Search

The dinucleosome is an informative unit for analysis of the higher-order chromatin structure. DNA fragments forming stable dinucleosomes were screened from a dinucleosome DNA library after the reconstitution of nucleosomes in vitro and digestion with micrococcal nuclease. Reconstituted dinucleosomes showed a diversity of sensitivity to micrococcal nuclease, suggesting that the biochemical stability of a dinucleosome depends, in part, on the DNA fragments. The DNA fragments after the screening were classified into three groups represented by clones bf10, af14 and af32 according to the sensitivity to micrococcal nuclease. Mapping of the nucleosome boundaries by Southern blotting of the DNA after restriction digestion and by primer extension analysis showed that each nucleosome position of clone af32 was fixed. Analysis of reconstituted dinucleosomes using mutant DNA fragments of clone af32 revealed a unique property characteristic of a key nucleosome, given that the replacement of a DNA fragment corresponding to the right nucleosome position resulted in marked sensitivity to micrococcal nuclease, whereas the replacement of the other nucleosome fragment had almost no effect on sensitivity as compared to the original af32 construct. The mutant construct in which the right nucleosome was removed showed multiple nucleosome phases, suggesting that the right nucleosome stabilized first each mononucleosome and then the dinucleosome. An oligonucleotide bending assay revealed that the DNA fragment in the right nucleosome included curved DNA, suggesting that the positioning activity of the nucleosome was attributed to its DNA structure. These results suggest that information for forming stable dinucleosome is embedded in the genomic DNA and that a further characterization of the key nucleosome is useful for understanding the building up of the chromatin structure.

We previously reported five DNA bend sites (ERB-4 to -1, and ERB+1) in the promoter region of the human estrogen receptor alpha (ERalpha) gene [FEBS Lett. 444 (1999) 117]. One of these sites, ERB-2, was accompanied by two half motifs of the estrogen response element (ERE) and several short poly(dA)(.)poly(dT) tracts including an A(4) tract located next to a half ERE motif. This A(4) tract and the 20 bp immediate flanking sequence containing a half ERE motif (T3B) exhibited DNA curvature. Transcription assays using luciferase as a reporter gene indicated that T3B sequence conferred positive estrogen responsiveness. Mutations introduced in this sequence indicated that both bendability and estrogen responsiveness were synergistically associated with the A(4) tract located next to the half ERE motif. This motif and a mutant sequence, GGTTA, had affinity for ERalpha protein, which seems to account for ERalpha protein binding to the region without an ERE motif. These findings suggest that some DNA curvature acts as a transcriptional modulator by modifying the state of ligand effects.

DNA bending has been observed in conjunction with transcription, replication, and recombination. Furthermore, nucleosomes in eukaryotic cells are positioned through DNA bending, suggesting an active role for DNA bending in the chromosome organization. We reported previously that DNA bend sites appear every 680 bp in the human epsilon- and beta-globin gene regions. Here we showed that these sites are present at an interval of roughly 700 bp in the G gamma-A gamma-psi beta-globin gene region and that they divide the region into units. They were conserved in the promoter regions of nearly all beta-like globin genes and between human beta- and mouse beta maj-globin genes, although the periodicity of the sites was locally disturbed at the junctions of the duplicated G gamma- and A gamma-globin genes and in their second introns. This suggested that the periodicity is ranked lower in the hierarchy of genomic DNA organization than genome rearrangement and gene expression. A close inspection of one of the sites in the A gamma-globin gene region indicated that a 20-bp sequence containing periodic short (dA)n tracts was partly responsible for the bending. This sequence was shown to phase nucleosomes in this region by preferential binding to the core histones.

