The structure of DNA in a nucleosome

Transcription

1 Proc. Nati. Acad. Sci. USA Vol. 87, pp , October 1990 iochemistry The structure of DNA in a nucleosome JEFFREY J. HAYES*, THOMAS D. TULLUSt, AND ALAN P. WOLFFE* *Laboratory of Molecular iology, National nstitute of Diabetes and Digestive and Kidney Diseases, National nstitutes of Health, uilding 6, Room 131, ethesda, MD 20892; and tdepartment of Chemistry, The Johns Hopkins University, altimore, MD Communicated by Martin Gellert, June 8, 1990 (received for review March 28, 1990) ASTRACT We describe the application of the hydroxyl radical footprinting technique to examine the histone-dna interactions of a nucleosome that includes part of the 5S ribosomal RNA gene ofxenopus borealis. We establish that two distinct regions of DNA with different helical periodicities exist within the nucleosome and demonstrate a change in the helical periodicity of this DNA upon nucleosome formation. n particular, we find that on average the helical periodicity of DNA in this nucleosome is ± 0.05 base pairs per turn. The same DNA, when bound to a calcium phosphate surface, has a periodicity of ± 0.05 base pairs per turn, similar to that of random sequence DNA. Modulations in minor groove width within the naked DNA detected by the hydroxyl radical are maintained and exaggerated in nucleosomal DNA. These features correlate with regions in the DNA previously suggested to be important for nucleosome positioning. The basic repeating unit of eukaryotic chromatin is the nucleosome (1, 2). Crystallization of nucleosome core particles reveals a histone core around which DNA is wrapped (3). Two central concepts have been proposed from a consideration of the constraints imposed on DNA by the histone core. First, DNA curvature may be a major determinant of nucleosome positioning (4). This is presumed to be a consequence of the fact that DNA is wrapped around the histone core in a shallow helical path with one complete turn comprising only 80 base pairs (bp). Second, the helical periodicity of nucleosome-wrapped DNA is thought to be different from that in free DNA. Such a difference in helical period would resolve the so-called linking-number paradox (5). This apparent discrepancy, that 1.75 turns of DNA around the nucleosome lead to only one DNA supercoil, has been explained by a change in helical periodicity from 10.5 bp per turn in solution to 10.0 bp per turn over the 146 bp of DNA believed to be associated with the histone core (6). oth of these concepts require alterations to the structure of DNA in a nucleosome compared to its state free in solution. These structural alterations are likely to affect the interaction of other DNAbinding proteins with nucleosomal DNA (7, 8). Whether or not a change in the helical periodicity of DNA occurs following incorporation into a nucleosome and whether this explains the linking-number paradox has led to much discussion (8-11). Measurements of the helical periodicity ofdna in the nucleosome with enzymatic probes such as DNase have shown that cleavage sites are spaced with an average periodicity of about 10.4 bp, not 10.0 bp as expected from theoretical predictions (5, 12). This discrepancy might be explained by steric hindrance of nucleases from attacking at all sites at an angle normal to the nucleosome surface (5). This explanation is supported by the sequence analysis of cloned nucleosomal DNA (13). n this work, we detail a more direct approach to investigate the interaction of DNA with the histone core, using a uniquely positioned nucleosome that includes DNA from a 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 solely to indicate this fact. Xenopus borealis 5S RNA gene (14) and the hydroxyl radical as a probe of DNA structure (15). A major advantage of this method is that hydroxyl radical cleavage of DNA avoids the problems associated with the large size and sequence specificity of enzymatic probes of DNA structure (5, 16). Thus it is now possible to analyze the structure of any unique sequence of DNA within a nucleosome at single-nucleotide resolution. The hydroxyl radical footprinting technique allows us to measure the helical periodicity of DNA in a nucleosome and reveals a clear change in the helical period of 5S DNA upon nucleosome formation. We also confirm and extend the concepts of DNA bending as a major determinant of nucleosome positioning. MATERALS AND METHODS DNA Fragments. Radiolabeled DNA fragments contained the X. borealis somatic 5S RNA gene inserted in either orientation within the vector psp64 at a amh site (17). These were used to