A Negative Cis-element Regulates the Level of Enhancement by Hypersensitive Site 2 of the b -Globin Locus Control Region*

The core of DNase hypersensitive site (HS) 2 from the b -globin locus control region is a potent enhancer of globin gene expression. Although it has been considered to contain only positive cis-regulatory sequences, our study of the enhancement conferred by segments of HS2 in erythroid cells reveals a novel negative element. Individual cis-regulatory elements from HS2 such as E boxes or Maf-response elements produced as great or greater enhancement than the intact core in mouse erythroleukemia (MEL) cells, indicating the presence of negative elements within HS2. A deletion series through HS2 revealed negative elements at the 5 * and 3 * ends of the core. Analysis of constructs with and without the 5 * negative element showed that the effect is exerted on the promoters of globin genes expressed at embryonic, fetal, or adult stages. The negative effect was observed in bipotential human cells (K562 and human erythroleukemia (HEL) cells), proerythroblastic mouse (MEL) cells, and normal adult human erythroid cells. The novel negative element also functions after stable integration into MEL chromosomes. Smaller deletions at the 5 * end of the HS2 core map the negative element within a 20-base pair region containing two conserved sequences. The human b -globin gene domain at chromosomal position 11p15.5 consists of a cluster of globin genes that are expressed only in erythroid cells. The genes

The core of DNase hypersensitive site (HS) 2 from the ␤-globin locus control region is a potent enhancer of globin gene expression. Although it has been considered to contain only positive cis-regulatory sequences, our study of the enhancement conferred by segments of HS2 in erythroid cells reveals a novel negative element. Individual cis-regulatory elements from HS2 such as E boxes or Maf-response elements produced as great or greater enhancement than the intact core in mouse erythroleukemia (MEL) cells, indicating the presence of negative elements within HS2. A deletion series through HS2 revealed negative elements at the 5 and 3 ends of the core. Analysis of constructs with and without the 5 negative element showed that the effect is exerted on the promoters of globin genes expressed at embryonic, fetal, or adult stages. The negative effect was observed in bipotential human cells (K562 and human erythroleukemia (HEL) cells), proerythroblastic mouse (MEL) cells, and normal adult human erythroid cells. The novel negative element also functions after stable integration into MEL chromosomes. Smaller deletions at the 5 end of the HS2 core map the negative element within a 20base pair region containing two conserved sequences.
The human ␤-globin gene domain at chromosomal position 11p15.5 consists of a cluster of globin genes that are expressed only in erythroid cells. The genes are arranged 5Ј-HBE-HBG2-HBG1-HBD-HBB-3Ј from the centromere to the telomere. The genes are temporally expressed in the order of their array along the chromosome, with ⑀-globin (encoded by HBE) produced embryonically, G ␥and A ␥-globin (encoded by HBG2 and HBG1) produced fetally, and ␦and ␤-globin (encoded by HBD and HBB) produced in the adult. Controlled expression of the genes occurs through proximal regulators, including the promoters of the individual genes, and distal elements, such as the dominant regulatory region known as the locus control region (LCR) 1 (1)(2)(3)(4). The LCR spans at least 17 kilobases of DNA, located 23 to 6 kilobases upstream of the ⑀-globin gene, and contains 5 DNase I-hypersensitive sites (HS1-5) (5-7). The LCR is defined by its ability to confer high level, tissue-specific expression of linked genes at all sites of integration examined in transgenic mice (8). It is a powerful enhancer (8 -11). Analysis of LCR deletions in thalassemic patients (12,13) and experiments testing LCR function at ectopic sites (e.g. 14, 15) have been interpreted as showing a domain-opening activity, but this has not been seen in directed deletions of the LCR (10,11). Thus, all the functions of the LCR have not been conclusively defined (16), but it is clear that it is a major cis-regulatory element for the ␤-like globin gene cluster. DNA fragments containing individual HSs can produce some of the effects of the LCR (17)(18)(19)(20), and growing evidence supports the model that multiple HSs work together in a holocomplex at the LCR (9,15,(21)(22)(23)(24)(25). A full set of cis-regulatory sites and the protein(s) that works at them is not yet known either for the function of an individual HS or for its interaction with other sites.
