In Vitro Chromatin Assembly of the HIV-1 Promoter

Nuclease hypersensitive sites exist in vivo in the chromatin of the integrated human immunodeficiency virus (HIV)-1 proviral genome, in the 5′-long terminal repeat (LTR) within the promoter/enhancer region near Sp1 and NFκB binding sites. Previous studies from the Kadonaga and Jones laboratories have shown that Sp1 and NFκB can establish hypersensitive sites in a truncated form of this LTR when added before in vitro chromatin assembly with Drosophila extracts, thus facilitating subsequent transcriptional activation of a linked reporter gene upon the association of additional factors (Pazin, M. J., Sheridan, P. L., Cannon, K., Cao, Z., Keck, J. G., Kadanaga, J. T., and Jones, K. A. (1996) Genes & Dev. 10, 37–49). Here we assess the role of a full-length LTR and 1 kilobase pair of downstream flanking HIV sequences in chromatin remodeling when these transcription factors are added after chromatin assembly. Using Xenopus laevis oocyte extracts to assemble chromatin in vitro, we have confirmed that Sp1 and NFκB can indeed induce sites hypersensitive to DNase I, micrococcal nuclease, or restriction enzymes on either side of factor binding sites in chromatin but not naked DNA. We extend these earlier studies by demonstrating that the process is ATP-dependent when the factors are added after chromatin assembly and that histone H1, AP1, TBP, or Tat had no effect on hypersensitive site formation. Furthermore, we have found that nucleosomes upstream of NFκB sites are rotationally positioned prior to factor binding and that their translational frame is registered after binding NFκB. On the other hand, binding of Sp1 positions adjacent downstream nucleosome(s). We term this polar repositioning because each factor aligns nucleosomes only on one side of its binding sites. Mutational analysis and oligonucleotide competition each demonstrated that this remodeling required Sp1 and NFκB binding sites.

In its most fundamental form, packaging of DNA in the nucleus of a eukaryotic cell into chromatin starts with the creation of the polynucleosomal chain. Each nucleosome consists of a core particle consisting of the histone octamer and 146 bp 1 of DNA wrapped in the left-handed superhelical sense, and a linker region of variable length but generally less than 50 bp, which interacts with histone H1 and/or other accessory proteins (reviewed in Ref. 1). This polynucleosomal structure is further looped and folded into various higher order structures, leading to the functional sequestration of the genome in the interphase nucleus at a DNA concentration of approximately 50 mg per ml (reviewed in Refs. 2 and 3).
The packaging of DNA into chromatin adds levels of complexity to the problem of transcriptional regulation. Nucleosome formation often hinders the accessibility of transcription factors to occupy their binding sites within promoters (reviewed in Refs. 4 -6). On the other hand, formation of the nucleosome can in some instances increase the affinity or action of factor binding (7)(8)(9). In addition, transcriptional elongation is attenuated by the formation of the polynucleosomal chain, leading to further potential regulatory opportunities (reviewed in Ref. 10). Although it seems clear that the eukaryotic transcription apparatus has evolved to work through chromatin and not simply naked DNA, we are only now beginning to develop an understanding of these mechanisms (reviewed in Ref. 11).
Windows in chromatin that have already bound transcription factors, or that will allow their binding, are operationally detected as nuclease hypersensitive sites (reviewed in Ref. 12). Several studies have focused on the mechanism of how these sites are generated. Theoretically one might imagine that DNA replication might be required for transcription factors to invade nucleosomes at the time of chromatin reassembly (13), but the only evidence for this mechanism that has thus far been obtained deals with the establishment of transcriptional repression (14,15 and references therein). Rather, DNA replicationindependent mechanisms seem to have evolved to allow transcription factors access to DNA sequences complexed to histone octamers. At least three different "protein machines" have been identified, by either biochemical or genetic experiments (or both), that utilize ATP to facilitate loading of transcription factors onto promoters originally packaged as nucleosomes: (i) the SWI⅐SNF complex (16; reviewed in Refs. 17 and 18); (ii) NURF (nucleosome remodeling factor) (19,20); and (iii) RSC (remodeling the structure of chromatin), a complex at least 10-fold more abundant than SWI⅐SNF in yeast (21).
