A major role for the TATA box in recruitment of chromatin modifying complexes to a globin gene promoter

Abstract

The developmentally regulated mammalian β-globin genes are activated by a distant locus control region/enhancer. To understand the role of chromatin remodeling complexes in this activation, we used stably replicated chromatin templates, in which transcription activation of the human embryonic ε-globin gene depends on the tandem Maf-recognition elements (MAREs) within the β-globin locus control region HS2 enhancer, to which the erythroid factor NF-E2 binds. The HS2 MAREs are required for nucleosome mobilization and histone hyperacetylation at the distant promoter. Nucleosome mobilization also requires the promoter TATA box, and is independent of histone hyperacetylation. In contrast, promoter hyperacetylation requires the promoter GATA-1, and CACC-factor activator motifs, as well as the TATA box. ChIP analysis reveals that NF-E2 is associated with the active ε-globin promoter, which lacks an NF-E2 binding sequence, in a TATA box and HS2/MARE-dependent fashion. NF-E2 association with the ε-globin promoter coincides with that of RNA polymerase II at both regulatory sites. The results emphasize MARE–TATA box interactions in the recruitment of complexes modifying promoter chromatin for transcription activation and imply close physical interaction between widely separated regulatory sequences mediated through these sites.

The basic unit of chromatin in vivo, the nucleosome, is inhibitory to transcription. Transcriptional activation involves relief of this repression by ATP-dependent SWI/SNF type nucleosome remodeling complexes (1), which alter nucleosome structure and/or position, and histone acetyltransferase (HAT) complexes (2), which acetylate N-terminal lysine residues of histone tails. The actions of these complexes are interdependent (3, 4), but the order in which they act is not invariant. In yeast, recruitment of SWI/SNF precedes histone acetylation by the SAGA complex at the cell-cycle-regulated HO gene (5), whereas histone acetylation precedes SWI/SNF-dependent remodeling at the PHO5 and PHO8 genes (6, 7). Diverse results have also been reported for mammalian promoters (8, 9). Indeed, remodeling itself is not essential for recruitment of the preinitiation complex to a promoter (10).

The paradigm for transcriptional activation in yeast is recruitment of remodeling complexes to promoters by transcription factors which bind within or nearby, at upstream activating sequences (UASs) (11). However, yeast UASs are highly distance dependent, whereas locus control region (LCR) enhancer sequences in metazoans may be tens of kilobases away from the genes they activate; yet LCR/enhancers act to relieve nucleosome repression at their distant target promoters, resulting in DNase I hypersensitive (HS) site formation and histone hyperacetylation that correspond with transcription activation (12–15). In the human β-globin locus, LCR sequences per se appear to be responsible for these changes (12, 16), whereas in the mouse the sequences responsible may be even more distant from the globin genes than the classically defined LCR (17). Although these data imply that the LCR or other distant sequences recruit remodeling complexes that modify promoters, it is not yet clear how the recruitment model can be extended to gene activation from a distance.

The five members, ε, ^Aγ, ^Gγ, δ, and β, of the human β-globin family of genes are expressed sequentially during development and depend for their high-level expression on the β-globin LCR (16). The LCR is composed of four core regions (HS 1–4), located between 10 and 50 Kb upstream of the globin genes, that are hypersensitive to DNase I in erythroid cell chromatin (18). These sites all contain Maf-recognition elements (MAREs, recognized by NF-E2 and other factors), and a small number of transcription factor-binding motifs that are common with the globin gene promoters (19). These motifs include GATA-1 and CACC [the latter recognized by proteins such as erythroid Krüppel-like factor (EKLF)]. Homo- and heteromeric interactions among these transcription factors have been proposed to contribute to enhancer-promoter communication as described in models invoking looping or linking of distant DNA sequences (20–22), or tracking along intervening DNA (23).