One of the most potent components of the LCR is HS2, which can produce high level, position-independent expression of linked genes in transgenic mice (17,20,26) and stably transfected erythroid cell lines, including both human K562 cells and mouse erythroleukemia (MEL) cells (21,26,27). It can also act without stable integration into chromosomes, conferring both strong enhancement (28,29) and heme-inducibility (30 -32) on linked globin genes in transiently transfected K562 cells. The core of HS2 is defined as the smallest region conferring position-independent expression of the ␤-globin gene; it is contained within a 374-base pair HindIII-XbaI fragment (26,33). Simultaneous alignment of multiple DNA sequences of the ␤-globin domain from mammals shows several conserved blocks, many of which have been confirmed as protein binding sites needed for the function of HS2 (4,34). A particularly dense cluster of sites, spaced at 10-base pair intervals within a 100-base pair segment of the HS2 core, is suggestive of a contiguous array of proteins on the same face of the DNA helix. A dimer of MAREs, or Maf-responsive elements (35), is crucial for the enhancement and heme-inducibility by HS2 in K562 cells (30,31) as well as high level expression in transgenic mice (26,33,36). The MAREs are binding sites for AP1, NFE2 (37)(38)(39), Nrf1 (40)/LCRF1 (41)/TCF11 (42), Nrf2 (43), and Bach1 (44). Most of these proteins function as heterodimers with small Maf proteins; hence, the binding site has been named for this common component. The MAREs are not sufficient for full-level enhancement (26,27,31,32), indicating that other proteins are also functioning at HS2. In particular, GATA1 and/or GATA2 (26,33,45,46), basic helix-loop-helix proteins and other E box-binding proteins (32), HS2NFE5 (47), and proteins binding to the highly conserved CAC motif (33) have all been implicated by a combination of mutagenesis, in vivo footprinting, in vitro binding coupled with antibody studies, and sequence conservation. Additional conserved sites are found throughout and beyond the HS2 core (4), and many of these regions have also been implicated in function of the LCR (33,48,49).
To better understand the contribution of the several conserved sequences in the HS2 core to enhancement in erythroid cells, we developed methods using cationic lipids to transiently transfect several erythroid cell lines (50). This has allowed us to test constructs containing promoters from several ␤-like globin genes in murine MEL cells (which produce ␤-globin (51)), human K562 cells (which produce ⑀and ␥-globin (52,53)), and human erythroleukemia (HEL) cells (which produce mainly ␥-globin (54)). The results identify cis-acting sequences within the HS2 core that negatively regulate expression of linked ␤-like globin genes in erythroid cell lines and in normal human adult erythroid cells. The negative effect is seen both in transiently transfected cells and after integration into stably transformed cells.

EXPERIMENTAL PROCEDURES
Transient Transfection of Erythroid Cells-The cationic lipid reagent Tfx50 (from Promega) was used to transiently transfect MEL, K562, and HEL cells (50) following the manufacturer's protocol for suspension cells. Optimal transfection was obtained at a 2:1 ratio (charge to mass) of cationic lipid to DNA. The reagent and DNA remained in the cell culture for 48 h, at which point the cells were harvested. Three different experimental designs were used to examine the level of expression in transient transfections. In the first, each plasmid containing a reporter gene was assayed as a titration of DNA mass from 0.25 g up to a maximum of 8.0 g (maintaining the 2:1 charge to mass ratio), with the results reported as luciferase activity in relative light units (RLU)/s. In the second, plasmids were transfected in triplicate at the single DNA mass most frequently seen to be optimal for enhancement in each cell line, which was 2.0 g for MEL cells and 3.0 g for K562 and HEL cells. The resulting luciferase activities were normalized to the amount of total protein in each sample, which was determined by the Bio-Rad BCA microprotein assay. In this design, Student's t test was used to determine the probability that paired samples did not belong to distinct, nonoverlapping data sets. Our initial attempts to use pRSVlacZ as a cotransfection control were unsuccessful, since the amount of pRSVlacZ required to obtain activity above the background were incompatible with experimentally determined transfection conditions, which limited test DNA to a maximum of 2 g total in MEL cells (50).
Subsequent testing showed that small amounts of a cotransfection control vector expressing the luciferase gene from Renilla (sea pansy, Promega) provided measurable activity without interfering unacceptably with the test plasmid. Thus the third design was to include the Renilla luciferase cotransfection control plasmid in the following amounts: for MEL cells, 1.75 g of test plasmid plus 0.25 g of control plasmid, and for K562 and HEL cells, 1.9 g of test plasmid and 0.1 g of control plasmid.
Normal human adult erythroid cells (hAEC) were cultured from human peripheral blood using the two-phase system of Fibach et al. (55). They were transiently transfected by electroporation using the methods described in Li et al. (56).
Stable Transfection of MEL Cells-MEL cells (1 ϫ 10 6 ) were stably transfected using the Tfx50 reagent to introduce 1.5 g of linearized test plasmid plus 0.5 g of pSV2neo, a neomycin phosphotransferase expression vector. One day after transfection, G418 was added to the culture at 0.6 mg/ml. Pools of stably transfected cells were harvested after 2 weeks of growth under selection to measure luciferase activity. Pools of transfected cells were induced by growth in 4 mM hexamethylene bisacetamide (HMBA) for 5 days; uninduced cultures were maintained for the same period in the absence of inducer.
Luciferase Assay-As described previously (32), up to 20 l of cellular extract was assayed in 100 l of Promega luciferase assay reagent.
