Biophysics Artificial nucleosome positioning sequences ( chromatin / histone - DNA binding / DNA bending )

We have used the emerging rules for the sequence dependence of DNA bendability to design and test a series of DNA molecules that incorporate strongly into nucleosomes. Competitive reconstitution experiments showed the superiority in histone octamer binding of DNA molecules in which segments consisting exclusively ofA and T or G and C, separated by 2 base pairs (bp), are repeated with a 10-bp period. These repeated (A/T)3NN(G/C)3NN motifs are superior in nucleosome formation to natural positioning sequences and to other repeated motifs such as AANNNTTNNN and GGNNNCCNNN. Studies of different lengths of repetitive anisotropically flexible DNA showed that a segment of "40 bp embedded in a 160-bp fragment is sufficient to generate nucleosome binding equivalent to that of natural nucleosome positioning sequences from SS RNA genes. Bending requirements along the surface of the nucleosome seem to be quite constant, with no large jumps in binding free energy attributable to protein-induced kinks. The most favorable sequences incorporate into nucleosomes more strongly by 100-fold than bulk nucleosomal DNA, but differential bending free energies are small when normalized to the number of bends: a free energy difference of only about 100 cal/mol per bend (1 cal = 4.184 J) distinguishes the best bending sequences and bulk DNA. We infer that the distortion energy of DNA bending in the nucleosome is only weakly dependent on DNA sequence. Since nucleosome formation can prevent the binding of regulatory proteins (1), the mechanism by which core histone binding sites are determined is of central importance to understanding gene regulation. Numerous authors have reported nucleosomes in regular arrays (phased nucleosomes) in the genome (2-5), and several contributing factors have emerged. One proposed cause is a boundary effect (6) in which a tightly bound nonhistone protein or histonenonhistone protein complex (7) forms an immovable anchor around which nucleosomes must organize. A second proposed model suggests that higher-order folding of chromatin modulates the placement of nucleosomes (8). Finally, there may be mechanisms for ordering nucleosomes in arrays based on DNA sequence-dependent interactions within each nucleosome. These could take the form of low-level, periodic signals that facilitate DNA bending, and thus the curving of the nucleic acid around the histone octamer (9, 10), or more conventional recognition of a short DNA region through specific contacts to the individual histones. The above effects are not mutually exclusive, and regions of phased nucleosomes with both sequence-dependent binding and boundary effects have been reported (11, 12). The sequence periodicities reported in the eukaryotic genome fall into two categories. In early work, Trifonov and Sussman (9) deduced periodicities in the eukaryotic genome that included repeats of the form AANNNTTNNN. The direction of bending of this motif has not been determined, but base-pair tilt (to yield bending toward the -backbone rather than compression of a groove) has been proposed (9). This directionality results from symmetry considerations if AA and TT segments, separated by half a helical turn, are both to bend preferentially toward the bound core histones. The second type of repeat to be reported, but one that has been characterized more thoroughly, consists of alternating A/Tand G/C-rich regions with a total period of 10 base pair (bp) (ref. 10; see also Fig. 1). These periodicities seem to exist over the entire '"145 bp of the DNA protected in mononucleosomes, with some irregularities near the dyad (13). Correlation of minor-groove accessibility with sequence shows that these fragments preferentially orient in nucleosomes such that G/C regions are found where the major groove is compressed (faces in as the DNA axis curves around the protein); A/T sequences are favored at sites of minor-groove compression (10). These preferred orientations also reflect flexibility rules determined for DNA bending by the catabolite activator protein (CAP) of the Escherichia coli lac promoter (14), suggesting that they are not nucleosomespecific but result from sequence-dependent anisotropies in the bendability of DNA. Along with these periodic sequence effects, there are several examples of specific nucleosome positioning sequences (15-17). The corresponding DNA fragments bind histones to form nucleosomes at defined positions in vitro and potentially could act as stable nucleosome boundaries in vivo. In at least one of these sequences the nucleosome occupies a position advantageous to the binding ofa regulatory protein (17). Mutagenesis studies on some fragments of this type revealed an essential positioning region that is reasonably short (40-50 bp), but no specific signal has been identified (16, 18). Inspection of these molecules shows some alternating A/Tand G/C-rich regions at majorand minorgroove compression sites, but nonperiodic signals have also been proposed (19). Natural nucleosomal DNA contains only modest sequence periodicity; our experiments probed the extent to which more powerful binding sequences can be designed. Additionally, we attempted to determine which of the experimentally determined sequence periodicities is most advantageous to histone binding and whether or not signals of this kind might be responsible for natural nucleosome positioning sequences. The strong nucelosome binding sequences we describe should be useful for further in vitro and in vivo investigation of the role of histone-DNA binding affinity in the packaging and function of genomic DNA. EXPERIMENTAL PROCEDURES Cloning Vectors. Two vectors, based on