription initiation complex capable of initiating transcription of the -globin gene, while in the A181 cells an inactive and repressor-like protein complex might be bound to the TATA box. We have already shown that K562 cells, in spite of having a phenotype similar to that of the A181 cells, neither have protein binding on the TATA box nor undergo hemoglobin switching (14). In contrast, the A181 cells reveal substantial protein binding to the TATA box and do switch from - to -globin. These observations suggest that there may be two steps involved in activating the -globin gene during the switch from fetal to adult stage. In the first step, the transcription initiation complex binds to the TATA box, but no or very low levels of gene transcription are initiated because of the complete lack of protein binding to the CCAAT and CACCC-2 boxes. In the second step, both the CCAAT and CACCC-2 boxes are bound by the respective protein factors, resulting in the efficient transcription of the -globin gene. The -globin gene in the fetal stage may be in a status of ``preactivation'' rather than an inactive status.

This study has shown that the most remarkable differences of in vivo footprints of the -globin locus between - and -globin-producing cells are observed at the promoter regions of the - and -globin genes. This suggests that the expression of the human -globin genes may be regulated primarily by protein-DNA interactions in promoter regions that are controlled in a stage-dependent manner. Similar conclusions have been reached in studies with transgenic mice which have shown that, in the absence of the LCR, the human - as well as -globin genes are expressed in the correct temporal patterns (63, 64). The LCR is indispensable for activating the domain of the -globin gene locus (7, 35) and plays a role in providing specificity in the developmental expression of globin genes (34, 65). Thus, it seems possible that in addition to promoters of the individual -globin genes, the LCR may also contribute to specifying which globin genes are expressed at a given time during development.

Our study demonstrates that in vivo footprints on GT motifs of the LCR vary depending on the globin genes expressed, but it remains unclear how protein-DNA interactions on GT motifs affect the function or conformation of the LCR. Characterization of transcription factors binding to GT motifs of the LCR should provide insights into the mechanisms by which the LCR communicates with the globin gene promoters and contributes to the developmental control of globin genes.