Identification of functional elements of the chicken epsilon-globin promoter involved in stage-specific interaction with the beta/epsilon enhancer.

Expression of the chicken globin genes is regulated in part by competition between the βA-globin and ε-globin promoters for the enhancer found between the genes. To understand the determinants of the enhancer-promoter interaction in stage-specific regulation, the functional elements of the embryonic chicken ε-globin promoter were characterized. In vitro assays demonstrated that: (a) the TATA motif at −30 bound GATA-1, (b) Sp1 bound to an element centered at −54, and (c) both Sp1 and another factor, designated CACCC (which appears related to erythroid Krüppel-like factor, EKLF) bound in the −120 to −128 region. The functions of these motifs were tested using transient expression in embryonic erythroid cells. In the absence of the enhancer, promoter point mutants showed that the TATA, Sp1, and CCAAT motifs (but not the CACCC motif) contributed to promoter activity. In contrast, in the presence of the enhancer, all four motifs (including the CACCC motif) contributed to transcription. Developmental regulation of the enhancer activity was observed, with enhancement decreasing sharply from 185-fold at 4 days (cells expressing ε-globin) to 16-fold at 10 days (when ε-globin is no longer expressed). Taken together, the data suggest that multiple transcription factors contribute to promoter-enhancer interaction and the developmental regulation of ε-globin expression, with EKLF-like factors having an especially important role. Regulation of stage specificity occurs at the level of enhancer/ε-promoter interaction, even in the absence of competition, and is not simply a property of the enhancer or promoter in isolation.

Vertebrates express different ␤-like globin genes at different stages of their development, a process termed gene switching (1,2). As a result of gene switching, organisms express hemoglobins whose physicochemical properties are adapted to the physiologic requirements of each developmental stage. In the chicken, the embryonic ⑀and -globin genes are expressed in primitive lineage erythroid cells at days 2-5 of development, while the adult ␤ A -and ␤ H -globins are produced in definitive lineage cells, beginning on embryonic day 5 (3).
The mechanisms responsible for developmental stage-specific expression of globin genes are beginning to be understood.
Interaction between the individual globin genes and distant regulatory elements is important for this process. Control by a series of upstream DNase I-hypersensitive sites, referred to as a locus control region (LCR), 1 was first identified in the human ␤-globin cluster (4). These hypersensitive sites are important for opening the globin cluster chromatin (5)(6)(7). The LCR is also needed for developmental regulation of the individual globin genes, whose promoters are thought to compete for interaction with the LCR (8 -11). The strength of a distant element's interaction with a particular promoter is determined by the distance between the element and the promoter (12), the order of the promoters within the cluster (13,14), and the developmental specificity intrinsic to the promoter and upstream site (15). One mechanistic model of LCR action is that the individual hypersensitive sites form simultaneous binary interactions with different globin promoters (16). Elegant evidence that the LCR functionally interacts with only one gene at a time is consistent with a model in which the LCR sites together form a single active complex (17).
Comparison of the chicken and human ␤-like globin clusters should yield important insights into gene cluster regulation. Unlike the human genes, the chicken ␤-globin genes (5Ј--␤ H -␤ A -⑀-3Ј) are not arranged in order of developmental expression. Also unlike the human cluster, a strong enhancer (the ␤/⑀ enhancer), with LCR-like properties, is located inside the cluster, between the ␤ A -and ⑀-globin genes (18 -20). Both the upstream sites and the ␤/⑀ enhancer contribute to the expression of all the genes in the cluster (21). Finally, developmental regulation of the ␤-globin cluster probably evolved independently in the two species (22,23).
Although the chickenand ⑀-globins are expressed simultaneously at similar levels and their proximal promoters are similar, the proposed mechanisms for their regulation are different. Transcription of -globin is largely dependent on a promoter cis-element that binds the GATA-1 transcription factor (24). The ⑀ promoter lacks this GATA site, and ⑀-globin expression is dependent on the ␤/⑀ enhancer located 2.4 kilobase pairs upstream. To understand more about stage-specific regulation, we have undertaken a detailed characterization of the ⑀-globin promoter and its interaction with the ␤/⑀ enhancer at different developmental stages.