The enhancer-binding protein NF-E2 is a heterodimer consisting of erythroid-specific p45 and ubiquitous p18 (MafK) (24–26). NF-E2 recognizes an extended AP-1/MARE motif in the LCR HSs and is a critical component of HS2 enhancer activity (27–29). Interestingly, NF-E2 is detected by chromatin immunoprecipitation (ChIP) at the active murine β-globin promoter, despite the absence there of a canonical NF-E2 motif (26, 30). NF-E2 interacts with CREB-binding protein (CBP)/p300 (31), a coactivator HAT found stably associated with a fraction of RNA polymerase II (pol II) holoenzyme (32–34), and with TAF _II 130 (35), a TATA box-binding protein (TBP)-associated factor. This finding raises the possibility that NF-E2 may interact with transcription activators and general transcription factors to recruit pol II. Studies with CB3 cells, a murine erythroleukemia (MEL) cell line depleted for p45/NF-E2, show that pol II is associated with HS2 in a p45/NF-E2-independent fashion (36). Together, these observations raise intriguing questions as to the degree of NF-E2/pol II interdependence at the gene promoter and the LCR, and as to the definition of sequences in the promoter and LCR that are involved in these associations.

We have addressed these questions by using stably replicated chromatin templates in human erythroid cells. The minichromosome model allows high-resolution structural analyses of the chromatin transitions that accompany transcription activation in organisms as diverse as yeast and humans (37, 38). Transcription activation of a 4-Kb embryonic ε-globin gene on minichromosomes in K562 cells is HS2-dependent, and is accompanied by DNase I HS formation, increased restriction enzyme accessibility, and histone hyperacetylation at the promoter (39, 40). Here we show that enhancer-dependent remodeling results in nucleosome mobilization at the ε-globin promoter, and, like histone hyperacetylation, requires the HS2 MAREs. Nucleosome mobilization requires an intact TATA box, but not prior histone hyperacetylation, whereas acetylation of promoter histones requires the TATA box, and interaction of transcription activators at the promoter. In addition, NF-E2 associates with the ε-globin promoter in a TATA box-dependent manner, whereas pol II associates with HS2 in a MARE-dependent manner. These data emphasize MARE–TATA box interactions, possibly through intermediates such as CBP/p300 and TAF_II130, in the recruitment of complexes that alter promoter chromatin structure and activate transcription.

Materials and Methods

Minichromosome Construction, Cell Culture Conditions, and Transfection. Minichromosomes contained the Epstein–Barr origin of replication, ori P, the EBNA-1 nuclear antigen and a hygromycin^R cassette (41). The construction of minichromosomes carrying a 4-kb ε-globin gene with or without HS2 (2-kb distant) has been described (29). Clustered point mutations were introduced by site-directed mutagenesis into HS2 or the ε-globin promoter to eliminate binding of NF-E2, GATA-1, and CACC factors. The mutations, transfection conditions, and culture conditions of K562 cells have been described (29, 39). Clones studied in this work stably carried 10–40 copies of minichromosomes.

RNase Protection Assay. RNA was prepared from 5 × 10⁶ K562 cells carrying various minichromosomes by using PUREscript (Gentra Systems). The episomal copy of the ε-globin gene is marked by a mutation in the 5′ UTR to distinguish its RNA transcripts from endogenous transcripts (29). RNase digestion and gel analysis were performed with an RPA II kit according to the manufacturer's recommendations (Ambion). The results were quantitated on a PhosphorImager (Molecular Dynamics). RNA levels were normalized to the actin signal and were corrected for minichromosome copy number. Endogenous ε-globin transcription is variable among clones and is unrelated to minichromosomal transcription.

Preparation of Nuclei and Nuclease Digestion. Nuclei of K562 cell clones (1 × 10⁸ to 1.5 × 10⁸ cells) were digested with 0–200 units of micrococcal nuclease per ml for 5 min at room temperature. For restriction-enzyme cleavage analyses, nuclei of 5 × 10⁷ cells were incubated with restriction enzymes (250 units/ml each enzyme) at 37°C for 30 min (29). DNA was purified, cut to completion with BglII, and analyzed by Southern blotting and hybridization with an internal XbaI–EcoRV fragment. The percent AvaII cutting [cut/(uncut + cut)] was quantitated on a PhosphorImager