We report here studies on chromatin remodeling within the HIV-1 5Ј-LTR where the viral promoter and enhancer are located. We selected this system because the LTR has been well characterized with respect to transcription factor binding sites (reviewed in Refs. [22][23][24] and nuclease hypersensitive sites in chromatin (25,26). Furthermore, recently the laboratories of Kadonaga and Jones have shown that Sp1 and NFB can establish hypersensitive sites in a 0.42-kbp truncated-LTR linked to a reporter gene when added before in vitro chromatin assembly with Drosophila extracts (27). Here we have taken advantage of these findings to explore further the mechanism for hypersensitive site formation and the potential for repositioning of nucleosomes along a full-length LTR, including 1 kbp of downstream HIV sequences, when factors are added either before or after chromatin assembly. Using Xenopus laevis oocyte extracts to assemble chromatin in vitro, we have confirmed that Sp1 and NFB can indeed establish nuclease hypersensitive sites, even when added after chromatin assembly. We demonstrate that the process is ATP-dependent, that nucleosomes upstream of NFB sites are rotationally positioned, and that factor binding leads to a novel polar repositioning of flanking nucleosomes.

EXPERIMENTAL PROCEDURES
Plasmid Constructs and DNA Labeling-The HIV-1 SphI/SphI fragment of pHXB2 (28) was inserted into pNEB193 (lacking a AvaI polylinker site). The newly generated plasmid termed pWLTR11 contained nucleotides 1-1456 of the HIV-1 genome (5Ј-LTR and 5Ј fragment of the gag gene) and was 4202 bp in length. Plasmids pWLTR12 and pWLTR13 containing mutations within Sp1 and NFB binding sites, respectively (29), were constructed similarly. Fig. 1 shows the positions of factor binding sites, key restriction enzyme sites, and the sequences mutated. Single site labeling of circular DNA utilized the procedure of Shimamura et al. (30). Briefly, plasmids were linearized at a single cutting restriction enzyme site and treated with alkaline phosphatase. DNA was then purified by phenol/chloroform extraction followed by ethanol precipitation and labeled with [␥-32 P]ATP using T4 polynucleotide kinase. DNA circles were generated in a T4 DNA ligasecatalyzed reaction. To remove radioactive label from non-ligated molecules, material was treated with alkaline phosphatase and DNA was purified by phenol/chloroform extraction followed by ethanol precipitation. Labeled DNA was dissolved in extraction buffer (see below) at 10 ng/l and stored at Ϫ80°C (used within 1-2 weeks without signs of significant radiolysis or nicking).
Preparation of Oocyte Extract and Assembly of Minichromosomes-Xenopus laevis oocyte extracts were prepared according to Shimamura et al. (30). Briefly, ovaries from adult frogs were treated for 3 h at room temperature with 0.15% collagenase type II (Sigma) in buffer OR-2 (5 mM HEPES, 1 mM Na 2 HPO 4 , 82 mM NaCl, 2.5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , pH 7.6). Dispersed oocytes were then washed with 10 changes of OR-2 buffer. Small oocytes were discarded, and large stage-6 oocytes were gently washed with three changes of ice-cold extraction buffer (20 mM HEPES, 5 mM KCl, 1.5 mM MgCl 2 , 1 mM EGTA, 10% glycerol, 10 mM ␤-glycerophosphate, 0.5 mM dithiothreitol, pH 7.6). An equal volume mixture of oocytes and extraction buffer was centrifuged in a Beckman SW55Ti rotor for 90 min at 40,000 rpm at 4°C. The interphase between the pellet and upper lipid layer was aliquoted and stored at Ϫ80°C. The protein concentration of extracts typically was about 3 mg/ml. Assembly and analyses of radioactive plasmid into chromatin was done according to the protocols of Worcel and co-workers (30 -33), in a final volume 10 -50 l. A 10-l reaction mixture consisted of 4 l of oocyte extract, 15-25 ng of labeled DNA (the optimal amount was experimentally determined for each DNA preparation), 10 ng of creatine phosphokinase, and 1 l of 10 ϫ chromatin assembly buffer (400 mM creatine phosphate, 20 mM ATP, 10 mM MgCl 2 ). All reagents were dissolved in extraction buffer (final pH 7.0). In some experiments chromatin was assembled in the