Plasmid Construction-Luciferase constructs containing the rabbit ⑀-globin promoter from Ϫ573 to ϩ85, without (⑀luc or pBS⑀-luciferase. 4) or with the human HS2 core (⑀HS2 or pBS⑀-luciferase-hHS2HX) have been described previously (48,57). The human HS2 core is the HindIII-XbaI fragment, HUMHBB positions 8486 -8860. The ␤-globin-luciferase reporter constructs contained the mouse ␤-major globin gene (Hbb-b1) promoter sequence from Ϫ106 to ϩ 26 (␤106), Ϫ340 to ϩ26 (␤340) or the HBB promoter from Ϫ374 to ϩ45 (h␤) fused to the luciferase-coding region of pGL2Basic (Promega) (58). The human HS2 core was added by subcloning it initially into pBluescriptII KSϪ (Stratagene), excising it by digestion with KpnI and SacI and ligating it to these same restriction cleavage sites within the ␤-globin-luciferase plasmids to generate ␤106HS2 and ␤340HS2. Plasmid ␥luc contains the human A ␥-globin gene promoter from, Ϫ260 to ϩ35, fused to the luciferase coding region of pGL3Basic (Promega), and plasmid ␥HS2 has in addition the KpnI to BglII DNA fragment (HUMHBB positions 7764 -9218) containing HS2 (47). ␥HX and BX contain the HindIII-XbaI and BalI-XbaI fragments of HS2, respectively. Constructs containing the human MAREs or multimers of the HS2 E box were made by inserting oligonucleotides, listed below, that had flanking sequence to allow directional cloning (via HindIII and SpeI sites) into either the ⑀or ␤-luciferase vectors.
Deletion Series through HS2-Beginning with the HindIII-XbaI fragment containing the human HS2 core cloned into pBluescript II KSϪ (pBSHS2HX), the desired deletion fragments were obtained by restriction digestion or PCR amplification and cloned into pBS⑀-luciferase.4 (48). The 5Ј deletion series contained four constructs: BX (BalI to XbaI), MX (Mares to XbaI), 8701X, and 8762X; the 5Ј ends are at HUMHBB positions 8568, 8658, 8701, and 8762 (respectively), and all 3Ј ends are at 8860. The 3Ј deletion series contained four constructs: H8750 (Hin-dIII-8750), HCAC (HindIII-CAC), H8650 (HindIII-8650), and HB (Hin-dIII-BalI); the 3Ј ends are at HUMHBB positions 8750, 8687, 8651, and 8568 (respectively), and all 5Ј ends are at 8486. The BX fragment of HS2 was directionally cloned into the ⑀luc plasmid by excising it from pBSHS2HX with MscI and PstI and ligating it into ⑀luc at SmaI and PstI sites to generate ⑀luc-Bal-Xba. The BX fragment was also inserted into the ␤106 luciferase expression plasmid by excising it from ⑀luc-Bal-Xba with PstI and SpeI, cloning it into pBluescriptIIKSϪ at the same restriction sites, excising again with SacI and BssHII, and ligating it into SacI and MluI sites in the ␤106 plasmid. The other fragments were amplified by PCR using the following primers (sequences are listed below): MX by primers "human MAREs" and T7; 8701X by primers 8701/BamHI and T7; 8762X by primers 8762 and T7; H8750 by primers 8750/PstI and T3; HCAC by primers CAC/PstI and T3; H8650 by primers 8650/PstI and T3; HBalI by primers BalI and T3. The resulting PCR products were digested with appropriate restriction enzymes to allow directional cloning into pBS⑀-luciferase.4. Three deletions between the HindIII and BalI restriction sites were generated that corresponded at their 5Ј ends to conserved sequences within the HS2 core. The template for amplification was the pBSHS2HX vector. All three constructs in this deletion series had the same sequences at their 3Ј end, generated by using the T3 primer of pBluescript. PCR products were prepared for subcloning into the ␤106-luciferase vector by digestion with restriction endonucleases. KpnI was introduced in the sequence of the forward primer, and SacI is found in the polylinker between the 3Ј end of the HS2 core and the T3 primer.
Nucleotides added at the 5Ј ends of the sequence are shown in lowercase. Consensus sequences for binding sites, conserved regions, or restriction cleavage sites are underlined.