ChIP and Analysis of DNA. Immunoprecipitation was performed by using modifications of published methods (42, 43) as described in detail (40). Input chromatin consisted predominantly of mononucleosome-sized fragments produced by micrococcal nuclease (MNase) digestion, and the number of PCR cycles was chosen to be within the linear response range (see Fig. 6, which is published as supporting information on the PNAS web site, www.pnas.org). Antibodies used were as follows: anti-diacetylated histone H3, anti-tetraacetylated H4 (Upstate Biotechnology, Lake Placid, NY), and anti-NF-E2, anti-pol II, and anti-BAF 155 (Santa Cruz Biotechnology). For NF-E2, pol II, and BAF 155 ChIPs, nuclei were first crosslinked with paraformaldehyde and mononucleosomes were gradient purified as described (44). PCR amplification, labeling of primers, and primer sequences used to amplify N1–, N2–, HS2, and ampicillin sequences in the plasmid backbone have been described (40). The amplification products (60–80 bp) were electrophoresed on 10% native polyacrylamide gels and the signals were quantitated on a PhosphorImager (Molecular Dynamics) by using IMAGEQUANT software.

Results and Discussion

Nucleosome Mobilization at the ε-Globin Promoter Depends on the HS2 MAREs. Previously, we had shown that an inactive ε-globin gene on minichromosomes was present in closed chromatin and contained histones that were not hyperacetylated, whereas the promoter of a transcriptionally active ε-globin gene linked to HS2 contained hyperacetylated histones H3 and H4 (40), indicating that one function of the enhancer is to recruit HAT activity to the promoter. These promoters were also highly accessible to restriction enzyme cleavage.

Complexes of the SWI/SNF type use the energy of ATP hydrolysis to alter nucleosome accessibility and/or mobility (45–47). We used MNase ladder disruption assays to investigate nucleosome mobilization at the ε-globin promoter. In this assay, a distinct ladder of mononucleosome and oligonucleosome fragments indicates that probe sequences are protected by a stable nucleosome. A smeared ladder indicates invasion and cleavage of the DNA corresponding to the probe by MNase, disruption of histone–DNA contacts, and mobilization of the nucleosome. We compared wild-type HS2ε genes with HS2ε genes in which the tandem HS2 MARE sites had been mutated, inactivating the enhancer (29). When MNase ladders were probed with promoter-proximal N1– sequences (Fig. 1B), actively transcribing HS2ε revealed a smeared pattern, which is indicative of nucleosome mobilization. In contrast, the MARE mutant MNase ladder remained distinct. When the same blot was reprobed with sequences distal to the promoter in nucleosomes N2–/N3–, clear nucleosomal ladders were observed for both clones. These data corroborate our earlier observations using indirect end labeling (29). We conclude that NF-E2 interacting at HS2 mediates mobilization of the promoter proximal nucleosome N1–, presumably by participating in recruitment of a SWI/SNF-like complex.

Fig. 1.

Promoter nucleosome mobilization depends on HS2/NF-E2 sites. (A) Positions of nucleosomes over the ε-globin promoter as determined by indirect end labeling are shown as shaded ovals (29). The N1– sequence is expanded to show the transcription activator sites it encompasses (shaded boxes). Filled bars represent probes within nucleosome N1–,orN2–/N3– used for hybridization to Southern blots of MNase ladders. (B) MNase ladders from nuclei of cells carrying either wild-type HS2ε or HS2εΔNF-E2 with disrupted NF-E2-binding sites, hybridized with the N1– probe, and subsequently hybridized with the N2–/N3– probe. (C) Representative ethidium bromide-stained agarose gel of CsCl₂ gradient fractions of MNase-digested chromatin. Lanes 1–4 illustrate fractions that were pooled and used as input DNA for ChIPs. M, markers in base pairs. (D) Linearity of the response range of PCR amplification signals with respect to input DNA. (E) Representative results of ChIPs performed with wild-type HS2ε and HS2εΔNF-E2 using antibodies to BAF 155. Specifically immunoprecipitated DNA (Ab) along with 0.02% of input DNA (In) was analyzed by PCR amplification using labeled primers within N1–,N2–, or HS2 sequences. To control for DNA sample size and minichromosome copy number, the immunoprecipitate and control (no, no antibody) signals were normalized to the input signal. Bars depict the results of three experiments ± SEM.