presence of total histone H1 isolated from human placenta, final concentration 0.7 ng/l (that was about 2 histone H1 molecules per nucleosomal repeat). The mixture was then incubated for 5 h at 27°C. Where indicated, chromatin was assembled in the presence of recombinant human transcription factors. Sp1, p50, and c-Jun were purchased from Promega. p65 and p50/p65 were produced in a baculovirus expression system using standard techniques (34). p65 was 6xHis-tagged, which allowed us to purify the p65 homodimer and p50/p65 heterodimer using Ni 2ϩ -NTA-agarose columns (Qiagen). TBP and Tat were produced as glutathione S-transferasefusions in a bacterial expression system and purified as described previously (35). In some experiments, factors were added to DNA 15 min before the oocyte extract was added. Final concentrations of the factors in the assembly mixtures were as follows: 10 nM Sp1, 50 nM NFB (homo-or heterodimer), 50 nM AP1 (c-Jun homodimer), 50 nM TBP, and 100 nM Tat. In other experiments Sp1 (20 nM) and NFB (100 nM) were added after chromatin assembly, and the reaction was allowed to proceed for another 30 min. ATP was removed from reaction mixtures that contained assembled minichromosomes according to Tsukiyama et al. (36); mixtures were either treated with apyrase (final concentration 0.01 unit/l) for 15 min at 27°C or filtered through spin columns containing Bio-Gel A-1.5m (Bio-Rad).
Analysis of Nucleosome Formation-Supercoils introduced into DNA circles upon minichromosome formation were analyzed by agarose gel electrophoresis. Aliquots of assembly reactions were mixed with 0.5 volume of stop solution (0.6% SDS, 50 mM EDTA, and 6 mg/ml proteinase K) and incubated for 1 h at 50°C. Gel loading dye buffer was added, and samples were then run on 1.25% SeaKem agarose gels (1 ϫ TAE) for about 12 h at 5 V/cm at 4°C. Alternatively, electrophoresis was performed in the presence of 15 g/ml (30 mM) chloroquine. Some DNA samples were electrophoresed in two dimensions. After the first dimension, gels were soaked in chloroquine-containing buffer, turned 90°, and run in a second dimension. When electrophoresis was completed, gels were fixed in 1% cetyltrimethylammonium bromide, dried, and exposed to x-ray films or PhosphorImager screens. To analyze the nucleosomal ladders of minichromosomes, CaCl 2 (final concentration 3 mM) and micrococcal nuclease (Worthington, final concentration 0.15 unit/l) were added to the assembly mixture. Digestion proceeded 1-90 min at room temperature and was stopped by adding 0.5 volume of stop solution (see above). After incubation at 50°C samples were run on 1.5% SeaKem agarose gels (1 ϫ TAE), and gels were fixed, dried, and exposed as described above.
Nuclease Hypersensitivity Site Analyses-Chromatin was assembled using plasmids labeled at the PmeI restriction site. After assembly was completed, CaCl 2 and MgCl 2 (final concentration 3 and 5 mM, respectively) were added and then chromatin was digested with DNase I (Worthington, final concentration 8 units/ml) or micrococcal nuclease (MNase) (Worthington, final concentration 4 units/ml) for 2 or 4 min at 27°C. In parallel experiments, naked DNA was digested with DNase I (0.04 unit/ml) or micrococcal nuclease (2 units/ml). Digestion was terminated by adding stop solution. After incubation for 1 h at 50°C, DNA was purified by phenol/chloroform extraction followed by ethanol precipitation. Solubilized DNA samples were then digested with EcoRI and electrophoresed on 0.5% SeaKem, 2% NuSieve agarose gels (1 ϫ TAE). Before analyzing for the accessibility of restriction sites, NaCl and MgCl 2 (final concentration 50 and 7 mM, respectively) were added to assembled minichromosomes. Restriction enzyme digestion (final concentration 1 unit/l) was for 7, 15, or 30 min at 27°C. Reactions were terminated by adding stop solution, and DNA samples were purified as above, digested with EcoRI, and electrophoresed on 1.5% SeaKem agarose gels (1 ϫ TAE). Gels were fixed, dried, and exposed to x-ray films and PhosphorImager screens as described above. Radioactivity in restriction fragments was quantified using ImageQuant software (Molecular Dynamics).