RESULTS
Weak Enhancement by HS2 in MEL Cells-Using cationic lipids to deliver DNA to erythroid cells lines for transient expression, we were surprised to find that the effects of HS2 on enhancement were quite modest in MEL cells. A series of luciferase reporter gene plasmids (Fig. 1A), driven by promoters from either the rabbit embryonic HBE gene or the mouse fetal/adult Hbb-b1 gene (encoding the ␤-major globin), were transfected into MEL and K562 cells at increasing concentrations of DNA to find optimal conditions for enhancement. Comparison of the resulting luciferase activities between plasmids with and without the HS2 core shows the expected robust enhancement by HS2 for both the HBE and the Hbb-b1 promoters in K562 cells (50-and 30-fold in this experiment, Fig. 1, B and C). However, enhancement in MEL cells is substantially lower, being only 12-and 3-fold for the HBE and Hbb-b1 promoters, respectively ( Fig. 1, B and C). Although the measured fold enhancement varies in different experiments, comparisons of transfections done at the same time invariably show a lower level of enhancement in MEL than in K562 cells. Also, as will be shown later, this modest effect of the HS2 core in MEL cells is seen in several different transfection protocols. Furthermore, a similar response from the promoters of rabbit, mouse, or human globin genes to enhancers containing fragments of the human HS2 core suggests that species-specific effects are not evident.
The data in Fig. 1 show that the transfected plasmid templates were active, even in cells in which the endogenous homologous genes are inactive. In particular, the plasmid with an HBE promoter without an enhancer is as active in MEL cells, which do not produce ⑀ y -globin from the endogenous Hbb-y locus, as in K562 cells, which produce ⑀-globin from the endogenous HBE gene (800 RLU/s in K562 and 500 RLU/s in MEL). However, the response to the HS2 core was considerably lower in transiently transfected MEL cells. Likewise, the plasmid with an Hbb-b1 promoter responded better to the HS2 core enhancer in K562 cells, where the endogenous homolog HBB is inactive, than in MEL cells. The Hbb-b1 promoter used for the data in Fig. 1C extended to nucleotide position Ϫ340 (with respect to the cap), and thus contained some sequences implicated in negative regulation of this gene (59,60). Deletion of these negative elements did not restore enhancement in MEL cells, since the Hbb-b1 promoter extending only to Ϫ106 still showed a 6-fold reduction in enhancement in MEL cells compared with K562 cells (Fig. 1D). Thus, these results indicate that the reduced enhancement in MEL cells is not related to the type of globin gene promoter used or the activity of the endogenous homologous genes.
The limited enhancement by the HS2 core in MEL cells is not an artifact of using only parts of the genes. The plasmid ⑀-luc (Fig. 1A) was designed to include almost all of the HBE gene (57,61). It has the rabbit HBE promoter segment extending from Ϫ573 to ϩ85, which includes multiple positive and negative elements (62) that are homologous to those of the human HBE gene (56,63,64). Since the luciferase-coding region replaces a segment extending from the 3Ј end of exon 1 to almost the 3Ј end of exon 2, the reporter gene retains most of the HBE exons, intron 2, and an extensive 3Ј-flanking region. The difference in enhancement between the two cell lines is seen both for expression plasmids containing the promoter and internal and 3Ј-flanking regions (of HBE) as well as for plasmids containing only the promoter (e.g. from Hbb-b1) but no other parts of the globin gene.
Greater Enhancement by Individual Cis-acting Elements than the HS2 Core in MEL Cells-The HS2 enhancer contains multiple cis-acting sequences, and thus, the reduced enhancement by HS2 in MEL cells could result from less activity from specific positive cis-acting elements, greater activity from negative cis-acting elements, or both. Therefore, we compared the activity of multimers of individual transcription factor binding sites with that of the entire enhancer in both cell lines. Previous observations (26,27,31,32) showed that in K562 cells the tandem MAREs are not sufficient for full enhancement. A similar result is obtained using the cationic lipid as the transfection reagent for K562 cells, assayed at increasing amounts of transfecting DNA ( Fig. 2A). A second cis-element of HS2 known to contribute to enhancement in K562 cells is the E box at position 8701 in the sequence file HUMHBB (32). We tested the ability of a multimerized E box sequence, containing five copies, to enhance independently of other elements in HS2 (Fig.  2B). The pentamer of the E box enhanced 3-fold, which is significant (p Ͻ 0.05, Table I) but much less than the enhancement by the HS2 core.
In contrast, in MEL cells, the MAREs enhance the HBB promoter more strongly than does the HS2 core (Fig. 2C). Also, the pentamer of the E box stimulates expression of the linked HBB promoter as strongly as the intact core does (Fig. 2D). The enhancements by both the E box multimers and the intact HS2 core are significant (p Ͻ 0.01), and they are not significantly different from each other. Thus the reduced enhancement in MEL cells is not exclusively from reduced activity of either of these transcription factor binding sites, since both the MAREs and the E boxes were as active or more active than the intact core. These results suggest that a cis-acting element in HS2 has a strong, negative effect in MEL cells.
Effects of HS2 Cis-elements in MEL Cells-To examine the contributions to enhancement by groups of cis-elements within the HS2 core, we tested the activities of a deletion series from the 5Ј and 3Ј ends of the HindIII-XbaI HS2 core. Members of each series were linked to the HBE-luciferase reporter gene and transfected into MEL cells. Two different experimental protocols were followed. In the first, the amount of DNA used in the transfection was titrated to find an optimal concentration for enhancement. In the second protocol, the transfections were repeated in triplicate at the optimal DNA concentration so that the results could be analyzed statistically. As shown in Fig. 3, the terminal deletions reveal negative cis-elements in HS2.