reconstitute nucleosomes and for hydroxyl radical footprinting. A 583-bp Hha -EcoR fragment from plasmid pxp-10 (17) was radiolabeled at the EcoR site. The axis of dyad symmetry of the nucleosome on the 5S RNA gene (14) is -72 bp from the radiolabel. A 293-bp Hind- EcoR fragment from plasmid pxp-14 (17) was radiolabeled at the Hind site. n this case the axis of dyad symmetry of the nucleosome on the 5S RNA gene is -84 bp from the radiolabel. Only a single nucleosome will bind to this particular DNA fragment (18). A Hpa -Rsa fragment from pxbs-1 (19), which extends from position -102 to +75 with respect to the start of transcription of the 5S gene, was used in the calcium phosphate experiments to investigate the helical periodicity of the upstream half (-80 to -3) of the nucleosome binding site. Likewise, the Hind-amH fragment from pxbs 201 (20) (-50 to +196), labeled at the Hind site, was used to investigate the periodicity of the downstream half (-3 to +70) of the nucleosome binding site (see Fig. 2). Nucleosome Reconstitution and Footprinting. Chicken erythrocyte histones were reconstituted onto the radiolabeled DNA fragments to form nucleosomes by salt/urea dialysis (21-23). The efficiency of reconstitution was monitored by electrophoresis (4, 24). We found that >95% of the labeled DNA needed to be reconstituted into nucleosomes to allow the appearance of a hydroxyl radical footprint (data not shown). Cleavage of DNA in nucleosomes by DNase or the hydroxyl radical was accomplished as described for other protein-dna complexes (16). Quantitative Analysis. Autoradiographs were scanned with a Joyce-Loebl Chromoscan 3 densitometer fitted with an aperture wide enough to allow measurement of the optical density across a whole lane. The area of each band was determined by integration using software included with the densitometer. ntegrals of bands from the control (free DNA) samples were subtracted from the integrals of corresponding bands in the nucleosome samples. The resulting values, which represent the amount of cleavage at each nucleotide in the nucleosome, were then smoothed by performing a threebond running average throughout the entire data set. Data 7405

2 7406 iochemistry: Hayes et al. sets were fitted with a sine function by using the program PASSAGE on a Macintosh computer. Data sets were also analyzed by Fourier transformation. The errors quoted in the text are ±2cr. RESULTS AND DSCUSSON The Helical Periodicity of DNA in the SS Nucleosome. The small size and lack of sequence specificity of the hydroxyl radical compared to enzymatic probes prompted us to attempt to analyze the structure of DNA in a nucleosome by using this reagent. Nucleosomes made up of histone cores assembled on 5S DNA were "footprinted" with both the hydroxyl radical and DNase. A clear modulation of cleavage is seen with both reagents (Fig. 1). The DNase footprint is identical to those previously reported for this gene (14). Regions of maximum hydroxyl radical cleavage coincide approximately with the DNase cleavage sites in the nucleosome (Fig. 1). Experiments in which the opposite strand was radiolabeled revealed a stagger in the positions of maximal cutting of 1-3 bases, characteristic of the minor groovecentered cleavage chemistry of hydroxyl radical (data not shown; ref. 16). Densitometric analysis (Fig. 2) of these A A ~!Eo j! m.*. ;.^ inn.:'m. 0 er4- sw -.. N3 N /7 *g Ei l '* *45 33: DYADif _i..: * r 4,.ik -4-:. E)YAL-) FG. 1. Hydroxyl radical footprints of a nucleosome incorporating the X. borealis 5S RNA gene. (A) For this experiment, the 583-bp fragment of pxp-10, radiolabeled on the noncoding strand of the 5S RNA gene, was used. Cleavage patterns are shown for naked 5S DNA generated by DNase (lanes 1 and 2) and by the hydroxyl radical (lane 6) as well as for DNA in the nucleosome generated by DNase (lanes 3 and 4) and by the hydroxyl radical (lane 7). Lane 5 contains G+A markers generated by chemical cleavage of 5S DNA. DNA fragments were separated by electrophoresis on a 6% polyacrylamide gel containing 7 M urea. Cleavage patterns resulting from DNase digestion for 30 sec (lanes 1 and 3) and 2 min (lanes 2 and 4) are shown. () The same set of samples was electrophoresed for a shorter time to highlight the smaller DNA fragments. The position of the axis of dyad symmetry of the nucleosome on the 5S RNA gene is indicated, as is the internal control region of the 5S RNA gene (stippled bar covering positions +45 to +95). The footprint of the nucleosome that is positioned over the beginning of the 5S gene is indicated by the black bar next to the