DNase I Footprinting Assay-Chromatin was assembled using plasmids labeled at either the BglII or AvaI restriction sites. After assembly was completed, minichromosome and naked DNA samples were digested with DNase I as described above. Purified DNA was digested either with BglII and SstI or with AvaI and ScaI. Heat-denatured DNA was run on 8 M urea, 6% polyacrylamide sequencing gels. Gels were dried and exposed as described above.
Gel Mobility Shift Analysis-The oligonucleotides used in gel mobility shift assays were 66-mer duplexes spanning NFB and Sp1 binding sites (Ϫ111 to Ϫ46 bp), with either wild type or mutated sites. Binding reactions between end-labeled double-strand probes (200 pmol) and purified proteins (500 pmol) were carried out for 30 min at 27°C in a final volume of 20 l of binding buffer (20 mM HEPES, 50 mM NaCl, 1 mM dithiothreitol, 0.5 mM EDTA, 3 mM MgCl 2 , 10% glycerol, 0.05% Nonidet P-40, pH 8.0) in the presence of different amounts of specific unlabeled competitors. Samples were run on native 5% polyacrylamide gels (0.25 ϫ TBE) at 10 V/cm at 4°C. Gels were fixed, dried, and exposed as described above.

Xenopus Oocyte Extracts Assemble Bona Fide
Chromatin-To characterize the kinetics and efficiency of chromatin assembly, we assayed for the production of negative DNA supercoils with time of incubation in Xenopus extracts by separating resulting purified DNA circles on agarose gels in the absence or presence of chloroquine (Fig. 2, A and B). The reactions reached completion after 5 h of assembly at 27°C. To determine the writhe for fully assembled molecules we performed two-dimensional electrophoresis in the presence of a standard that contained an array of topoisomers differing in linking numbers by single integers. As shown in Fig. 2C, the number of negative DNA supercoils introduced into the 4.2-kbp circles exhibited a distribution with an average between Ϫ23 and Ϫ24. Assuming 1 negative supercoil per nucleosome (2), these circles possessed nucleosomes about every 175 to 185 bp. To investigate further the assembly process we digested the fully assembled minichromosomes with micrococcal nuclease (MNase) and estimated the DNA lengths of the purified DNA products by gel electrophoresis. As shown in Fig. 2D, the size of the protected mononucleosome DNA fragments is close to 146 bp, as expected for core particles, whereas oligonucleosomes exhibit repeat lengths between 165 and 175 bp, with an interband background indicative of slightly irregular spacing. Taken together, the DNA supercoiling and nucleosome ladder results indicate that the Xenopus oocyte system works to assemble bona fide chromatin in our hands.
ATP-dependent Induction of DNase I-hypersensitive Sites in the HIV-1 Promoter in Chromatin by Sp1 and NFB-To determine if Sp1 and NFB could create DNase I-hypersensitive sites within the HIV 5Ј-LTR in the Xenopus chromatin assembly system like they can in the Drosophila chromatin assembly system (27), these recombinant proteins were added individually or together before or after chromatin assembly. As shown in Fig. 3A, when added before chromatin assembly individually, each protein created hypersensitive sites apparently flanking both sides of their DNA binding sites within the LTR, and when added together the hypersensitive sites, which were centered at positions Ϫ40 and Ϫ160 bp (large arrowheads), exhibited a broadened intervening footprint (for purposes of presentation, unless otherwise stated, the term NFB will refer to the p50/p65 heterodimer). NFB also created hypersensitive sites in flanking vector sequences in a region coincidentally exhibiting a 9/11 bp match to its consensus binding site (small arrowheads). It is significant that all of these sites are chromatin-specific and do not appear when these factors are complexed with naked DNA (Fig. 3B), a point not addressed by the earlier study (27). Furthermore, if chromatin was first assembled in the absence of exogenous Sp1 and NFB, the later addition of these factors induced the hypersensitive sites in an ATP-dependent fashion (Fig. 3C, asterisks) (for purposes of presentation, when factors are added after chromatin assembly they are designated with asterisks). Although three different forms of NFB (p50/p65, p50/p50, and p65/p65) each stimulate in vitro transcription from the HIV enhancer on naked DNA (37,38), the p50/p50 form cannot activate transcription in vitro on chromatin templates (27). However, we found that all three of these forms of NFB were active in forming DNase I-hypersensitive sites when added either before or after chromatin assembly, in the presence or absence of Sp1 (data not shown; see below). We also tested recombinant forms of other transcription factors (TBP, AP1(c-Jun), and Tat) and found that these proteins (either alone or in combination with Sp1 and NFB) did not create any chromatin structural alterations ( Fig.  3A and data not shown). We conclude that Sp1 and NFB are key players in chromatin remodeling within the HIV 5Ј-LTR in the Xenopus system just as they are in the Drosophila system (27).