Preparation of Nuclear Extracts-Nuclear extracts (26) from embryonic chicken erythroid cells were prepared at 4°C with protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 0.5 g/ml aprotinin, 0.5 g/ml leupeptin, 0.7 g/ml pepstatin, and 0.5 g/ml chymostatin) in all * This work was supported in part by a grant from the Lucille P. Markey Charitable Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. solutions. Washed cells were swollen (10 min, 0.67% phosphate-buffered saline), centrifuged, resuspended (10 9 cells/ml in 10 mM HEPES, pH 7.9, 3 mM MgCl 2 , 0.5 mM DTT, and protease inhibitors), and homogenized with 12 strokes in a Dounce homogenizer (type B pestle). After centrifugation (25,000 ϫ g, 4°C, 20 min), the pellet was resuspended (1.67 ϫ 10 9 cells/ml in 20 mM HEPES, pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM DTT, and protease inhibitors). The suspension was homogenized as above, gently stirred for 30 min, centrifuged as above, and the supernatant was dialyzed (6 h, 4°C against 50 volumes of 20 mM HEPES, pH 7.9, 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, and protease inhibitors). Aliquots (200 -500 l) were frozen on dry ice and stored at Ϫ70°C. Protein concentrations were determined using the Bio-Rad protein assay.
Deletion plasmids are named for the most upstream base of the ⑀ promoter that is still present. ⑀73 (p1307) was produced by BglII digestion of m74 followed by ligation. ⑀142 (p1309) was produced from m143 by HindIII digestion, Klenow fill-in, StuI digestion, and ligation. To produce ⑀51 (p1306) and ⑀88 (p1308), the ⑀ promoter HpaI/NcoI fragments (from m52 and m89, respectively) were inserted into the modified pGL2-Basic between the SmaI and NcoI (initiation codon) sites.
Transient Expression-Transient transfection in erythroid cells from 4-, 5-, and 6-day chicken embryos was performed essentially as described previously (19,32). Typically, 125 ng of luciferase reporter plasmid and 125 ng of RSV-cat control plasmid were transfected into 3 A 412 units of cells using osmotic shock. Transient expression in 10-day erythroid cells was performed by electroporation. Cells (10 8 in 0.5 ml of L-15 medium) were mixed with 2.5 g of RSV-cat and 10 g of test plasmid. After 5 min at room temperature, the cells were electroporated (0.4-cm cuvette, 500 microfarads, 450 V, giving a of ϳ9 ms) using a Bio-Rad Gene Pulser. The cells were cultured with 1.4 ml of complete medium at 37°C for 40 -48 h, as described for the cells transfected by osmotic shock. Luciferase activity was measured (Luciferase Assay System, Promega), normalized to the activity from the RSV-cat (33) internal control, and expressed as a percent of the wild type promoter activity in the same assay. The assay background was determined from transfections with pBluescriptIISK Ϫ and was Յ1% of the ⑀510 level. Plasmids were assayed in duplicate in at least three independent experiments. Results are expressed as the mean Ϯ S.E.

RESULTS
Enhancerless ⑀and -Globin Promoters Express at Similar Levels-In embryos, ⑀and -globin are strongly expressed and at similar levels (see Ref. 21 and references therein). To compare the promoter strength of these genes independent of distant elements, we transfected plasmids carrying a luciferase reporter driven by the or ⑀ promoter into 4-day erythroid cells. We expected that ⑀ expression in the absence of the ␤/⑀ enhancer would be much lower than expression. However, the enhancerless constructs were expressed at similar levels ( Fig.  1). Inclusion of the ␤/⑀ enhancer in the ⑀ construct greatly stimulated ⑀ expression (Fig. 1).
