Gamma rays and bleomycin nick DNA and reverse the DNase I sensitivity of beta-globin gene chromatin in vivo

The nucleolar shell provides anchoring sites for DNA untwisting

DNA underwinding (untwisting) is a crucial step in transcriptional activation. DNA underwinding occurs between the site where torque is generated by RNA polymerase (RNAP) and the site where the axial rotation of DNA is constrained. However, what constrains DNA axial rotation in the nucleus is yet unknown. Here, we show that the anchorage to the nuclear protein condensates constrains DNA axial rotation for DNA underwinding in the nucleolus. In situ super-resolution imaging of underwound DNA reveal that underwound DNA accumulates in the nucleolus, a nuclear condensate with a core-shell structure. Specifically, underwound DNA is distributed in the nucleolar core owing to RNA polymerase I (RNAPI) activities. Furthermore, underwound DNA in the core decreases when nucleolar shell components are prevented from binding to their recognition structure, G-quadruplex (G4). Taken together, these results suggest that the nucleolar shell provides anchoring sites that constrain DNA axial rotation for RNAPI-driven DNA underwinding in the core. Our findings will contribute to understanding how nuclear protein condensates make up constraints for the site-specific regulation of DNA underwinding and transcription.

Transcriptionally Active Chromatin-Lessons Learned from the Chicken Erythrocyte Chromatin Fractionation

The chicken erythrocyte model system has been valuable to the study of chromatin structure and function, specifically for genes involved in oxygen transport and the innate immune response. Several seminal features of transcriptionally active chromatin were discovered in this system. Davie and colleagues capitalized on the unique features of the chicken erythrocyte to separate and isolate transcriptionally active chromatin and silenced chromatin, using a powerful native fractionation procedure. Histone modifications, histone variants, atypical nucleosomes (U-shaped nucleosomes) and other chromatin structural features (open chromatin) were identified in these studies. More recently, the transcriptionally active chromosomal domains in the chicken erythrocyte genome were mapped by combining this chromatin fractionation method with next-generation DNA and RNA sequencing. The landscape of histone modifications relative to chromatin structural features in the chicken erythrocyte genome was reported in detail, including the first ever mapping of histone H4 asymmetrically dimethylated at Arg 3 (H4R3me2a) and histone H3 symmetrically dimethylated at Arg 2 (H3R2me2s), which are products of protein arginine methyltransferases (PRMTs) 1 and 5, respectively. PRMT1 is important in the establishment and maintenance of chicken erythrocyte transcriptionally active chromatin.

Intercalation of small molecules into DNA in chromatin is primarily controlled by superhelical constraint

The restricted access of regulatory factors to their binding sites on DNA wrapped around the nucleosomes is generally interpreted in terms of molecular shielding exerted by nucleosomal structure and internucleosomal interactions. Binding of proteins to DNA often includes intercalation of hydrophobic amino acids into the DNA. To assess the role of constrained superhelicity in limiting these interactions, we studied the binding of small molecule intercalators to chromatin in close to native conditions by laser scanning cytometry. We demonstrate that the nucleosome-constrained superhelical configuration of DNA is the main barrier to intercalation. As a result, intercalating compounds are virtually excluded from the nucleosome-occupied regions of the chromatin. Binding of intercalators to extranucleosomal regions is limited to a smaller degree, in line with the existence of net supercoiling in the regions comprising linker and nucleosome free DNA. Its relaxation by inducing as few as a single nick per ~50 kb increases intercalation in the entire chromatin loop, demonstrating the possibility for long-distance effects of regulatory potential.

Effects of DNA supercoiling on chromatin architecture

Disruptions in chromatin structure are necessary for the regulation of eukaryotic genomes, from remodelling of nucleosomes at the base pair level through to large-scale chromatin domains that are hundreds of kilobases in size. RNA polymerase is a powerful motor which, prevented from turning with the tight helical pitch of the DNA, generates over-wound DNA ahead of itself and under-wound DNA behind. Mounting evidence supports a central role for transcription-dependent DNA supercoiling in disrupting chromatin structure at all scales. This supercoiling changes the properties of the DNA helix in a manner that substantially alters the binding specificity of DNA binding proteins and complexes, including nucleosomes, polymerases, topoisomerases and transcription factors. For example, transient over-wound DNA destabilises nucleosome core particles ahead of a transcribing polymerase, whereas under-wound DNA facilitates pre-initiation complex formation, transcription factor binding and nucleosome core particle association behind the transcribing polymerase. Importantly, DNA supercoiling can also dissipate through DNA, even in a chromatinised context, to influence both local elements and large chromatin domains. We propose a model in which changes in unconstrained DNA supercoiling influences higher levels of chromatin organisation through the additive effects of DNA supercoiling on both DNA-protein and DNA-nucleosome interactions. This model links small-scale changes in DNA and chromatin to the higher-order fibre and large-scale chromatin structures, providing a mechanism relating gene regulation to chromatin architecture in vivo.