autoradiograph. The approximate positions of two other nucleosomes detected by the hydroxyl radical, which are located downstream of the 5S nucleosome, are indicated by the black bars labeled N2 and N3. Proc. Natl. Acad. Sci. USA 87 (1990) autoradiographs shows that the hydroxyl radical cleavage pattern is modulated with a period of bases over extended regions on either side of the center of dyad symmetry of the nucleosome (14), which is marked by the large 'A c c) c.7'1 t ".7 f~~~~~~~~~~~~~~~~~~~1 ase Position f.t fl... A.11,.. ) 01.!..4 j. 0 CR., f. f Free DNA DNA on Calcium Phosphate DNA in Nucleosome FG. 2. (A) Densitometer scans of the hydroxyl radical footprint of the 5S nucleosome. Cleavage patterns for the noncoding strand of 5S DNA are shown. The upper tracing is the hydroxyl radical footprint of the nucleosome, and the lower tracing is the pattern of hydroxyl radical cleavage of this same DNA when free in solution. The small arrows mark the positions of maximum hydroxyl radical cleavage of nucleosomal DNA, which were determined after subtraction of the trace of naked DNA. Nucleotide positions are numbered relative to the start of the 5S RNA gene. Note that there is no zero position in this figure. The corresponding positions on naked DNA are shown for reference. The large vertical arrows indicate the position of the axis ofdyad symmetry of this nucleosome as previously determined (14). The horizontal black bar indicates the region that appears to be internucleosomal. () Plots of the hydroxyl radical cleavage frequency at each nucleotide of the noncoding strand of the X. borealis 5S RNA gene free in solution, bound to calcium phosphate, and assembled in a nucleosome. Sites are numbered relative to the start of transcription of the 5S gene (+1). Positions -n (upstream) actually correspond to positions -(n - 1) in the usual numbering scheme for upstream sequences, due to the necessity of the plotting program to include a point at 0. Note that the "DNA on Calcium Phosphate" data are actually comprised of two independent data sets, which were joined at position -3 as indicated by a horizontal dash. Each data set covers one-half of the nucleosome binding site and is not expected to have any relation to the other. The horizontal bar above the curve indicates the approximate position of the 5S gene internal control region (CR). The vertical lines are included to highlight the change in helical periodicity between DNA bound to calcium phosphate and in the nucleosome. They also show how the periodic features in the cleavage pattern of naked DNA correlate with the other two patterns. 120 i

3 iochemistry: Hayes et al. vertical arrow in Fig. 2A. t is also clear from this scan that the 10.0-base periodicity does not extend across the center of dyad symmetry, because the modulation patterns to either side of the center are not in phase with each other (see below). We assume that the modulation of the cleavage pattern gives directly the helical repeat of the DNA (16) in the local frame of reference [i.e., with respect to the surface of the histone core (6, 25)]. Cleavage by the hydroxyl radical can be attenuated by alterations in minor groove width as a consequence of DNA bending (26) or by the presence of protein bound to the helix (16). The hydroxyl radical cleavage data (Fig. 2) was quantitated (see Materials and Methods). To evaluate the periodicity for the entire core DNA within the 5S nucleosome, it is most convenient to treat separately the three turns of DNA that span the center of dyad symmetry (located at position -3), since this region has a helical periodicity that clearly differs from that in the rest of the nucleosome. Fig. 2A shows that regions to either side of the dyad have a periodicity of 10.0 bp that are out of phase by -2 bp if extended across the center of dyad symmetry. Cleavage data from each of the flanking regions, which cover positions +14 to +72 and positions -17 to -75 in the 5S gene, were analyzed independently by Fourier analysis of the smoothed data (Fig. 3A) or by the fitting of a sine function to the data set (not shown). oth analyses yield identical results. We find that both of the flanking regions have a periodicity of ± 0.04 bp per helical turn. The helical period of the small central region can be most accurately determined by comparing the phases of sine functions that were fitted to the adjoining flanking regions. The difference in phase between these regions was found to be 68 ± 20, or about 0.2 turns (680/3600) of the DNA helix. Therefore, the region of DNA that contains the central 32 bp in the 5S