Protein Binding to the ⑀-Globin Promoter-To identify potential regulatory elements in the ⑀-globin promoter, in vitro DNase I footprinting was performed using nuclear extracts from embryonic blood cells. Protected regions included the noncanonical TATA box located at bases Ϫ24 to Ϫ33, a GC-rich, Sp1-like motif (bases Ϫ51 to Ϫ66), and a distal region (Ϫ115 to Ϫ142) that contained CACCC/Sp1-like motifs (Figs. 2 and 3). Footprints were also detected in two regions of the promoter (bases Ϫ86 and Ϫ105 and Ϫ155 to Ϫ183), which contained no readily identified consensus sites. No footprint was evident in the region of the CCAAT box (Ϫ77 to Ϫ73). There were no apparent differences between the footprints obtained with primitive versus definitive extracts.
The protein-binding regions of the ⑀ promoter were charac-terized further using electrophoretic mobility shift assays. Like many other globins, the ⑀ promoter contains a GATA motif in place of a canonical TATA box. When this region was used as a probe (oligonucleotide A), a specific complex with the mobility of cGATA-1 bound to DNA was formed (Fig. 4, lanes 1-6).
Mutation of the TATA element (G 3 T at Ϫ33, oligonucleotide Am33) destroyed the GATA motif and prevented formation of the GATA-1⅐DNA complex (Fig. 4, lanes 7-10). These results indicate that the cGATA-1 protein binds to the GATA/TATA motif of the ⑀ promoter. Oligonucleotide B, spanning the GC-rich, Sp1-like region immediately upstream of the TATA box (Fig. 3), yielded a band with the mobility and binding specificity of an Sp1⅐DNA complex (Fig. 4, lanes [11][12][13][14][15][16][17][18][19]. Antibody to Sp1 inhibited formation of this complex (Fig. 4, lanes 20 -22). These data demonstrate that this region of the ⑀ promoter can bind Sp1 (or an antigenically related protein). Oligonucleotide E corresponds to the protected region of the promoter containing EKLF/CACCC-like (34,35) and Sp1-like sequence motifs and yielded two distinct complexes in mobility shift assays (Fig. 5A). The slower migrating complex had the mobility of Sp1 bound to probe; its formation was competed by an authentic Sp1 oligonucleotide, by oligonucleotide B (shown above to contain an Sp1 site), and by antibodies to Sp1. In contrast, the faster migrating complex was not competed by Sp1 oligonucleotides or antibodies to Sp1. These results demonstrate that this region binds two distinct factors in the erythroid nuclear extracts; the slower complex is due to Sp1 (or a related factor that is indistinguishable by antibody reactivity and binding site specificity), while the faster migrating band is not due to Sp1 and is here designated CACCC.
The CACCC motif centered at Ϫ128 of the chicken ⑀-globin promoter is a perfect match (CCACACCCT) to the EKLF-binding sequence of certain mammalian globin promoters (34,35). In addition, the methylation interference pattern of CACCC is FIG. 1. Enhancerless ⑀and -globin promoters express at similar levels. Transient expression in erythroid cells from 4-day embryos was performed as described under "Experimental Procedures." The -globin promoter and 5ЈUTR (726), the ⑀-globin promoter and 5ЈUTR (⑀WT510), or the same ⑀ promoter with a marked ⑀ 5ЈUTR (⑀510) were linked to a luciferase reporter to produce the plasmids for transfection. Shown for comparison are expression data for RSV-driven luciferase (RSV) and a construct that includes the ␤/⑀ enhancer in addition to the ⑀ promoter (⑀3229). Data are the mean Ϯ standard error from 3-12 independent experiments. Note that there is a 100-fold difference between the right and left scales. identical to that reported for murine EKLF (36). Thus, we examined binding of recombinant murine EKLF to this region (Fig. 5D). EKLF formed a complex with oligonucleotide EЈ (the wild type chicken sequence) that comigrated with EKLF bound to the classical murine ␤ maj -globin promoter site. The point mutation, which inhibited Sp1 binding but not CACCC complex formation (EЈm120), did not significantly affect EKLF binding. Conversely, the mutation that abolished CACCC binding with little effect on Sp1 (EЈm124c) did abolish EKLF binding. These data suggest that the faster migrating complex (designated CACCC) in chicken erythroid extracts may be due to a chicken homologue of EKLF or to a closely related protein.