For the triplicate transfections shown in Fig. 3B, the Hin-dIII-XbaI HS2 core enhanced HBE-luciferase expression 6.6fold (Fig. 3B). Deletion from the 5Ј end increased enhancement to 11.7-fold for the BalI-XbaI fragment. This is a significant increase (p Ͻ 0.01 for comparison with the intact core; Table  II). A similar result was obtained in the DNA titration experiment shown in Fig. 3C. In this case, the enhancement increased 4-fold by deleting the HindIII-BalI fragment (from 2.5-fold for the HindIII-XbaI HS2 core to 10-fold for the BalI-XbaI fragment). Thus cis-acting elements located between Hin-dIII and BalI exerted a negative effect in this assay. Additional deletions into the HS2 core caused a progressive loss of enhancement but no deletion completely removed activity (Fig. 3 and Table II).
Deletions from the 3Ј end of the HS2 core revealed a second negative element in MEL cells. For the triplicate transfections shown in Fig. 3B, removal of the 3Ј fragment caused an increase in enhancement from 6.6-to 9.5-fold; this increase is significant (p Ͻ 0.02; Table II). This result was confirmed by the DNA titration experiment in Fig. 3C in which deletion of the 3Ј fragment doubled the level of enhancement, from 3-to 6.8-fold. Similar to the results with the 5Ј deletion series, additional deletions from the 3Ј end caused a progressive reduction in enhancement (Fig. 3), but again, all fragments re-tained significant activity (p Ͻ 0.001 for each in comparison with ⑀-luciferase, Table II). These data show that truncation of either the 5Ј or 3Ј end of the core of HS2 partially relieves repression in MEL cells.
The activity of a particular segment of HS2 depends on the context. Deletion of the 5Ј region of HindIII to XbaI the 3Ј region of 8750 to XbaI shows that both have a negative effect on enhancement when the rest of the HS2 core is present. In contrast, both have a small but significant positive effect when assayed by themselves. This suggests that the negative effect could arise by interference with positive elements within the rest of HS2.
Negative Effect of HindIII-BalI Region on Both HBE and Hbb-b1 Genes in Three Different Erythroid Cell Lines-To examine whether the cis-regulatory sequences at the 5Ј end of the HS2 core had a negative effect only on the embryonic ⑀-globin gene promoter in MEL cells, the effect of its deletion was tested on promoters from the ⑀-, ␥-, and ␤-globin genes in K562, HEL, and MEL cell lines (Fig. 4A). In this series of experiments, the test plasmid with the globin gene promoter-driving expression from the firefly luciferase gene was cotransfected with a control for transfection efficiency, which was the Renilla luciferase gene expressed from a tk promoter. The results are plotted as a ratio of firefly luciferase activity to Renilla luciferase activity.
As shown in Fig. 4B, removal of the HindIII-BalI fragment doubled the level of enhancement in K562 for both the HBE promoter (from 6-fold in its presence to 13-fold in its absence)  HEL cells produce mainly ␥-globin and a small amount of ⑀-globin, but no ␤-globin, from its endogenous genes (54). Results of transient transfections (Fig. 4B) show that the HindIII-XbaI fragment comprising the HS2 core had almost no effect on either the HBE or Hbb-b1 promoters. However, removal of the HindIII-BalI negative element increased enhancement substantially, to 3.6-fold for HBE and 5.0-fold for Hbb-b1. In contrast to the results with K562 cells, the HBG1 promoter responded more strongly to the BalI-XbaI fragment of HS2 (lacking the negative element) than to the core HindIII-XbaI fragment (3.1-and 2-fold enhancement, respectively, Fig. 4B). This small but significant (p Ͻ 0.01) difference indicates that the negative element can decrease enhancement of any of the three promoters tested, although its effect is stronger on HBE and Hbb-b1 than on HBG1.
Results of transfections in MEL cells were similar to those in HEL cells for all three promoters (Fig. 4B). Little to no enhancement of the Hbb-b1 or HBE promoters was observed for plasmids containing the HS2 core (HindIII-XbaI fragment), but both promoters showed substantial enhancement with the BalI-XbaI fragment of HS2 (5.8-fold for HBE and 7.0-fold for Hbb-b1). In contrast, the HBG1 promoter was enhanced by the HS2 core (3.5-fold), but even this increased to 8.8-fold upon deletion of the negative element. Thus the 5Ј HindIII-BalI fragment is exerting a negative effect on all three promoters in MEL cells, with its strongest effect on HBE and Hbb-b1. A striking decline in enhancement is observed in MEL cells for the KpnI-BglII fragment containing HS2 and the HBG1 promoter (Fig. 4B, plasmid ␥KB). Transfection of both K562 and HEL cells with this same plasmid DNA showed strong enhancement by the KpnI-BglII fragment. Further investigation would be required to determine whether this corresponds to a difference in species specificity, promoter preference, or stage of differentiation.