pGEM-2 (Promega), were used throughout this study. The majority of the fragments were made by cloning multimers of oligonucleotides into pGEM2-Ava. This vector was produced from pGEM-2 in two steps. First, the unique, symmetric Ava I site in the original vector was destroyed by cutting with Ava I, filling in the resulting overhangs with the Klenow fragment of 7418 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Proc. Natl. Acad. Sci. USA 86 (1989) 7419 DNA polymerase I, and reclosing. A new, asymmetric Ava I site (20) was then introduced into this intermediate plasmid by cloning the sequence GATCGCTCGGGTG into the BamHI site. Due to the lack of restriction sites on the SP6 promoter side of the Ava I site of the pGEM2-Ava polylinker, a second vector was used for very short inserts. This chimeric vector was constructed by cloning the 1425-bp EcoRI-Ava I fragment of pBR322 into pGEM2-Ava. This new vector has a Sty I site =80 bp from the asymmetric Ava I site. Oligonucleotide Inserts. Oligonucleotides were synthesized on an Applied Biosystems DNA synthesizer and purified by electrophoresis in 12-15% polyacrylamide gels containing 50% (wt/vol) urea in TBE (45 mM Tris/45 mM boric acid/ 1mM EDTA, pH 8.3). The purified single strands were phosphorylated with polynucleotide kinase and then annealed by slow cooling from 90'C to 40C. Approximately 10 gg of this double-stranded oligonucleotide was polymerized for 15 min at room temperature, and the ligation mixture was fractionated in a 6% polyacrylamide gel run in TBE. The desired multimer band was visualized by staining with ethidium bromide, excised, and soaked overnight at 45-55°C. The samples were spun twice to remove gel slices and the DNA was precipitated with ethanol. Purified multimers were cloned into the Ava I site of the appropriate vector. Clones were screened by the alkaline lysis method and verified by dideoxy sequencing (technical manual, Promega). Reconstitution, Binding, and Footprinting. The procedure used for reconstituting nucleosomes was similar to histone exchange methods used previously (16). The major differences were the addition of bulk competitor DNA and much lower levels of total protein. In brief, polynucleosomes stripped of histones H1 and H5 were produced by the standard methods (21) from chicken blood purchased from Pel-Freeze Biologicals. The material was checked for histone content and integrity, divided into aliquots, and frozen at -20°C for future use. In binding studies, -5 ,ug of this stripped chromatin was mixed with r20 ,g of bulk DNA (isolated from chicken erythrocytes) and 0.1 ,ug of a labeled DNA fragment in 20 mM Tris/1 M NaCl containing 100 ,g of albumin per ml and 0.1% Nonidet P-40 (Shell Chemicals, London). The mixture (70 ,l) was incubated at room temperature for 30 min before the salt concentration was lowered to 0.1 M with three additions of 210 ,ul of 20 mM Tris buffer (20 min apart, room temperature). The percentage of a DNA fragment that had been incorporated into nucleosomes was assayed by separating free DNA from complexed DNA in 5% polyacrylamide gels (75:1 acrylamide/N,N'-methylenebisacrylamide weight ratio) run at 20°C in TBE. The gels were dried and autoradiographed for 1-2 hr. The resulting films were used as templates to locate the radioactive bands, which were excised and quantitated in a Packard 1500 scintillation counter. The free energy reported for a given sequence (Seq) was calculated from the equation E(Seq) = RTln[a(TG)] RTln[a(Seq)], where a(TG) is the ratio of counts in the nucleosomal complex band for the pentamer of the TG reference oligonucleotide (see Fig. 1 and Table 1) to counts in the free TG pentamer band, and a(Seq) is the analogous ratio for Seq reconstituted under identical conditions. Each fragment was reconstituted in two separate experiments and the results were averaged. The two numbers generally agreed to within =100 cal. For hydroxyl-radical footprinting studies, the nucleosomes were produced by histone exchange without added competitor DNA. The salt was lowered from 1 M to -25 mM by slow dialysis. This resulted in >90% incorporation of the labeled fragments into nucleosomes. The hydroxyl-radical footprinting procedure was as described elsewhere (22). RESULTS Competitive Reconstitution Approximates an Equilibrium Distribution. Our experimental approach (Fig. 1) measures the ability of a labeled DNA fragment to compete with bulk DNA for a limited number of histone octamers during the nucleosome reconstitution procedure. The results are given as apparent free energy differences in the binding of a variety of sequences relative to binding of a standard reference sequence. It should be recognized that the partitioning of DNA molecules during binding in a salt gradient could, in principle, reflect either kinetic or equilibrium factors, or some combination of the two. Although most of the major points of this study depend on the ranking of sequences and would not change if the reconstitution efficiencies partly reflected a nonspecific, kinetic process, control experiments implied that the free energy differences closely approximate a true equilibrium and are probably relevant to nucleosome formation at physiological salt levels. Test reconstitutions reached the same final (labeled nucleosome/labeled free DNA) ratios regardless of whether the radioactive fragment was initially added as free DNA or in nucleosomal complexes. This convergence to common percentages of incorporation starting from either very high or very low fractions of labeled nucleosomes suggests a true equilibrium. Additionally, neither extending the incubation time in the high-salt buffer nor changing the rate at which low-salt buffer was added affected the fraction of nucleosomes produced, to