Despite the presence of footprints in at Ϫ86 to Ϫ105 and Ϫ155 to Ϫ183, no mobility shifts were observed with oligonucleotides comprising these regions of the ⑀-globin promoter (data not shown). Under the conditions used, the CCAAT box probe (oligonucleotide C) did not produce a protein-DNA complex (Fig. 4, lanes 23-28), a result consistent with our failure to detect a footprint at this site. In summary, the protein binding studies have identified cGATA-1 binding to the TATA box centered at Ϫ30 and an Sp1 binding site at Ϫ54. The region at Ϫ120 to Ϫ131 has overlapping binding sites for Sp1 and CACCC, an EKLF-like factor. Under the assay conditions used, no protein binding was observed to the CCAAT box at Ϫ75.
Effect of Deletions on ⑀ Promoter Activity-To examine the  4 g, lanes 1-10, 16 -22, 27, and 28), embryonic chicken erythroid cells, or were omitted (lanes 11 and 23). Competitor oligonucleotides were added in a 10-or 100-fold molar excess as indicated. The positions of free probe and authentic GATA-1⅐DNA and Sp1⅐DNA complexes are indicated. Lanes 20 -22 contained no antiserum, preimmune serum, or anti-Sp1 as indicated. In these three lanes, the free probe was run off the gel.
functionally important regions of the ⑀-globin promoter, a series of deletion mutants were tested with and without the ␤/⑀ enhancer. In reporter plasmids lacking the enhancer (Fig. 6, solid bars), deletion upstream of Ϫ142 did not reduce expression. Deletion to Ϫ88, removing the CACCC/Sp1 element, reduced activity 2-to 5-fold in the primitive cells but to a lesser extent in definitive cells. Deletion to Ϫ73, destroying the CCAAT motif, and to Ϫ51, ablating the Sp1 site, each reduced expression further. These data suggest that full promoter activity resides in the proximal 142 bp of promoter. To examine the effect of the enhancer on ⑀-globin expression, plasmids containing the enhancer (as a 2.7-kilobase pair fragment, from Ϫ3229 to Ϫ510) was cloned upstream of the various promoter deletions. Constructs containing at least 73 bp of promoter  , lanes 1 and 11; panel C, lanes 1, 10, and 16). Competitor oligonucleotides were added in a 10-or 100-fold molar excess. The positions of free probe, a complex with the CACCC site, and authentic Sp1⅐DNA complexes are indicated. Lanes 12-14 of panel A contained no antiserum, preimmune serum, or anti-Sp1. B, methylation interference patterns of DNA from protein complexes with oligonucleotide E (with the upper strand labeled). The regions of a gel corresponding to the free probe and the protein⅐DNA complexes (labeled Sp1 and CACCC in C) were eluted, the DNA was cleaved with piperidine and electrophoresed on 8% denaturing acrylamide gel. Bases whose relative intensity was decreased maximally are designated by solid circles; open circles indicate bases that were reduced in intensity, but to a lesser extent. D, electrophoretic mobility shift assays using recombinant EKLF. Purified GST-EKLF (34) was provided by J. Bieker, and electrophoretic mobility shift was performed as described (36). Competitor oligonucleotides were added as indicated in a 100-fold molar excess.
showed strong enhancement (Fig. 6, open bars; note the different scales), suggesting that the proximal 73 bp of the ⑀-globin promoter is sufficient for productive interaction with the enhancer.