Assays described so far test human HS2 on promoters of globin genes from other species, i.e. rabbit and mouse. To test whether the human HS2 core works more effectively with a human HBB promoter, MEL cells were transfected with con-  structs containing an HBB promoter in place of the Hbb-b1 promoter. In this plasmid, the HS2 core was unable to enhance expression from the human HBB promoter, but the BalI-XbaI fragment still conferred 3.8-fold enhancement ( Fig. 5; hu␤HX compared with hu␤BX). These results are similar to the results seen with the mouse Hbb-b1 promoter, showing that the reduced enhancement is not an artifact of using a heterologous promoter. The fold enhancements should not be compared between the results in Fig. 4 and those in previous figures because of the difference in experimental protocol. The inclusion of a cotransfection control does affect the level of expression from the test plasmid (65), and thus, it could affect the magnitude of the enhancement. However, the increase in enhancement upon removal of the HindIII to BalI fragment of HS2 is consistently seen for HBE and Hbb-b1 promoters in all three cell lines. It is seen with or without cotransfection controls in HEL and MEL cells (data not shown) and when using a different control (pRS-VlacZ) in K562 cells (data not shown).
In summary, the data in Figs. 4 and 5 show that the 5Ј HindIII to BalI fragment of HS2 has a negative effect upon transient transfection of three immortalized erythroid cell lines from two different species (human and mouse), affecting four different globin genes. Hence the effect is not species-specific nor is it unique to one promoter.
Negative Effect of HS2 HindIII to BalI Fragment in Normal Human Adult Erythroid Cells-To test whether the negative effect of the 5Ј end of HS2 could be seen in normal human cells rather than in continuously growing cell lines, normal hAEC were isolated by culturing peripheral blood by the procedure of Fibach et al. (55). The hAEC were transfected with plasmids containing the ␤106-luciferase reporter gene with or without DNA fragments containing segments of HS2. The results of the transient expression assays (Fig. 6) show that the HindIII to XbaI HS2 core segment had no effect on the reporter gene, whereas the BalI-XbaI DNA fragment boosted expression over 7-fold. Thus the negative effect seen for the 5Ј HindIII-BalI fragment is not an artifact of transfecting immortalized cell lines.
Negative Effect of HS2 HindIII to BalI Fragment on Stable Expression after Integration into MEL Cells-The effect of the 5Ј end of the HS2 core was tested on stably integrated DNA constructs in MEL cells. As shown in Fig. 7A and Table III, in three independently generated pools of stably transfected MEL cells, the BalI-XbaI DNA fragment from HS2 enhanced expression of the ␤106-luciferase reporter gene much more than did the HindIII to XbaI HS2 core segment. Although the fold enhancement for each HS2-containing construct varied among the three pools, the BalI-XbaI fragment consistently produced a substantially greater enhancement than did the HS2 core fragment. Thus the negative effect of the 5Ј HindIII-BalI frag- FIG. 5. Negative effect of the Human HindIII to XbaI fragment on reporter genes carrying the human ␤-globin promoter. A, MEL cells were transiently transfected with the HBB promoter-luciferase reporter fusion carrying either no enhancer (hu␤), the (hu␤HX), or BalI-XbaI fragment (hu␤BX). Transfections were done in triplicate, and the firefly luciferase activity (FFLUC) from each sample was normalized to the amount Renilla luciferase activity as described previously. B, diagram of the HindIII-XbaI and BalI-XbaI fragments of HS2 upstream of the HBB promoter-luciferase reporter fusion.

FIG. 4. Comparison of the effects of deleting the HindIII to BalI fragment of HS2 on three different globin genes in three different cell lines.
A, diagram of DNA fragments within or containing HS2, which were ligated to luciferase reporter genes driven by promoters from HBE, HBG1, or Hbb-b1. Conserved blocks in the HS2 core are indicated by the shaded rectangles; the darker-shaded rectangles are the MAREs (labeled) and E box at 8701. B, the graphs compare the expression from plasmids containing various combinations of promoter and enhancers after transfection of K562, HEL, or MEL cells. Expression is plotted as the ratio between the activity of firefly luciferase (FF luc) encoded in the test plasmids and Renilla luciferase (R luc) encoded in a cotransfection control. All transfections were in triplicate; the mean is plotted, and the error bars show the S.D. The firefly luciferase in the HBG1 constructs is from pGL3Basic (Promega), which encodes a luciferase enzyme with higher activity and, thus, results in higher RLU output than do the pGL2derived firefly luciferases in the HBE and Hbb-b1 constructs. ment is seen on templates integrated into the mouse chromosomes.