Effects of DNA supercoiling on chromatin architecture

Disruptions in chromatin structure are necessary for the regulation of eukaryotic genomes, from remodelling of nucleosomes at the base pair level through to large-scale chromatin domains that are hundreds of kilobases in size. RNA polymerase is a powerful motor which, prevented from turning with the tight helical pitch of the DNA, generates over-wound DNA ahead of itself and under-wound DNA behind. Mounting evidence supports a central role for transcription-dependent DNA supercoiling in disrupting chromatin structure at all scales. This supercoiling changes the properties of the DNA helix in a manner that substantially alters the binding specificity of DNA binding proteins and complexes, including nucleosomes, polymerases, topoisomerases and transcription factors. For example, transient over-wound DNA destabilises nucleosome core particles ahead of a transcribing polymerase, whereas under-wound DNA facilitates pre-initiation complex formation, transcription factor binding and nucleosome core particle association behind the transcribing polymerase. Importantly, DNA supercoiling can also dissipate through DNA, even in a chromatinised context, to influence both local elements and large chromatin domains. We propose a model in which changes in unconstrained DNA supercoiling influences higher levels of chromatin organisation through the additive effects of DNA supercoiling on both DNA-protein and DNA-nucleosome interactions. This model links small-scale changes in DNA and chromatin to the higher-order fibre and large-scale chromatin structures, providing a mechanism relating gene regulation to chromatin architecture in vivo.

DNA of a circular minichromosome linearized by restriction enzymes or other reagents is resistant to further cleavage: an influence of chromatin topology on the accessibility of DNA

The accessibility of DNA in chromatin is an essential factor in regulating its activities. We studied the accessibility of the DNA in a ∼170 kb circular minichromosome to DNA-cleaving reagents using pulsed-field gel electrophoresis and fibre-fluorescence in situ hybridization on combed DNA molecules. Only one of several potential sites in the minichromosome DNA was accessible to restriction enzymes in permeabilized cells, and in growing cells only a single site at an essentially random position was cut by poisoned topoisomerase II, neocarzinostatin and γ-radiation, which have multiple potential cleavage sites; further sites were then inaccessible in the linearized minichromosomes. Sequential exposure to combinations of these reagents also resulted in cleavage at only a single site. Minichromosome DNA containing single-strand breaks created by a nicking endonuclease to relax any unconstrained superhelicity was also cut at only a single position by a restriction enzyme. Further sites became accessible after ≥95% of histones H2A, H2B and H1, and most non-histone proteins were extracted. These observations suggest that a global rearrangement of the three-dimensional packing and interactions of nucleosomes occurs when a circular minichromosome is linearized and results in its DNA becoming inaccessible to probes.

Signal transduction pathways and the modification of chromatin structure

Mechanical and chemical signaling pathways are involved in transmitting information from the exterior of a cell to its chromatin. The mechanical signaling pathway consists of a tissue matrix system that links together the three-dimensional skeletal networks, the extracellular matrix, cytoskeleton, and karyoskeleton. The tissue matrix system governs cell and nuclear shape and forms a structural and functional connection between the cell periphery and chromatin. Further, this mechanical signaling pathway has a role in controlling cell cycle progression and gene expression. Chemical signaling pathways such as the Ras/mitogen-activated protein kinase (MAPK) pathway can stimulate the activity of kinases that modify transcription factors, nonhistone chromosomal proteins, and histones. Activation of the Ras/MAPK pathway results in the alteration of chromatin structure and gene expression. The tissue matrix and chemical signaling pathways are not independent and one signaling pathway can affect the other. In this chapter, we will review chromatin organization, histone variants and modifications, and the impact that signaling pathways have on chromatin structure and function.