Enhancement Decreases with Development-The decrease in ⑀-globin transcription that occurs with development could be due to diminished enhancer activity or to diminished promoter activity. Promoter activity, in the absence of the enhancer, was comparable in erythroid cells from days 4 -6 ( Fig. 6). (Because different methods were used to transfect the 10-day cells, the data from that stage cannot be compared directly with the results from primitive cells.) To examine the role of the enhancer, enhancer activity (defined as the ratio of activity of constructs containing the promoter and enhancer to those with the promoter only) for each of the promoter deletions was plotted as a function of the age of the cells used for transfection (Fig. 7). In the four promoter constructs containing at least 73 bp of promoter, the average enhancement was 185-, 66-, 18-, and 16-fold for expression in 4-, 5-, 6-, and 10-day cells, respectively. These data suggest that the decrease in ⑀-globin expression with development is due to a decrease in the effect of the enhancer on the ⑀ promoter, and not to a decrease in intrinsic ⑀ promoter activity.
Expression of ⑀-Globin Promoter Point Mutants-We next studied the effect of clustered point mutations on transcription from the ⑀-globin promoter in chicken erythroid cells (Fig. 8). Sequencing of the mutated constructs revealed a T at Ϫ123 in our parent construct, rather than the reported C (31). However, introduction of the C into several constructs showed no difference in expression in the absence of the enhancer (compare m123c versus ⑀510, m111c versus m111, and m124c versus m124). When not otherwise noted, the sequence is a T at this position.
In the absence of the enhancer, mutation of the TATA site (m25, m31), of the Sp1 motif (m52), or the CCAAT box (m74) reduced promoter activity. Interestingly, mutation of sites farther upstream than the CCAAT box did not affect expression levels. In addition, conversion of the TATA site from a GATA to a canonical TATA (m33) had no significant effect on ⑀ expression.
In plasmids containing the enhancer, like those without the enhancer, point mutations in the TATA (m25, m31), Sp1 (m52), and CCAAT (m74) motifs reduced promoter activity (Fig. 9, upper panels). However, in contrast to the enhancerless plasmids, mutations in the CACCC site (m111c, m123c, m124, m124c), in the presence of the enhancer, also decreased reporter expression. Consistent with the greater enhancement in earlier stage cells noted earlier with the deletion mutants, the reduced activity resulting from the point mutants was more striking in 4-day than in 10-day cells.
A measure of the relative contribution of the promoter sites to enhancer-promoter interaction was obtained by comparing the effects of mutated promoter elements in enhancerless and enhancer-containing constructs (Fig. 9, lower panels). Mutations of the CACCC element (m111c, m123c, m124, m124c) affected enhancement more than mutations of the CCAAT (m74), Sp1 (m52), and TATA (m25, m31, m33) regions. Since mutations affecting CACCC factor binding lowered expression more than the mutations affecting Sp1 binding (compare m124 and m124c with m120), promoter-enhancer interaction at this site is probably due to the EKLF-like proteins rather than Sp1. Enhancer effect (defined as the ratio of the luciferase activity from constructs containing the promoter and enhancer to that from constructs containing only the promoter) was calculated from the data in Fig. 6. Data are grouped according to the age of the transfected cell. The promoters in each series are, from left to right, ⑀510, ⑀142, ⑀88, ⑀73, and ⑀51.

DISCUSSION
Functional Elements in the Chicken ⑀-Globin Promoter-Using a combination of protein binding and transient expression assays, we have characterized the cis-elements and trans-factors of the ⑀-globin promoter. The proximal 142 bp were sufficient for full promoter activity, with four sequence motifs contributing to this activity: the TATA box at Ϫ30, an Sp1-binding site (GGTGGG) at Ϫ54, a CCAAT sequence at Ϫ75, and a CACCC motif at Ϫ127. These elements are similar to those observed in other globin promoters, albeit with differences in the order and spacing of the sites (Ref. 37,reviewed in Ref. 38). The ⑀ promoter, like the ␤ A promoter, contains a non-canonical TATA element capable of binding GATA-1. However, our transient expression experiments did not demonstrate any difference between the ⑀ TATA and a canonical TATA in either promoter activity or promoter-enhancer interaction, unlike results from ␤ A expression (39,40).