DNase I-hypersensitive Site Formation by Sp1 and NFB Requires Their DNA Binding Sites in Cis and Can Be Competed by an Excess of Their Binding Sites in Trans-To gain further evidence that Sp1 and NFB must bind to their target sites in the HIV 5Ј-LTR to create hypersensitive sites, we repeated the experiments shown in Fig. 3 but with DNA constructs that possessed base substitution mutations in the corresponding binding sites known to inhibit transcription in vivo (29) (see Fig. 1C). When DNA containing mutated Sp1-binding sites was mixed with Sp1 in the absence of NFB and chromatin was assembled, hypersensitive sites characteristic for Sp1 binding were not induced (Fig. 4). However, as expected, NFB was still able to create its characteristic hypersensitive sites in this mutant construct in the presence or absence of Sp1 (Fig. 4). When DNA containing mutated NFB-binding sites was mixed with Sp1 and assembled into chromatin, the specific pattern of DNase I hypersensitivity characteristic for Sp1 was present but weaker as compared with that of the wild type control, suggesting that the GC-rich nature of wild type NFB sites may help preserve Sp1 occupancy after chromatin assembly (Fig. 4). NFB was still able to create weak hypersensitive sites spaced even broader than those characteristic for NFB binding per se, even though its binding sites had been mutated, presumably because residual half-sites were still present in this mutant and secondary binding occurred through sequence similarities between NFB and Sp1 binding sites ( Fig. 4; see Fig. 1C). Gel mobility experiments using a 66-mer duplex spanning these mutations verified that this sequence still had residual affinity to bind NFB (data not shown). Interestingly, when both factors were combined with the construct carrying the mutated NFB sites and chromatin was assembled, hypersensitive sites characteristic (although somewhat weaker) of the wild type construct were generated (Fig. 4). Apparently the known cooperative interactions between Sp1 and NFB (39,40) leads to synergism in occupancy of the weakened binding sites. We also performed competition experiments with 33-mer duplex oligomers containing wild type NFBor Sp1-binding sites (or both). The presence of both competitors completely abolished the formation of these hypersensitive sites, while either alone ablated the formation of its target hypersensitive sites (data not shown).
Rotational Positioning of Nucleosomes before and after Occupancy of Sp1 and NFB Binding Sites-To investigate the possibility of rotational positioning of nucleosomal DNA on the surface of histone octamers as well as prove that Sp1 and NFB actually target their binding sites in the LTR when added either before or after chromatin assembly, we separated DNase I digests on sequencing gels. When NFB bound to naked DNA it protected from Ϫ109 to Ϫ77 nt corresponding to its two binding sites (non-coding strand), whereas Sp1 protected from Ϫ87 to Ϫ64 nt corresponding to the highest affinity binding site of the triplet array (41) (Fig. 5A). Binding of both factors gave the expected additive footprint spanning from Ϫ109 to Ϫ64 nt, in either naked DNA or chromatin (Fig. 5, A and B). Very similar footprints were seen on chromatin substrates with the three different forms of NFB whether the factors were added before or after nucleosome assembly ( Fig. 5B; data not shown). We conclude that these transcription factors can target their binding sites in chromatin. In contrast to the similarities in the footprinted regions in these factor binding sites between naked DNA and chromatin, the chromatin but not naked DNA samples exhibited approximately a 10-nt periodicity in DNase I cutting upstream of Ϫ220 nt (Fig. 5, B and C, closed circles), even in the absence of added factors, immediately adjacent to a short footprinted region spanning Ϫ205 to Ϫ220 nt (Fig. 5, B and C, X), suggesting the presence of rotationally positioned nucleosomes. It is striking that this presumptive rotational positioning extends beyond the limits of the 5Ј-LTR (Ϫ454 nt) onto flanking vector sequences (Fig. 5C, closed circles). In contrast to these results, no rotational positioning was observed downstream of Sp1 binding sites, in the presence or absence of Sp1 or NFB (or both) (Fig. 5B, data not shown).