Formation and repair of DNA double-strand breaks in gamma-irradiated K562 cells undergoing erythroid differentiation

Cellular differentiation is accompanied by gross changes in nuclear organization, metabolic pathways and gene expression characteristics. To investigate, whether the response to radiation damage is altered during cellular differentiation, we studied the formation and repair of DNA double-strand breaks in gamma-irradiated K562 erythroleukemia cells induced to differentiate by exposure to butyric acid. We applied an assay based on pulsed-field gel electrophoresis and Southern hybridization to measure break induction in several genomic restriction fragments. Pulsed-field gel electrophoresis of (14)C-labelled unrestricted DNA was used to study the rejoining of gamma-radiation-induced breaks in the whole genome. Total rejoining and joining of correct break ends in specific genomic regions was monitored by hybridization analysis of blots of unrestricted and restriction digested DNA with single-copy probes. The yields of gamma-ray-induced DNA double-strand breaks were found to decrease with differentiation by about 20%. Correct rejoining of radiation-induced breaks, as measured by the reconstitution of broken restriction fragments, was unaltered in differentiating cells compared to actively proliferating precursor cells. Total rejoining, however, appeared to be retarded in differentiating cells. The results suggest that in spite of the fundamental changes accompanying differentiation, the cellular damage response pathways are not essentially affected throughout erythroid differentiation.

DNAs with unusual properties revealed by field inversion gel electrophoresis of agarose-encapsulated DNA from mammalian cells

Distinct DNA fractions (fr-DNAs), moving separately from bulk DNA, were revealed by field inversion gel electrophoresis of DNA from intact cells lysed and deproteinized in agarose plugs. These fr-DNAs (approximately 2% of the total DNA) were ubiquitously present in nuclei of all mammalian cells studied, including human normal and tumor tissues, and showed a typical electrophoretic pattern (three bands with constant mobilities termed a-, b-, and c-DNA). Characteristic mobility shifts induced by gamma irradiation of a- and b-DNAs suggest their non-linear conformation. In fact, electron microscopy of a- and b-DNAs from Namalwa cells revealed rosette-like structures stabilized by a central protease-resistant knob. Comparative PCR analysis revealed qualitative differences between genomic fingerprints of a- and b-DNAs on the one hand and chromosomal DNA on the other. The results obtained suggest that fr-DNAs originate either from some specific chromatin regions due to non-random cleavages or from an autonomous intranuclear structure, not identified as yet.

Developmental changes in transcription factors associated with the nuclear matrix of chicken erythrocytes

The nuclear matrix has roles in organizing nuclear DNA and in controlling transcription. Transcription factors are associated with the nuclear matrix, with the spectra of transcription factors differing from one cell type to another. In this study we identified the transcription factors and enzymes functioning in the regulation of gene expression that were associated with nuclear matrix and nonmatrix nuclear fractions in erythrocytes isolated from chick embryos at different stages of development, anemic and normal adult birds. We found that the primitive erythroid nuclear matrix had the greatest histone deacetylase activity and highest levels of several transcription factors, including GATA-1, CACCC-binding proteins, and NF1. These transcription factors have key roles in erythroid-specific gene expression. The levels of these transcription factors were lower in the nonmatrix and matrix fractions isolated from definitive erythrocytes. For primitive and definitive erythrocytes, the level of CACCC-binding proteins in the nuclear matrix fraction was greater than that of Sp1. The relative levels of these transcription factors were reversed in the nonmatrix fraction. Casein kinase II was not found in erythroid nuclear matrices. The observed erythroid lineage specific alterations in erythroid nuclear matrix transcription factor composition and abundance may be involved in erythroid-specific gene expression.

Effect of differential gene expression on the chromatin structure of the DHFR gene domain in vivo

Photoactivated psoralen was used to probe region-specific chromatin structure in Chinese hamster ovary (CHO) cells. Specifically, the chromatin structure of six regions within the dihydrofolate reductase (DHFR) gene was probed with photoactivated psoralen in cells cultured in such ways as to differentially express the DHFR gene. Cells were irradiated with X-rays prior to the psoralen photocross-linking reaction in order to eliminate the influence of any DNA torsional tension on the psoralen binding and the sequence-specificity of psoralen binding was adjusted for. It was found that a region encompassing the promoter of the serum-regulated DHFR gene was about 50% more accessible to psoralen photocross-linking in serum-stimulated cells and about 90% more accessible in serum-starved cells than the other five regions of the DHFR gene analyzed and the genome overall. Treating serum-stimulated cells with the RNA polymerase II transcriptional inhibitor 5,6-dichloro-1-beta-D-ribofuranosyl-benzimidazole (DRB) or the topoisomerase I inhibitor camptothecin reversed the elevated accessibility of the DHFR promoter region. These results suggest that the accessible chromatin structure of the DHFR promoter is not dependent on serum-stimulated poising of the gene for transcription, but may reflect the ability of the RNA polymerase to clear the promoter.