A total of seven DNA bend sites were mapped in the 4.4-kilobase human beta-globin gene region by the circular permutation assay. The periodicity of these sites (except one) was about every 700 (average 685.5 +/- 267.7) base pairs. All of the sites contained the sequence feature of short poly(dA) tracts, which are typical of DNA bending. The relative positions of the sites to the cap site were identical to those in the epsilon-globin gene region, suggesting that the bend sites were conserved during molecular evolution of the two globin genes. To explain this periodicity and conservation of the sites within the evolutionary unstable noncoding regions, we focused upon the appearance of a potential bend core sequence, A2N8A2N8A2 (A/A/A), and its complement, T2N8T2N8T2 (T/T/T). These sequences appeared in or very close to most of the bend sites of the globin gene regions, whereas other A+T-rich sequences or candidates for DNA bending did not. The distances between any two of the core sequences in the entire beta-globin locus showed a strong bias to a length of about 700 base pairs and its multiples, suggesting that the periodicity exists throughout the locus. The data presented here strengthen the idea of sequence-directed nucleosome phasing.

Analysis by the circular permutation assay of the human epsilon-globin gene region revealed that the DNA bend sites were located every 682.5 +/- 132.0 base pairs on average, separating the region into domains. Among 10 major and 1 minor bend sites mapped in the region, the transcription initiation and termination sites of the epsilon-globin gene were located close to the bend sites, and the first and the second exons of the epsilon-globin gene were separated from the third exon by another site. The bend sites were also located anterior to the two Alu family sequences. Short poly(dA).poly(dT) tracts typical for DNA bending were not always present in the sites. Fine mapping of a bend site having no poly(dA).poly(dT) tracts with concatenated oligonucleotides and analysis by S1 nuclease nicking assay indicated that the unusual structure, a base slippage or a partial triplex DNA structure, formed by a polypurine.polypyrimidine sequence in the region is the basis of bending. The bend sites were mapped in the promoter region (within approximately 300 base pairs from the cap site) of the human beta-globin and in c-myc and erythropoietin receptor genes, as well as in the mouse beta maj-globin gene. The conservation and the periodicity of the bend sites in the noncoding region suggest the active role of the sites that is a signal for nucleosome phasing.

Dinucleosome formation is the first step in the organization of the higher order chromatin structure. With the ultimate aim of elucidating the dinucleosome structure, we constructed a library of human dinucleosome DNA. The library consists of PCR-amplifiable DNA fragments obtained by treatment of nuclei of erythroid K562 cells with micrococcal nuclease followed by extraction of DNA and adaptor ligation to the blunt-ended DNA fragments. The library was then cloned using a plasmid vector and the sequences of the clones were determined. The dominating clones containing the Alu elements were removed. A total of 1002 clones, which comprised a dinucleosome database, contained 84 and 918 clones from the clones before and after removing Alu elements, respectively. Approximately 70% of the clones were between 300 and 400 bp in size and they were distributed to various locations of all chromosomes except the Y chromosome. The clones containing A 2 N 8 A 2 N 8 A 2 or T 2 N 8 T 2 N 8 T 2 sequences were classified into three types, Type I (N shape), Type II (V shape) and Type III (M shape) according to DNA curvature plots. The locations of experimentally determined curved DNA segments matched well with the calculated ones though the clones of Types I and III showed additional curved DNA segments as revealed by the curvature plots. The distributions of complementary dinucleotides in the nucleosome DNA, at the ends of the dinucleosome DNA clones, allowed us to predict the positions of the nucleosome dyad axis, and estimate the size of the nucleosome core DNA, 125 nt. The distributions of AA and TT dinucleotides, as well as other RR and YY dinucleotides, showed a periodicity with an average period of 10.4 bases, close to the values observed before. Mapping of nucleosome positions in the dinucleosome database based on the observed periodicity revealed that the nucleosomes were separated by a linker of 7.5+∼10× n nt. This indicates that the nucleosome–nucleosome orientations are, typically, halfway between parallel and antiparallel. Also an important finding is that the distributions of AA/TT and other RR/YY dinucleotides, apparently, reflect both DNA curvature and DNA bendability, cooperatively contributing to the nucleosome formation.