Polar Repositioning of Nucleosomes by Sp1 and NFB-To determine if nucleosome positions within the LTR are modulated by Sp1 and NFB binding when the factors are added either before or after chromatin assembly, we mapped the positions of MNase cutting sites. In the absence of these factors, almost all MNase cutting sites were common between the naked DNA controls and chromatin samples (Fig. 6, lanes  11-13 and 18 -20), indicating that the nucleosomes were not translationaly positioned along the DNA sequence. However, binding of Sp1 and NFB to DNA prior to chromatin assembly generated a pair of major MNase-hypersensitive sites in positions essentially identical to the DNase I-hypersensitive sites that flank factor binding sites (Fig. 6A, lanes 7-9; large arrowheads); these sites were not generated upon factor binding to naked DNA controls (Fig. 6B, lanes 14 -16). In addition, Sp1 and NFB generate minor MNase-hypersensitive sites at ϩ140 and Ϫ340 bp, respectively (Fig. 6A, lanes 3-6; closed circles, small arrowheads). Interestingly, these minor sites correspond to nucleosome length distances from the major hypersensitive sites, suggesting that binding of Sp1 and NFB each position adjacent nucleosomes, one upstream of NFB binding sites and the other downstream of Sp1 binding sites. Consistent with this interpretation is the observation that cutting sites in the regions of these putative positioned nucleosomes centered at positions Ϫ290 and ϩ40 bp become protected when factors bind (Fig. 6A, lanes 3-6; open circles, bars). It is significant that these same major and minor hypersensitive sites and footprints are generated when factors bind after chromatin assem- bly (Fig. 6C, compare lanes 21-23 with lanes 24 -26; closed  circles, arrowheads, open circles, and bars). Furthermore, MNase ladder experiments reveal that nucleosomes indeed reside on sequences downstream of the Sp1 binding sites (minichromosomes labeled at BglII site located at ϩ20 bp, data not shown). We conclude, therefore, that Sp1 and NFB each remodel chromatin structure by repositioning nucleosomes with polarity.
Quantitative Assessment of Hypersensitive Site Formation Using Restriction Enzymes-To more accurately judge the increase in nuclease accessibility generated by Sp1 and NFB binding before or after chromatin assembly, we digested minichromosomes to completion with several different restriction enzymes and determined the cleavage efficiencies. As shown in Fig. 7, binding of Sp1 to DNA before chromatin assembly increased the accessibility of the PvuII site located at Ϫ21 bp downstream of Sp1-binding sites but did not affect the efficiency of cleavage at the ScaI site located at Ϫ140 bp upstream of NFB-binding sites. However, binding of NFB to DNA before chromatin assembly increased the accessibility at both these sites. Maximal cutting at both sites occurred when both Sp1 and NFB were added prior to chromatin assembly (Fig. 7) or when Sp1 was added prior to and NFB was added after chromatin assembly (data not shown). Adding both factors after chromatin assembly (Fig. 7) or NFB prior to followed by Sp1 after chromatin assembly (data not shown) were less efficient in increasing accessibility. Cleavage results obtained at this ScaI site were very similar to those exhibited at the AvaI site at Ϫ158 bp (data not shown). Furthermore, no significant differences were observed when the different forms of NFB were used in similar experiments (data not shown). In addition, increased restriction site accessibility was lost as expected when constructs containing mutated binding sites were used or when ATP was depleted prior to adding factors to chromatin assembled on the wild type construct (data not shown). Similar studies at numerous other sites revealed no increase in accessibility upon factor binding (see Fig. 7 legend), except for an increase in cutting at HinfI (ϩ124 bp) caused by Sp1 (data not shown), correlating well with the 3Ј border of the positioned nucleosome detected by MNase cleavage experiments (Fig. 6). In conclusion, the results obtained with restriction enzymes are in complete agreement with the results obtained using both DNase I and MNase as probes for chromatin structure.