Effects of Induction of MEL Cells-We used these pools of stably transfected MEL cells to test whether induction by HMBA could overcome the negative effect of the HindIII-BalI fragment. As shown in Fig. 7B and Table III, expression of the ␤106-luciferase reporter gene increased in each of two pools of transfected cells in response to the inducer, HMBA. However, the amount of induction was similar for each construct (almost all within a range of 5-10 -fold) in the absence or presence of a DNA fragment containing HS2. Thus, for these constructs, the major cis-acting sequence affecting induction appears to be the Hbb-b1 promoter. In particular, induction by HMBA did not overcome the negative effect of the HindIII-BalI fragment in HS2.
Deletion Series through the HindIII to BalI Region of the HS2 Core-The 5Ј negative element was mapped more precisely using additional deletions at a finer resolution. As shown in Fig. 8, the 5Ј end of each deletion corresponds to one of several sequences in the HS2 core that are conserved among mammalian globin loci (4). MEL cells were transfected with the series of constructs that contained no enhancer (␤106), the HS2 core (␤HX), or subfragments of the HS2 core beginning at conserved region 2, 3, or 4 (␤CR2, -3, or -4, respectively) or the BalI-XbaI region (␤BX). All fragments had the same 3Ј end, the SacI site in the polylinker downstream of the HS2 core. The HS2 core conferred no enhancement on the expression of the Hbb-b1 promoter, but CR2-, -3, and -4 constructs all showed a substantial increase of 6.1-, 4.9-, and 5.9-fold (respectively) over the enhancerless construct ␤106. This is similar to the 5.2-fold enhancement obtained with the BalI-XbaI fragment. Thus the sequence responsible for the negative effect on HS2 enhancement is localized to a region of 20 nucleotides at the 5Ј end of HS2. Two conserved elements that fall within this area are candidates for the negative cis-regulatory activity, one at the HindIII restriction site and one in a neighboring T-rich motif. DISCUSSION Many studies show that HS2 of the ␤-globin LCR strongly stimulates the expression of linked globin genes in transgenic mice (17,26,33,66), in stably transfected MEL (21, 26) and K562 cells (27,67), and in transiently transfected K562 cells (29,30). Multiple cis-acting sequences in HS2 contribute to and modulate the enhancement activity (26,32,33,36,47,49,68). Our studies reported here show that a 20-base pair region within the HindIII to BalI fragment at the 5Ј end of the HS2 core exerts a negative effect, whereas the remaining segment of HS2 strongly stimulates expression.
The negative effect was observed in four different types of erythroid cells, including three cell lines (MEL, HEL, and K562) as well as normal hAEC. All of these cells are derived from adults, but HEL and K562 cells are not fully committed to the erythroid lineage and express globin genes that are maximally expressed in fetal and embryonic development. This indicates that the negative effect of the HindIII-BalI fragment can be exerted at many stages of adult erythroid differentiation (e.g. before and after commitment and at various stages of erythroid maturation). Although developmental specificity cannot be definitively addressed with these transfection experiments, we note that a globin gene expressed predominately during primitive erythropoiesis (rabbit HBE) and one expressed during definitive erythropoiesis (mouse Hbb-b1) responded strongly to this negative element. This was seen both in cells expressing the endogenous homolog as well as in cells in which the homolog is silent (e.g. the effect was seen on Hbb-b1 in MEL cells, where the endogenous gene is expressed, and in HEL and K562 cells, where the homologous gene HBB is not expressed). Thus the HS2 negative element is active on genes expressed at different stages of development. This is consistent with studies of the human HS2 in transgenic mice; both gainof-function (19,20) and loss-of-function (15,69) experiments show a comparable effect of HS2 at all developmental stages.
Human HBG1, which is expressed predominately in fetal life, also responded to this negative element in HEL and MEL cells, albeit less dramatically than the other genes tested. Indeed, transfection of a HBG1 promoter-luciferase reporter in K562 cells showed no effect of the HS2 HindIII-BalI fragment. The absence of an effect in K562 cells and the rather modest effect in MEL cells helps explain why this effect was not reported previously, since the HBG1 promoter has been commonly used in transfections of these cells (27,30,47). Furthermore, the HBE gene also is frequently used in transfections of K562 cells (28,56,64,65,70), and its expression is enhanced by the HindIII-XbaI core of HS2, although the enhancement is increased after deletion of the HindIII-BalI negative element. The observation that the HindIII-BalI fragment of HS2 has only a modest, and sometimes no, effect on HBG1 in the same cells where it has a strong effect on HBE and Hbb-b1 shows that this part of HS2 has some promoter specificity.