To confirm interaction of human SWI/SNF components, we performed ChIPs on wild-type HS2ε and MARE mutant clones. Nuclei were crosslinked with paraformaldehyde, digested with MNase, and gradient purified. Mononucleosomal fractions (Fig. 1C) were pooled and reacted with antibodies to BAF 155, a component of human SWI/SNF, and of the ERC-1 erythroid-specific remodeling complex (48). The resultant DNA was amplified by quantitative PCR (Fig. 1D) using labeled primers within N1–. To control for DNA sample size and minichromosome copy number, the immunoprecipitate signal was normalized to the input signal. BAF 155 is recruited to HS2 in a MARE-dependent manner (Fig. 1E) and its association at the ε-globin promoter is reduced by more than half in the absence of intact HS2 MAREs. At N2– sequences, between HS2 and the TATA box, detection of BAF155 is almost 4-fold lower than at N1–, and without the MAREs, BAF 155 is at background level. Thus, recruitment of BAF155 to HS2 and the proximal promoter is heavily dependent on the HS2 MAREs.

ε-Globin Promoter Nucleosome Mobilization Requires the TATA Box. These results indicate that a functional enhancer is important for both HAT and SWI/SNF activities at the promoter. To try to dissociate and independently observe the activities of SWI/SNF and HAT histone-modifying complexes, we studied ε-globin genes with mutations at promoter transcription-activator sites that weaken enhancer-promoter communication. These sites (see Fig. 1 A) include the conserved GATA-1 site at –165, the CACC motif at –110, the TATA box (39), and a +6 GATA-1 motif within a putative initiator element of the core promoter (49). The mutant genes were linked to a wild-type HS2 located 2.2 kb upstream of the promoter. RNase protection analysis of ε-globin transcription (Fig. 2A) indicated that all of the promoter mutations were deleterious to transcription to varying degrees (10–30% of the wild-type level), but that no single mutation completely eliminated transcription. In parallel with the diminution in transcription, accessibility of the promoter AvaII site (see Fig. 1 A) was reduced to 10–20% of the wild-type (Fig. 2B). Thus, the mutations resulted in a structural and functional state of the promoter that was intermediate to an inactive and a fully active promoter.

Fig. 2.

TATA box and activator motif mutations decrease ε-globin transcription and promoter accessibility. The promoter transcription activator motifs indicated in Fig. 1 A were destroyed by clustered point mutations (38). (A) Representative RNase protection analysis of RNA from K562 clones carrying HS2ε-globin genes with mutations in promoter GATA, CACC, or TATA motifs. Controls include wild-type HS2ε, an enhancerless ε-globin gene (ε), and one linked to MARE mutant HS2 (ΔNF-E2). The endogenous and minichromosomal ε-globin signals are indicated. M, markers between 75 and 298 bp. (B) Representative analysis of AvaII accessibility within N1– (see Fig. 1A) for the mutant clones. M, markers between 1.6 and 3 kb. (C) The RNase protection results illustrated in A are depicted graphically on the left (averages of two or three determinations for each clone ± SEM). Quantitation of the AvaII cutting [cut/(uncut + cut)] illustrated in B is depicted graphically at the right (two or three determinations for each clone ± SEM).