DISCUSSION
Our results on how the Xenopus system designs or remodels the chromatin structure of the HIV-1 promoter/enhancer and flanking regions in response to human transcription factors are schematically presented in Fig. 8. In the absence of transcription factors, the nucleosomes along the HIV-1 5Ј-LTR are rotationally positioned upstream of Ϫ220 bp (R ovals) but randomly positioned elsewhere. Binding of NFB results in translational positioning of these upstream nucleosomes while still maintaining rotational positioning (RT ovals), while binding of Sp1 translationally positions a downstream nucleosome (T oval). Each factor creates hypersensitive nuclease cutting sites on either side of their binding sites in chromatin but not in naked DNA. Together these factors create a local chromatin structure that resembles the in vivo state (see below).
Our study extends the work of the laboratories of Kadonaga and Jones who used Drosophila extracts for chromatin assembly experiments (27). We have confirmed that Sp1 and NFB create hypersensitive sites and have been able to examine repositioning of nucleosomes on flanking bona fide HIV sequences that was not studied previously. We show that this process is ATP-and target site-dependent, and we have quantitated site generation using restriction enzymes. In the Kadonaga, Jones, and co-workers' study (27), chromatin structural analysis was only performed when factors were bound to DNA before chromatin assembly, not after; in addition, nucleosomes were already positioned downstream of Sp1 binding sites on the flanking vector sequences in the absence of this transcription factor, and no change in positioning was detected upon prebinding Sp1 per se; this is not the case in the Xenopus system with HIV sequences, because Sp1 binding, even after chromatin assembly, leads to polar repositioning of downstream nucleosome(s). The sensitivity of our studies also permitted observing rotational positioning upstream of NFB sites not looked for in the earlier work. Finally, repositioning of nucleosomes by NFB per se was not studied previously, which we found here also to exhibit polarity.
Statistical positioning theory predicts that occupied factor binding sites would create boundary constraints that would exclude nucleosomes and passively lead to positioning on both sides of the boundary (42,43). However, we observed positioning only downstream of occupied Sp1 sites and only upstream of occupied NFB sites. Furthermore, in other in vitro chromatin assembly experiments using the mouse immunoglobulin gene, we have observed the same polarity of nucleosome positioning with respect to the orientation of the NFB binding site adjacent to the intronic enhancer (data not shown). This polarity suggests that specific segments of these transcription factors interact with specific segments of the histones. This positioning is absolutely precise with respect to the DNA helix for the rotationally positioned nucleosomes upstream of NFB sites (Fig. 8, RT ovals). Possibly the footprinted region at Ϫ205 to Ϫ220 nt ( Fig. 5; Fig. 8, black lines below) sets up this positioning because of a need for a specific exposed helix surface to bind factors in the oocyte extracts; this footprinted sequence is AT-rich and has been reported to bind NF-AT (23,24) as well as nuclear matrix protein(s) (44). The interacting Xenopus protein(s) remains to be identified as well as the significance, if any, of such rotational positioning.
Chromatin structure appears to be a major control point in regulating HIV-1 gene expression. In vivo analysis has revealed hypersensitive sites near Sp1 and NFB binding sites for integrated provirus exhibiting low basal expression, similar to but apparently somewhat broader than the zone described in our studies, in cell lines in which NFB is thought not to have been activated (25,26). Perhaps the transient activation of this factor can create sites that become "memorized" after factor disappearance, and this chromatin structure is templated to daughter cells; alternatively, the promoter region possesses a battery of other factor binding sites, including TFIID, LEF-1, Ets-1, and USF (see Fig. 1A). Binding of some of these factors may be responsible for generating and/or broadening the hypersensitive site in vivo (45), whereas AP-3-like, DBF1, possibly together with Sp1 binding may create another downstream site not seen in our studies (46).