The nuclear matrix and the regulation of chromatin organization and function

Nuclear DNA is organized into loop domains, with the base of the loop being bound to the nuclear matrix. Loops with transcriptionally active and/or potentially active genes have a DNase I-sensitive chromatin structure, while repressed chromatin loops have a condensed configuration that is essentially invisible to the transcription machinery. Core histone acetylation and torsional stress appear to be responsible for the generation and/or maintenance of the open potentially active chromatin loops. The transcriptionally active region of the loop makes several dynamic attachments with the nuclear matrix and is associated with core histones that are dynamically acetylated. Histone acetyltransferase and deacetylase, which catalyze this rapid acetylation and deacetylation, are bound to the nuclear matrix. Several transcription factors are components of the nuclear matrix. Histone acetyltransferase, deacetylase, and transcription factors may contribute to the dynamic attachment of the active chromatin domains with the nuclear matrix at sites of ongoing transcription.

Role of the liver-enriched transcription factor HNF-1 alpha in expression of the CYP2E1 gene

The role of the trans-acting factor HNF-1 alpha in activating CYP2E1 gene expression was confirmed by transient co-transfection of an HNF-1 alpha expression plasmid and the CYP2E1 promoter fused to the chloramphenicol acetyl transferase (CAT) reporter gene. Only HNF-1 alpha, and not HNF-1 beta, HNF-4, C/EBP alpha, C/EBP beta, or DBP, was able to activate the CYP2E1 promoter. The extent of activation was proportional to the number of copies of the HNF-1 binding sequence upstream of the promoter. Removal or mutation of the HNF-1 binding sequence led to inactivation of the promoter in response to HNF-1 alpha. Gel-shift Western blot analysis using a synthetic HNF-1 binding sequence derived from CYP2E1 and rat liver nuclear extract revealed that the protein-DNA complex obtained with adult rat liver nuclear extract consisted of both HNF-1 alpha and HNF-1 beta proteins. The shifted bands produced by nuclear extracts from adult, where the endogenous CYP2E1 gene is active, and fetal rat liver, where the gene is inactive, were found to migrate differently, suggesting that the population of factors, possibly including different ratios of HNF-1 alpha and HNF-1 beta proteins, may change during development. However, the co-transfection study did not show cooperativity between the two factors. Elements upstream of the HNF-1 binding site were found to affect the activity of the promoter negatively in the transfection assay. DNase I hypersensitive site mapping revealed a hypersensitive site in this inhibiting element in the adult rat liver sample but not in liver from newborn animals.

DNA stainability with base-specific fluorochromes: dependence on the DNA topology in situ

The influence of DNA topology on stainability with the externally binding fluorochromes Hoechst 33258 (HO) and mithramycin (MI) was investigated in HeLa nuclei in comparison with the intercalating dye propidium iodide (PI). Changes in DNA topology were induced with a mild DNAse I treatment. Stainability properties of untreated and nuclease-treated nuclei were compared with those of the supercoiled-circular and the relaxed-linear forms of the plasmid pBR322. DNAse-treated nuclei stained with HO showed a higher fluorescence intensity than control samples, independently of the dye concentration, in contrast with the findings obtained with PI. Similar behaviour was observed with the relaxed-linear form of pBR322, compared with the supercoiled-circular molecule. With MI, the stainability of HeLa nuclei did not depend on the DNA topology, whereas the stainability of the plasmid was similar to that of HO. In order to assess whether this discrepancy depended on differences in the availability of DNAse-sensitive sites to the fluorochromes, fluorescence resonance energy transfer (FRET) studies were performed in nuclei stained with HO+PI, or with HO+MI dye pairs. After DNAse I digestion, the relative FRET efficiency between donor (HO) and acceptor molecules (PI or MI) was reduced significantly only when MI was the acceptor. This result may be due to greater stainability of DNAse-sensitive sites with HO than with MI. These findings indicate that DNA stainability with base-specific fluorochromes may be affected by the topology of chromatin regions.

Photosensitized DNA damage in human cells is localized in chromatin sensitive to DNAse I digestion

Human NHIK 3025 cells were sensitized to light by incubation with Photofrin II (PII) and aluminum phthalocyanine tetrasulphonate (AlPCS4). Light exposure of sensitized cells at 1 degree C induced DNA strand breaks causing unwinding of DNA in alkali. Subsequent to light exposure, cells were treated with deoxyribonuclease I (DNAse I) which produces DNA strand breaks in active chromatin. The combination of photo-damage, killing > 95% of the cells, and treatment with DNAse I caused the same degree of DNA unwinding as DNAse I treatment alone. This was not the case for X-irradiation. Thus, the photosensitized DNA strand breaks were selectively induced in or close to DNAse I-sensitive sites.