A considerable body of evidence indicates a role for promoter transcription activators in targeting chromatin-modifying complexes (50, 51). We probed MNase ladders generated by the promoter mutant clones with N1– sequences and control N2–/N3– sequences to determine whether activator binding participated in targeting SWI/SNF activity to the ε-globin promoter. When probed with N1– sequences (Fig. 3A), the GATA-1 and CACC mutant promoters displayed a smeared pattern similar to that of wild-type HS2, whereas the TATA mutant promoter, like the enhancerless gene and the MARE mutant clone (Fig. 1B), retained a distinct ladder pattern. The smeared or disrupted pattern observed with the GATA and CACC mutant promoters was not due to degradation of the DNA, because all yielded substantially nucleosomal ladders when probed with N2–/N3– sequences (Fig. 3B; for scans, see Fig. 7, which is published as supporting information on the PNAS web site). We conclude that nucleosome mobilization at the ε-globin promoter requires HS2 and the TATA box, but not the individual promoter transcription activator sites. Although we cannot exclude the possibility that a subset of transcription factor binding sites is sufficient for SWI/SNF recruitment to the promoter, a double mutant lacking both the –165 GATA-1 and the CACC sites demonstrated a smeared nucleosome ladder at N1– (see Fig. 8, which is published as supporting information on the PNAS web site).

Fig. 3.

Promoter nucleosome mobilization requires the TATA box. MNase ladders were generated from nuclei of cells carrying either wild-type, HS2ε,or HS2ε-globin genes with mutations in promoter GATA, CACC, or TATA motifs. (A) Hybridization was performed with the N1– probe. imagequant data and excel and sigmaplot software were used to produce scans of the most highly digested lane for each of the clones. (B) Blots were stripped and rehybridized with the N2–/N3– probe. For scans, see Fig. 7.

That SWI/SNF recruitment did not depend on promoter-bound transcription activators is unusual (50), and implies that the TATA box can be recognized by TBP, the TATA-binding component of TFIID, before promoter remodeling. This possibility is supported by recent evidence that silenced chromatin does not always exclude preinitiation complex formation on a promoter (52), and that a complete preinitiation complex including pol II was assembled at the α₁-antitrypsin promoter before promoter remodeling had occurred (10). Despite these examples, it remains unclear how TBP would discriminate among inactive promoters to recognize its target.

Nucleosome mobilization at N1– was concomitant with a partial increase in restriction enzyme accessibility, and low-level transcription of the GATA and CACC mutant promoters. However, it was not the result of transcription per se, because the double –165 GATA/CACC mutant gene displayed undetectable transcription (see Fig. 6). The modest change in restriction-enzyme accessibility of these mutant promoters on mobilization may be attributable to the central location of the AvaII site near the dyad axis of N1–. The results suggest that enhancer-dependent recruitment of SWI/SNF may be an initial remodeling event at a globin promoter facilitating factor binding, which is consistent with the observation that in the absence of SWI/SNF, EKLF displays negligible binding to the β-globin promoter CACC site on chromatin templates in vitro (53).

Promoter Histone Hyperacetylation Requires the TATA Box and GATA-1 and CACC Factors. To determine the requirements for promoter histone acetylation, we performed ChIPs on the promoter mutant clones. Soluble chromatin was precipitated with antibodies directed against diacetylated H3 or hyperacetylated-H4. The resultant DNA was amplified by quantitative PCR using labeled primers within N1– sequences, or within control ampicillin sequences in the episome backbone. Fig. 4A shows that, compared with HS2ε, both the enhancerless gene and one linked to an inactivated HS2 with mutated MAREs, display greatly reduced levels of acetylated histones at N1–. Strikingly, all of the mutant promoters displayed levels of H3 and H4 acetylation at N1– similar to the inactive promoters. In contrast, at control ampicillin sequences there was equivalent histone acetylation in all clones. We conclude that hyperacetylation of H3 and H4 at N1– is strongly dependent on transcription activators. This result suggests that activators recruit, stabilize, or directly affect the activity of a HAT complex at the promoter (54).

Fig. 4.

Promoter H3 and H4 acetylation requires the transcription activators and the TATA box. Representative ChIPs performed on nuclei of cells carrying either wild-type HS2ε, or HS2ε-globin genes with mutations in promoter GATA, CACC, or TATA motifs using antibodies to diacetylated H3 (A) or hyperacetylated H4 (B). The results of three experiments with N1– (black bars, ± SEM) or ampicillin primers (gray bars, ± SEM) are depicted graphically. Band intensities for immunoprecipitate (Ab) and control (no, no antibody) samples are expressed relative to the signal from 0.5% of input (In) DNA to correct for amount of chromatin and minichromosome copy number and are normalized to wild-type HS2ε.