Previous in vitro transcription studies on chromatin templates have shown that the p65 form of NFB, which possesses a transactivation domain, is required for recruitment of the basal machinery for synergistic transcriptional activation with Sp1 (27). In addition, direct protein interactions between Sp1 and p65 have been identified and shown to be required for efficient activation of HIV transcription (39,40). Interestingly, we found that Sp1 binding to DNA before chromatin assembly potentiated the ability of NFB to generate hypersensitive sites FIG. 7. Digestion of reconstituted chromatin with restriction nucleases. A, the position of key restriction sites (arrows), factor binding sites (boxes), and DNase I-hypersensitive sites (DHS) (black bars); B, effects of Sp1 and NFB upon accessibility of ScaI and PvuII restriction sites. The percent of cleavage upon reaching complete digestion is shown below the indicated restriction fragments. Factors added after assembly are marked with asterisks. In experiments where chromatin was formed in the presence of Sp1, NFB, and AP1, the susceptibility of the following restriction sites for cleavage did not change significantly: PstI (Ϫ476; vector polylinker), EcoRV (Ϫ341), BsaI (ϩ2), BglII (ϩ20), SacI/SstI (ϩ34), HindIII (ϩ78), NarI (ϩ185), SacI/SstI (ϩ226), HinfI (ϩ243), XmnI (ϩ386), HinfI (ϩ465), PstI (ϩ505), SspI (ϩ514), StyI (ϩ627), HindIII (ϩ633), PvuII (ϩ693 and ϩ699), and PstI (ϩ696). after chromatin assembly and observed that Sp1 enhanced the ability of NFB to bind to partially mutated NFB sites (Fig. 4).
Activation of HIV-1 infected T-cells leading to subsequent increases in HIV-1 transcription results in rearrangement or displacement of the repressing nucleosome immediately downstream of the transcription start site (26). Our in vitro assembly conditions did not result in this disruption. Further analysis will be needed to determine whether this change in chromatin structure is involved in activation of transcription by the HIV-1 transactivator Tat. Nucleotide level resolution revealed that the region surrounding the TATA box did not become footprinted after chromatin assembly in the presence of a mixture of transcription factors (data not shown), indicating that TFIID complexes did not form over the core promoter. It is possible that Dr1, a TBP-binding repressor present in Xenopus oocytes, may account for this fact (47). In addition, we have found that including ␣-amanitin (final concentration 0.1 M) had no effect on hypersensitive site formation when Sp1 and NFB were added either before or after minichromosome formation (data not shown), thus arguing that hypersensitive site formation occurs independent of transcription.
An interesting unanswered question is whether histones still remain bound within the hypersensitive region of the HIV-1 5Ј-LTR after ATP-dependent nucleosome disruption by factor binding. In a model in vitro system Owen-Hughes et al. (16) have shown that the SWI⅐SNF complex can displace histones from a GAL4-occupied hypersensitive site and that the hypersensitivity remained even after removal of bound transcription factors by oligonucleotide competition. We attempted related experiments in the Xenopus system with the HIV LTR, where we performed gel mobility analysis on the hypersensitive region after its excision by restriction enzymes from purified minichromosomes. Although the hypersensitive region exhibited a gel mobility indistinguishable from naked DNA associated with Sp1 and NFB, after oligonucleotide competition this complex was converted into others and not simply to either naked DNA or nucleosome bands (data not shown).
The Xenopus oocyte extracts used in our chromatin assembly experiments contain only an embryonic form of histone H1, termed B4 (48), whose binding to nucleosomes is insufficient to repress the transcription of certain genes that can be repressed by adult histone H1 homologs during development (49). Therefore, we have repeated many of these experiments with chromatin assembled with added histone H1 and found similar remodeling whether the factors were added before or after assembly, by DNase I-hypersensitive site analysis, MNase footprinting assays, and by the observation of disruption of the MNase ladder for minichromosomes labeled at the AvaI site located at Ϫ158 bp, and by noting an increase in accessibility to restriction enzyme sites at Ϫ21 bp (PvuII) and Ϫ140 bp (ScaI) (data not shown). This is in confirmation with the results of Kadonaga, Jones, and co-workers (27) who used histone H1 containing chromatin. Thus, even though histone H1 is known to be functionally important in repressing gene expression and reducing nucleosome mobility (49,50), the ATP-dependent remodeling activity in oocyte extracts was fully capable of overriding any potential inhibition by this protein. Clearly, the nature of the remodeling activity remains to be elucidated but may be amenable to study in these extracts.