The HindIII-BalI fragment of HS2 decreases enhancement after stable integration into MEL cell chromosomes as well as in transiently transfected cells. The unintegrated DNA in transiently transfected cells is readily packaged into nucleosomes (71,72), and thus, the effects of enhancers observed in transiently transfected cells do not result from proteins interacting with naked DNA. However, in these experiments, it is not clear that all templates are in the same chromatin structure. We examined pools of stably transfected MEL cells to test the effects of different segments of the HS2 core on templates integrated into chromosomes, and hence fully packaged into chromatin. Pools were tested so that many different integration sites could be assayed at once. These experiments showed that the negative effect of the HindIII-BalI fragment at the 5Ј end of the HS2 core is exerted even when the template is in chromatin.
Changes in the level of enhancement conferred by fragments of the HS2 core support the results of studies where transcription factor binding sites were specifically targeted by point mutation. For instance, mutations in the 8701 E box reduce HS2 enhancement of the human HBE promoter in K562 cells (32). Multimers of this same E box sequence confer a 3-fold increase in the level of expression from the mouse Hbb-b1 promoter in MEL cells (Fig. 2C). Deletion of the 8701 E box (plus GATA motifs and other sequences) decreases the level of enhancement of HS2 (compare 8701-XbaI versus 8762-XbaI fragment in Fig. 3). This is consistent with previous mutagenesis results. However, a deletion of same region from the 3Ј end of HS2 had no effect on enhancement. One explanation is that the presence of MAREs in the HindIII-8750 and HindIII-CAC fragments may influence the ability to see an effect of E boxes in the 3Ј deletion series. Thus the effect observed for cis-elements is dependent on their context, i.e. other binding sites.
Interestingly, the negative effect of the 5Ј end of HS2 has been recorded in previous experiments with transgenic mice but interpreted differently. Liu et al. (36) compared the activity of the HindIII-XbaI HS2 core to that of a BalI-SnaBI fragment, which has the same 5Ј end but extends 49 base pairs further 3Ј than the BalI-XbaI HS2 fragment used in our experiments. The fragment lacking the HindIII-BalI fragment had a 3-fold higher average activity. The authors suggested that this may result from an undocumented mosaicism in some of the mouse lines. Also, Talbot et al. (26) assayed a number of HS2 segments for enhancement of the Hbb-b1 gene in stably transfected MEL cells. A HaeIII-XbaI fragment (of similar end points to the BalI-XbaI fragment discussed here) produced a slightly greater level of enhancement than did the HindIII-XbaI fragment in several pools of clones. Although these differences were not considered significant in the Talbot et al. (26) paper, they are consistent with our demonstration of a negative element at the 5Ј end of HS2.
Although our studies utilizing fragments of the LCR in proximity to various globin gene promoters are useful for dissecting regulatory components, they leave open a large number of possibilities for how these components could function during physiological regulation. For instance, the negative function of the HindIII-BalI fragment could be utilized to help keep globin genes silent in nonerythroid cells, to turn them off in erythroid cells at appropriate stages, or to attenuate the activation by HS2 early in erythroid differentiation. The situation is further complicated by the fact that multiple HSs function together in the LCR, and our experiments do not address how this part of HS2 functions in this context. However, it is important to know FIG. 7. Negative effect of the Hin-dIII to BalI region in stably transfected MEL cells before and after induction. MEL cells were transfected ␤106, ␤HX (containing the ␤106 promoter and the HindIII-XbaI core of HS2), and ␤BX (containing the ␤106 promoter and the BalI-XbaI fragment of HS2) along with a neomycin phosphotransferase expression vector as a selectable cotransfection marker. Pools of drug-resistant cells containing multiple clones were assayed for the amount of luciferase activity in triplicate. The means are plotted, and the S.D. are shown as error bars. A, results for three sets of stably transfected pools of cells assayed without induction. Pool 1 was the initial pool of stably transfected cells, and pools 2 and 3 were additional ones generated in a second, independent experiment. B, pools 2 and 3 of stably transfected cells were subsequently grown in the absence and presence of the inducer HMBA for 5 days and assayed for luciferase activity (triplicate assays for each pool). The fold enhancement (luciferase activity for the indicated construct divided by the luciferase activity for ␤106 under the same condition, i.e. induced or uninduced) is shown at the top of each bar. a P E is the probability that the observed mean is not different from that with no enhancer, based on Student's t test with df, 4. b P HX/BX is the probability that observed expression levels from constructs with the HindIII-XbaI enhancer are not different from those with the BalI-XbaI enhancer. that this segment can play a negative role.
A potential practical application of the identification of negative elements in the LCR HSs may be found in improving vectors for globin gene therapy. Inclusion of LCR HSs in expression constructs, including retroviral vectors, can greatly increase the level of expression of the target globin gene (21,25,73,74), but the set of DNA fragments needed for optimal expression has not yet been defined (4). Our characterization of negative elements within the conventional HS2 and the indication that they also may be operative in transgenic mice (36) raise the possibility that re-engineering LCR constructs to remove all such negative regions could generate an even more potent enhancer of expression.