The CACCC sequence at Ϫ123 to Ϫ131 is a 9/9 match to the EKLF motif in the mammalian ␤-globin promoters, binds recombinant murine EKLF, and mediates enhancer activity. This is curious since, in mice, EKLF is a transcription factor essential for definitive erythropoiesis (41,42) and may (43) or may not (41) be present in primitive erythroid cells. One possible explanation is that chicken EKLF has an embryonic developmental specificity. Alternate hypotheses are that (a) CACCC sites of embryonic genes bind embryonic-specific Krü ppel family transactivators other than EKLF and/or (b) CACCC sites of embryonic genes bind definitive stage-specific Krü ppel family members that act as repressors, for example due to Krü ppelassociated box domains (44,45). Our mutagenesis experiments with the chicken gene are not consistent with the last explanation. Elucidation of the actual mechanisms in the chicken and assessment of their generality requires characterization of chicken EKLF and of the proteins binding to the CACCC-like motifs that control embryonic globin expression in mammals.
Comparison of ⑀and -Globin Regulation-The chickenand ⑀-globins are expressed with the same developmental pattern and at similar levels. The proximal promoters and some of the exons and introns of these genes are very similar in sequence, probably due to gene conversion (23,46). As expected  Fig. 3. Results are the mean Ϯ standard error of 3-9, 4, and 3-6 independent assays for the 4-, 5-, and 10-day data, respectively. NA, construct was not assayed.
FIG. 9. Activity of ⑀-globin promoter point mutations in the presence of the enhancer. Luciferase activity of the mutated constructs containing the enhancer is shown in the top panels. Data are the mean Ϯ standard error of 3-9 and 3 independent assays for the 4-and 10-day data, respectively. The lower panels show the enhancer activity (defined in Fig. 7), calculated using the data from the top panels and Fig. 7. NA, construct was not assayed. from the /⑀ sequence similarity and from functional studies of the promoter (24), the ⑀ promoter Ϫ54 element binds Sp1 and mutation of either the TATA or Sp1 motifs impairs promoter function. From the ⑀ promoter results, it is likely that, in the promoter, GATA-1 binds to the TATA motif, and that the -globin CCAAT box, whose function has not been examined, contributes to promoter function.
It has been proposed that high level expression of ⑀-globin requires the ␤/⑀ enhancer (8,30), but that expression of -globin is driven chiefly by its promoter (24). We found that in transient expression assays in the absence of an enhancer, ⑀ and were expressed at similar levels. Inclusion of the enhancer in the ⑀ construct greatly increased ⑀ expression, suggesting that , like ⑀, also uses distant elements to achieve its high in vivo expression level. The upstream hypersensitive sites are the obvious candidates for this function since they are known to increase -globin expression in transgenic mice (21).
The differences between the regulation ofand ⑀-globins are also interesting and informative. Transcription of -globin is largely dependent on a GATA-1 site at Ϫ214, which is absent from the ⑀ promoter (24). In contrast, the ⑀ promoter CACCC/ Sp1 motif at Ϫ127, which we showed plays a role in enhancer/ promoter interaction, is not present in . It is possible that the promoter GATA site interacts with a distant element(s) just as the ⑀ promoter CACCC site interacts with the enhancer. For example, in the distal promoter of the human ⑀-globin gene, a GATA site interacts with a distant regulatory element (47).
Enhancer-Promoter Interaction-To determine which promoter elements interact with the enhancer, we studied deletion and point mutants of the ⑀ promoter. In the 5Ј deletion mutants, full enhancer activity remained upon removal of the CACCC/Sp1 and CCAAT motifs and was lost only after deletion past the Sp1 site at Ϫ54. Thus, only 73 bp of promoter, containing Sp1 and TATA motifs, were required for full interaction with the enhancer.