Human and mouse globin genes were separated approximately 200 million years ago but still share homology and synergism in many aspects including DNA structure. We first mapped DNA bend sites in the mouse epsilon-globin gene and found that these sites were distributed in a regular manner except in the coding region and their overall average distance was 650.7 bp. The first bend site upstream of the cap site (MepsilonB-1, -334 to -147 bp) was found to contain A + T-rich sequences and features contributing to DNA curvature by computer analysis. Transcription assays using deletion constructs indicated strong promoter activity up to bp -215 in erythriod K562 cells. Therefore, the MepsilonB-1 site was located immediately upstream of the promoter region. A reporter gene assay using a series of constructs containing the promoter region revealed that the MepsilonB-1 site showed repressor activity, and on replacement of the DNA curvature with one from another source the activity was retained. A similar feature was found in the other conserved B-1 sites in the human, bovine, and rabbit beta-like globin genes, with the exception of an unconserved B-1 site in the chicken beta-globin gene. A common feature of these conserved B-1 sites was not the nucleotide sequences but the DNA curvature. Furthermore, a unique nucleosome phase at the MepsilonB-1 site was likely to be directed by DNA curvature. Based on these results, DNA curvature is one of the major features of these promoter regions which might influence transcription through nucleosome positioning.

We reported previously that DNA bend sites appear in the human beta-globin locus at an average distance of 680 bp. The relative locations of the sites were conserved among the five active beta-like globin genes and one pseudogene. Here, we mapped the sites in the beta-like globin genes from various species and examined their conservation. The locations of the bend sites in the bovine, rabbit and chicken beta-globin genes mapped here showed marked conservation in their locations relative to the cap site and showed similar locations to the previously mapped sites in the human beta- and mouse betamaj-globin genes. Further analysis of the first bend sites from the cap site (B-1 sites) indicated that they contained tracts of adenines and thymines longer than or equal to two bases. This sequence feature contributed mostly to the curvature profiles revealed by gel assays and/or by computer-based TRIF analysis. TRIF analysis indicated that most of the B-1 sites showed right-handed superhelical twists accompanied by left handed twists. This was confirmed by the effect of ethidium bromide on the superhelical twists in the assays.

We report here the locations of curved DNA in the human erythropoietin receptor gene. A total of 13 DNA bend sites were mapped by circular permutation assays, appearing at an average interval of 651.2+/-214.6 (S.D.) in the 8-kb region. The bend centers in these 13 bend sites were confirmed by oligonucleotide-based assays where most of these centers had bend angles higher than that shown by (AAACCGGGCC) x (A)20 and lower than that shown by (AAACCGGGCC)2 x (A)10. DNA curvature mapping by TRIF software, which is based on the distribution of dinucleotides, primarily AA and TT, provided a highly accurate prediction for the locations of the bend sites. They showed approximately 20 degrees to 40 degrees of bend angles demonstrated by the oligonucleotide assays and by computer analysis.

We determined DNA bend sites in the promoter region of the human estrogen receptor (ER) gene by the circular permutation assay. A total of five sites (ERB−4 to −1, and ERB+1) mapped in the 3 kb region showed an average distance of 688 bp. Most of the sites were accompanied by short poly(dA)·poly(dT) tracts including the potential bend core sequence A2N8A2N8A2 (A/A/A). Fine mapping of the ERB−2 site indicated that this A/A/A and the 20 bp immediate flanking sequence containing one half of the estrogen response element were the sites of DNA curvature. All of the experimentally mapped bend sites corresponded to the positions of DNA curvature as well as to nucleosomes predicted by computer analysis. In vitro nucleosome mapping at ERB−2 revealed that the bend center was located 10–30 bp from the experimental and predicted nucleosome dyad axes.