Nucleosome spacing is compressed in active chromatin domains of chick erythroid cells

We have cleaved the chromatin of embryonic and adult chicken erythroid cells using a novel nuclease that is capable of resolving clearly the nucleosomes of active chromatin. We found that in active chromatin, nucleosomes are spaced up to 40 base pairs closer together than in inactive chromatin. This was true for both "housekeeping" and "luxury" genes and was observed whether the digestion was carried out on isolated nuclei in vitro or by activating the endogenous nuclease in vivo. The close spacing extended several kilobases into flanking chromatin, indicating that this is a domain property of active chromatin, not just a characteristic of regions disrupted by transcription. A simple interpretation of our results is that the nucleosomes of active chromatin are mobile in vivo and, not being constrained by linker histones, freely move closer together.

The DNA intercalator, ethidium bromide, alters the pattern of DNAse I hypersensitive sites of the beta A-globin gene in chicken erythrocytes

We have analysed the effects of a DNA intercalator, ethidium bromide (EB), on chromatin structure in nuclei from both chicken mature erythrocytes (RBC) and reticulocytes (Ret). A differential release of nuclear proteins was obtained from both types of nuclei exposed to EB. Among these proteins, a species of 45 kDa is the major component. Furthermore, in the 10 mM EB-treated nuclei, the pattern of DNAse I hypersensitive sites (DHS) around the chicken beta A-globin gene were significantly altered, i.e., the original set was replaced by a new set of DHS. We have discussed the implications of our observations, in the light of current concepts of functional aspects of the conformational heterogeneity of DNA in both protein-DNA interactions and chromatin structure, as well as the effects of DNA intercalators on DNA conformation.

Transcription-induced nucleosome 'splitting': an underlying structure for DNase I sensitive chromatin

Utilizing yeast strains containing promoter mutations, we demonstrate that transcription of the HSP82 gene causes nucleosomes toward the 3'-end to become DNase I sensitive and 'split' into structures that exhibit a 'half-nucleosomal' cleavage periodicity. Splitting occurs even when only a few RNA polymerase II molecules are engaged in basal level transcription or during the first round of induced transcription. The split nucleosomal structure survives nuclear isolation suggesting that it may be stabilized by post-translational modifications or non-histone proteins, and may require DNA replication for reversal to a whole nucleosomal structure. Split nucleosomes represent a structure for DNase I sensitive chromatin and are probably of common occurrence but difficult to detect experimentally. We suggest that transient positive supercoils downstream of traversing RNA polymerase lead to nucleosome splitting.

Nuclease-induced DNA structural changes assessed by flow cytometry with the intercalating dye propidium iodide

A flow cytometric analysis of DNA structural changes induced by cleavage with nucleases was performed on isolated HeLa nuclei by assessing changes in stainability with the DNA-specific fluorochrome propidium iodide (PI). After mild digestion with DNAse I, micrococcal nuclease, or with the single-strand-specific S1 and Neurospora crassa nucleases, fluorescence intensity of nuclei stained with PI increased by about 15-30% above the value of undigested control samples. No significant modifications were observed with the restriction enzymes Eco RI, Alu I, and Not I. The DNAse I-induced increase in fluorescence intensity was also observed with the non-intercalating dye Hoechst 33258, but not with mithramycin. Nuclease-induced fluorescence intensity changes as determined with PI were found to be dependent on the dye concentration. A constant increase (about 20%) was measured at dye/DNA-P ratios greater than 0.11. Below this value (2 micrograms/ml PI), the fluorescence intensity of digested samples was 15-30% lower than that of undigested controls. This behaviour towards intercalating dyes is similar to that of the relaxed (nicked) vs. the supercoiled (intact) form of circular DNA. These results suggest that conformation- but not sequence-specific nucleases induce a relaxation of DNA supercoils.