However, the point mutants suggest a more complicated situation. In the context of the complete promoter, alterations of the CCAAT or CACCC elements decreased enhancer activity. Nonetheless, despite results showing that the enhancer can interact productively with a 73-bp promoter, point mutation of the proximal Sp1 or TATA motifs in the context of a longer promoter did not reduce enhancer activity. These results suggest that multiple promoter elements can interact with the enhancer and do so in a redundant manner. As a result of this redundancy, alteration of any one promoter motif causes only a partial reduction in fold enhancement.
It is difficult to reconcile the deletion and point mutant data, if one assumes that the deletion series promoters are interacting with the enhancer in the same way. We therefore suggest that they do not. One possible explanation is that, in the intact promoter, the enhancer interacts with the distal promoter elements (such as CACCC and CCAAT), which then influence the more proximal promoter elements. In contrast, in the deletion constructs missing distal elements, the proximal elements interact with the enhancer. Under this model, the data from the point mutants may be a more accurate reflection of the in vivo regulation of the ⑀-globin gene.
Our data suggest that the CACCC site is preferentially involved in interaction with the distant enhancer in primitive cells. The interaction may include EKLF-like proteins binding (48) to enhancer-bound GATA-1 (49). While the simplest interpretation of our data is that the EKLF-like factor(s) is an activator of transcription, we cannot exclude more complex mechanisms in which the EKLF-like factor(s) prevents the binding of inhibitory factors.
Developmental Regulation of ⑀-Globin Transcription-Ex-pression of ⑀ is restricted to primitive erythroid elements and is not expressed in definitive or adult erythroid cells. Our data demonstrate that developmental stage-specific regulation of ⑀ expression is effected through ⑀ promoter/enhancer interaction independent of promoter strength; enhancer activity dropped from 185-fold in primitive cells to 16-fold in definitive cells. Enhancer activity is determined by factors binding to the promoter and to the enhancer, and by factors not binding directly to these elements. Developmental changes in any of these groups of proteins could account for the regulation of ⑀. Two lines of evidence indicate that the mechanism is not regulation of a developmental stage-specific factor binding directly to the ⑀ promoter. First, we were unable to detect stage-specific factor binding to the promoter. Second, no promoter mutant showed stage-specific regulation when lacking the enhancer. Moreover, our studies revealed no evidence of a silencer element. In contrast, regulation of the human ⑀-globin gene does involve a silencer, which can be detected using transient expression (50). Transient expression assays using competition between two promoters for an enhancer have provided evidence for a stage selector protein binding to the ␥-globin genes in human ␥-␤ switching (51,52) and to the ␤ A -globin gene in the chicken (NF-E4; Refs. 28 and 53). Whether these or similar proteins are involved in developmental regulation of the chicken ⑀-globin gene is unclear. Our results define a system in which developmental specificity can be examined in the absence of the confounding effects of a second promoter.
Previous experiments in transgenic mice and using transient expression have demonstrated that competition between promoters for an enhancer/LCR contributes to stage specificity. Since our reporter plasmids contained only a single promoter, competition cannot explain the observed developmental specificity. Moreover, since the ␤/⑀ enhancer functions effectively in definitive erythroid cells to increase ␤ A -globin expression, an intrinsic, absolute stage specificity inherent in the enhancer can also be excluded. Rather, our results suggest that regulation of stage specificity occurs at the level of enhancer/promoter interaction, even in the absence of competition, and is not simply a property of the enhancer or promoter in isolation. This conclusion is consistent with a competition mechanism for the regulation of globin expression. Indeed, developmental regulation of the enhancer/⑀-promoter interaction is likely to be the basis for the decreasing expression of ⑀-globin and the increasing expression of ␤ A -globin seen in competition assays and in vivo.