DNA bend sites appear periodically at average intervals of 680 bp, corresponding to a length of four nucleosomes, in the human epsilon-, beta- and Ggamma-Agamma-psibeta-globin gene regions. We found that the HS2 region flanked by two DNA bend sites accommodated five nucleosomes and they were regularly phased throughout the region with the exception of that located in the middle, which corresponded to the precise location of HS2 and included the binding site for NF-E2. There appeared to be several phases in this region in the reconstituted chromatin and in erythroid K562 cells where the globin genes are expressed, whereas only one phase was adopted in non-erythroid HeLa cells. Meanwhile, almost unique phases were adopted at the flanking bend sites in vitro as well as in vivo. Sequences of 30 bp containing the bend centers cloned into the vector alone showed identical nucleosomal phases to those observed with the in vitro and in vivo experiments and removing the bend sites caused disruption of the phases at the bend sites as well as those in their direct vicinity. Finally, the nucleosome in this HS2 region had an inhibitory effect on NF-E2 binding, although remodeling occurred with the nuclear extract from K562 cells in the presence of ATP. This suggests that HS2 is placed at a region of weak nucleosome phasing activity along with factor binding sites.

DNA bend sites appear every 680 bp on average in the human epsilon- and beta-globin gene regions. Although most of their molecular nature has not been unraveled, a potential bend core sequence A2N8A2N8A2 (A/A/A) and its complementary T2N8T2N8T2 (T/T/T) appeared preferentially either in or very close to most of the bend sites, whereas other combinations of A2 and T2 dinucleotides, A/T/T + A/A/T, T/T/A + T/A/A and A/T/A + T/A/T, did not. The distances between any two of the core sequences in the entire beta-globin locus showed a strong bias to a length of 701-800 bp and multiples thereof, suggesting that there is periodicity throughout the locus. This bias was not found for other combinations of A2 and T2. Again, this periodicity was identified in many eukaryotic genes, whereas the tendency was absent in mRNAs and prokaryotic as well as viral genomes.

Recent progresses in the study of histone modifications in gene networks and signaling revived the importance of chromatin for controlling gene functions. However, the difficulty of chromatin study lies first in the number of necleosomes in a cell: roughly 107 nucleosomes, each containing a different DNA sequence, are continuously tied as ”beads and strings”, interacting each other, between the neighboring ones as well as the distant ones. Chromatin organization and nucleosome stability is crucial in massive changes of chromatin enviroment such as in DNA replication and formation of chromosomes at cell division as well as in the local changes as observed in transcription of genes, changes of active/inactive chromatin regions and their boundary formation. One of the best ways to study this complex state is first to classify these nucleosomes based on some definitive criteria. Realization that all nucleosomes are not behaving identically is not new. However, explanation at the molecular level of such differences needs a comprehensive analysis of the genome structure by molecular biological experiments and by computational extractions which became available only in recent years. We provide here evidence for a conserved regulatory element for transcription of the β-like globin gene as a summary of our recent studies of a total of over 30 genes using experimental and/or computational approaches (Onishi & Kiyama. Nucl. Acids Res.29, 3448–3457, 2001; Onishi & Kiyama J. Biol. Chem.278, 8163–8171, 2003; Kato et al. J. Mol. Biol.332, 111–125, 2003; Wanapirak et al. J. Mol. Evol.56, 649–657, 2003; Onishi & Kiyama. Recent Res. Devel. Nucl. Acids in press). The element is characterized by the appearance of AA- or TT-dinucleotides in the A+T-rich region located 200 to 400 bp upstream of cap sites. G tracts 3 to 5 nucleotides long exist between the A+T-rich region and the conserved transcription factor binding sites (GATA-1 site and the CACCC, CCAAT, and ATA boxes). An average periodicity of AA- or TT- dinucleotides in the region from a total of eighteen β-family globin genes from four species was approximately 13 bp, suggesting DNA in these regions to show right-handed superhelicity. A proposed biological meaning of this element is to adjust the spatial positions for the first interaction of the transcription factors which should recognize specific DNA sequences in the presence of packed chromatin.