Topological properties of bovine papillomavirus type 1 (BPV-1) DNA in episomal nucleoprotein complexes: a model system for chromatin organization in higher eukaryotes

Sedimentation analysis of isolated episomal bovine papillomavirus type 1 (BPV-1) nucleoprotein complexes in sucrose gradients and subsequent separation of the purified DNA in chloroquine gels revealed different classes of molecules, varying in their degree of superhelicity. Since torsionally stressed DNA favors the adoption of secondary structures, we employed the single-strand-specific S1 nuclease to detect such structural alterations in both naked DNA and native chromatin. Direct examination of nuclease digestion products in chloroquine gels showed that neither the naked DNA nor the BPV-1 nucleoprotein complexes in isolated nuclei were cleaved randomly by the enzyme. Instead, there was a strict dependence on nuclease susceptibility and the degree of supercoiling, strongly suggesting that the structural features detected by S1 nuclease are due to the occurrence of torsionally stressed viral chromatin. Mapping analysis using the indirect end-labeling method demonstrated an S1-nuclease cleavage site adjacent to 20 homopurine residues known to be hypersensitive to S1 attack. Furthermore, direct methylation experiments with viral chromatin in isolated nuclei indicated that only circular, covalently closed nucleoprotein complexes served as substrate, whereas linearized BPV-1 chromatin was not susceptible to exogenously added Hhal methylase. This observation raises the possibility that the modulation of topology in nucleosomally organized DNA might also play a role in eukaryotic DNA methylation.

Inhibition of MEL cells' capacity to undergo erythroid differentiation by chemicals added during induction

Erythroid differentiation of murine erythroleukemia (MEL) cells, as induced by dimethyl sulfoxide, can be suppressed by chemicals at very low concentrations, not affecting cell viability and proliferation, if present in the culture medium between 18 and 24 h after addition of the inducer. The effect is apparent on the progeny of the treated cells and is determined, between day 3 and 5 following DMSO induction, as percent value of cells expressing the erythroid phenotype. Cultures showing decreased values are no longer terminal and a large number of clones, incapable of expressing the erythroid phenotype, can be isolated from them. In contrast, induced cultures are terminal if the added chemicals do not decrease the expression of the erythroid phenotype. Incorporation of thymidine into induced cultures reveals that maximal sensitivity of MEL cells to chemicals coincides with DNA duplication. In all affected cells, the inhibition to undergo erythroid differentiation is transmitted from one cell generation to the next.

Characterization of a topoisomerase-like activity at specific hypersensitive sites in the Drosophila histone gene cluster

It is well known that treatment of DNA-topoisomerase complexes with SDS induces cleavage of the DNA by trapping a reactive intermediate in which the topoisomerase is covalently linked to the terminal phosphates of the cut DNA. I have used this technique to examine potential topoisomerase binding sites in the histone gene chromatin of Drosophila Kc cells. Treatment of Kc nuclei with SDS induces Mg++-dependent DNA cleavage near the borders of two nuclease-hypersensitive sites located 5' and 3' of histone H4. It is likely that the SDS-induced cleavage at these hypersensitive sites is due to a topoisomerase because protein becomes tightly bound to the ends of the cleaved DNA fragments. Preliminary experiments suggest that a type II topoisomerase may be responsible for the cleavage.

5' structural motifs and Xenopus beta globin gene activation

We have analysed the structure of the Xenopus beta globin gene 5' flanking region in erythroid and non-erythroid chromatin, in supercoiled plasmids and in minichromosomes assembled in HeLa cell transfections. We have identified two erythroid chromatin-specific, nuclease-hypersensitive sites (HSs), one centred on the cap site, the other located 1000 base-pairs further upstream. An (AT)n tract is located 200 base-pairs upstream from each of these sites. In supercoiled plasmids, the (AT)n tracts, and not the chromatin HSs, are preferentially cleaved by single strand and double strand-specific nucleases. Using restriction enzymes, we have looked at the structure of the cap site HS in minichromosomes assembled in HeLa cell transfections. We find that the structure is indistinguishable from that found in erythroid chromatin, thus reinforcing our previous suggestion, based only on DNase I studies, that the formation of this HS is not dependent on erythroid-specific factors. In view of this close structural mimicry of the situation in vivo, we have used the HeLa cell model system to study the sequences required for cap site HS formation. We find that deletion of the (AT)n tract immediately upstream influenced neither the formation of the HS nor transcription of the globin gene. Indeed, these features remained unaffected by further deletion of upstream sequences, including 50 base-pairs of the HS itself. In this construct, the dimensions of the HS remained the same as in the undeleted construct, with the plasmid sequences that replaced the deleted Xenopus sequences becoming hypersensitive. Thus, HS formation is directed by sequences downstream from --116 acting over a distance of at least 50 base-pairs.