Figs. 3 and 4 together indicate that promoter nucleosome mobilization does not require prior histone acetylation and imply that mobilization may occur first at the ε-globin promoter. Among the genes for which detailed studies exist, there does not seem to be a required order in which SWI/SNF and HAT complexes function (55). Theoretically, either type of complex could alter the chromatin template to render it a better substrate for the other (4). SWI/SNF could loosen histone–DNA contacts, allowing transcription factors to interact with the promoter to recruit HATs, because surprisingly, in our study all of the GATA-1 and CACC motifs in the promoter were required for histone hyperacetylation. Promoter histone acetylation may stabilize in vivo the partially open nucleosomal state produced by SWI/SNF action. Alternatively, acetylation may stabilize the association of a remodeling complex at the promoter. In support of this possibility, histone acetylation increased retention of SWI/SNF at the adenovirus promoter in vitro (56). Hyperacetylation may be a later step in ε-globin gene activation, which is consistent with a role in signaling to additional factors needed to boost transcription directly (57). If this observation can be generalized, it suggests that histone hyperacetylation may then be a key rate-limiting step in high-level transcription activation.

NF-E2 and Pol II Association with the ε-Globin Promoter and HS2 Is Interdependent. The above results emphasize MARE–TATA box interactions in the recruitment of remodeling complexes. To investigate the roles of NF-E2 and pol II in this recruitment, we used paraformaldehyde crosslinked chromatin of mononucleosome size (see Fig. 1C) from wild-type HS2ε, and MARE and TATA mutant clones, and performed ChIPs with antibodies to NF-E2 and pol II. PCR amplification was carried out by using primers within N1–, HS2, and control ampicillin sequences. Fig. 5 A–C are representative results and the data are summarized graphically in D. As expected (26, 58, 59), NF-E2 was associated with HS2 in wild-type HS2ε. Mutation of the MAREs eliminated this interaction (Fig. 5A). NF-E2 was associated with N1– in actively transcribing ε-globin genes, as was found at the active β-globin gene (26, 30); interestingly, this association depended on the HS2 MAREs. Further, when the TATA box was mutated, NF-E2 retained association with HS2, but its detection at N1– was greatly reduced. There was no association of NF-E2 with ampicillin sequences.

Fig. 5.

NF-E2 and pol II association with HS2 and the ε-globin promoter. Representative ChIPs performed with wild-type HS2ε, HS2εΔNF-E2, and HS2εΔTATA using antibodies to NF-E2 (A) and pol II (B). Specific immunoprecipitates (Ab) and an aliquot of input DNA (0.02%, In) were amplified with primers within N1–, HS2, or control ampicillin sequences. Audiographic images of the products separated on 10% polyacrylamide gels are shown. (C) ChIP results with two additional HS2εΔTATA clones (denoted as ΔTATAa and ΔTATAb) and a clone with HS2 only (no linked globin gene), using antibodies to NF-E2 or pol II. (D) Graphical comparison of signals obtained by using HS2 and N1– primers and anti-NF-E2 (black bars), anti-pol II (gray bars), or no antibody (white bars) expressed relative to 0.02% the input. Results of three determinations ± SEM are presented.

These results suggest that NF-E2 requires the HS2 MAREs to access N1–. It has been proposed that NF-E2 could be recruited independently to the β-globin promoter at the endogenous murine locus, because in mice harboring a deletion of the β-globin LCR, association of NF-E2 with the promoter is still observed (60). However, in the LCR deletion mice, β-globin transcription, although very low, is still detectable, and promoter DNase I hypersensitivity and hyperacetylation are retained (17). Given that other sequences outside the LCR, possibly further upstream of the globin genes may be involved in remodeling of the mouse β-globin promoter, it is possible that promoter association of NF-E2 may depend on NF-E2 sites contained therein or elsewhere in the locus.

The association of NF-E2 with N1 in a TATA box-dependent fashion led us to investigate the distribution of pol II in our system. ChIPs with antibodies to pol II (Fig. 5B) revealed strong crosslinking to the promoter of wild-type HS2ε, whereas there was a lack of crosslinking for the nontranscribing MARE mutant clone, in agreement with the dependence on p45/NF-E2 of pol II crosslinking at the promoter of the β-globin gene in the mouse locus in MEL cells (36). Association of pol II with N1– sequences was eliminated when the TATA box was mutated and we did not detect pol II at ampicillin sequences.

We, like others, detected pol II association with HS2 (36). This association depended on the HS2 MAREs (Fig. 5B). Further, when the promoter TATA box was mutated, the pol II/HS2 association was greatly diminished. To extend this observation, we asked whether pol II could be recruited to HS2 in the absence of a linked globin gene, and we studied two additional TATA mutant clones. The results shown in Fig. 5C indicate that, in all cases, the loss of the TATA box greatly reduces detection of pol II at HS2 (4–5 fold, Fig. 5D). This loss occurs even though NF-E2 is detectable at HS2, albeit at somewhat reduced levels (Fig. 5D). The diminution (to ≈60% of the wild-type level) of NF-E2 binding is interesting in light of the critical role played by NF-E2 in HS2 formation, and in light of our previous finding (39) that restriction enzyme accessibility of HS2 decreased 60% when the TATA box of a linked gene was mutated. This result supports a direct interaction in chromatin between the promoter and enhancer. Overall, the results suggest that HS2 may not serve as an independent entry point for pol II. In summary, our results argue that association of pol II with HS2 is interdependent with promoter association, and association of NF-E2 at the promoter is interdependent with its HS2 association.

In contrast to these results, pol II recruitment to HS2 in the mouse globin locus was still observed in CB3 MEL cells depleted for p45/NF-E2 (36). However, this association may be nonetheless through the MAREs, because HS2 forms a DNase I HS site in these cells, and the MAREs are occupied by an unknown factor (61). Such a scenario is unlikely in our studies, because the HS2 NF-E2 mutation eliminates all binding to that site and abrogates DNase I HS site formation at HS2 (29). Strikingly, we did not observe substantial pol II/HS2 association in the absence of pol II/promoter association, whereas, in CB3 cells, pol II was associated with HS2, but not with the β-globin promoter. Thus, it appears that ablation of the HS2 MAREs, and depletion of the HS2 MARE-binding component p45/NF-E2 have different consequences that we may not be able to understand without knowing what occupies the MAREs in CB3 cells. It is also possible that in the mouse locus, the LCR HS sites do serve as independent entry points for pol II. Alternatively, aspects of the regulation of the ε-globin versus β-globin genes may differ.

We have used a model chromatin system in vivo to examine the recruitment of remodeling complexes to the promoter of a human globin gene when it is activated for transcription. A major role for the TATA box in enhancer-dependent recruitment of SWI/SNF and HAT activities to the promoter was found, raising the possibility that both types of complexes may be recruited in association with the pol II holoenzyme. NF-E2 and pol II were associated with both HS2 and the active ε-globin gene promoter. Detection of pol II at HS2 was greatly reduced in the absence of a linked globin gene with an intact TATA box and recruitment of NF-E2 to the promoter depended on the HS2 MARES. We suggest that HS2-bound NF-E2 becomes associated with the promoter and promoter-bound pol II becomes associated with HS2, because the enhancer and transcribing promoter are near to each other in chromatin, as a looping or linking model would predict. One possible participant in this interdependent association is the HAT CBP/p300, which interacts with both pol II and NF-E2, and that participates in bridging between the enhancer and promoter of the PSA gene (62). We consider it unlikely that the minichromosome structure per se dictates this juxtaposition because small clustered point mutations in either promoter or enhancer are sufficient to alter it. Extension of our observations in a model system suggests the conclusion that the LCR exists in close proximity to an actively transcribing globin gene. While this work was under review, strong physical evidence that this is the case has been presented (63).