nucleosome (positions -17 to + 15) has a periodicity of ± 0.06 bp per turn. The entire core DNA region is then calculated to have an average periodicity of ± 0.05 bp per turn, in close agreement with earlier estimates from sequence analysis of mixed-sequence nucleosome cores, which gave a periodicity of bp per turn (4, 13, 27). The Helical Periodicity of 5S DNA Changes Upon Nucleosome Formation. The average helical periodicity of bp per turn found for the DNA in the 5S nucleosome would be strong evidence for a change in helical period upon nucleosome formation, if, as would be expected, the periodicity of this same DNA is different when free in solution (8, 25). The helical periodicity of naked DNA in solution has been measured by cleaving DNA bound to calcium phosphate crystals with DNase (28-30), or with the hydroxyl radical (15). These experiments have yielded a range of values for the helical period of random sequence DNA from ± 0.1 bp per turn (28), to 10.5 bp per turn (15), to 10.4 bp per turn (29, 30). The average value from these experiments for the helical periodicity of DNA, bp per turn, is in good agreement with the value of ± 0.01 bp per turn that was determined by a completely different experimental approach (31). We consequently have analyzed the hydroxyl radical cleavage pattern of 5S DNA bound to calcium phosphate crystals to determine the helical periodicity of this DNA. Fig. 2 shows plots of the hydroxyl radical cleavage frequency of 5S DNA when free in solution, bound to calcium phosphate, and bound in a nucleosome. The frequency of a sine curve fitted to the cleavage pattern in 5S DNA bound to calcium phosphate yields a value of ± 0.05 bp per turn for the helical repeat of the 150 bp of 5S DNA (positions -80 to +70) that associates with the histone core (Fig. 3). To test our method of analysis, we have treated in the same way DNase cleavage data for random-sequence nucleosomal DNA bound to calcium phosphate (15). For this DNA we derive a helical periodicity of 10.5 bp per turn (results not shown), in substantial agreement with previous measurements (15, 28). A en -5 ~00 Cr C1) Proc. Natl. Acad. Sci. USA 87 (1990) ase Pairs per Turn ase Position FG. 3. (A) Plots of the frequency spectra from Fourier analysis of hydroxyl radical cleavage data. The data set analyzed for the peak labeled "DNA on Calcium Phosphate" was derived from the data shown in Fig. 2 for the DNA downstream of the start site of the 55 gene. The peak labeled "DNA in Nucleosome" is derived from data covering positions +16 to +74 in the "DNA in Nucleosome" data set shown in Fig. 28. The width of the Fourier transform peak is a function of the size of the data set used to calculate the modulus of the transform and is that expected for these particular data sets. () Least-squares fit of a sine function to a portion of the "DNA on Calcium Phosphate" data set (Fig. 2), which corresponds to the downstream half (positions -3 to +72) of the nucleosome binding site on the 55 gene. A similar analysis of the part of the "DNA on Calcium Phosphate" data set that corresponds to the upstream half of the nucleosome binding site (-3 to -78) yields an identical result. Shown is a plot of the original data set and the fitted sine curve. The fitted period, which gives the helical periodicity of this segment of DNA in solution, is indicated. We conclude that the segment of 55 DNA that forms a nucleosome, when free in solution, has a helical periodicity very close to that found for random sequence -form DNA in solution. Comparison of the cleavage patterns of 55 DNA in the nucleosome and bound to calcium phosphate demonstrates directly that the helical periodicity of 55 DNA changes significantly as a consequence of its interaction with the histone core. The change in periodicity is clearly evident in Fig. 28, where peaks in the hydroxyl radical cleavage frequency in the calcium phosphate and nucleosome plots are roughly aligned at approximately position +20, but are clearly out of phase a few helical turns away at +50 or +60. n the region encompassing positions +15 to +75 in the 55 gene, the helical period changes from about bp per turn in solution to bp per turn in the nucleosome (Fig. 3A). However, the average helical periodicity of DNA in the nucleosome is not as drastically different. The three helical turns that cross the center of

4 7408 iochemistry: Hayes et al. dyad symmetry, which have a periodicity of bp per turn, make the average periodicity of 5S DNA in the nucleosome core bp per turn. Evidence for the existence of a central region of altered periodicity in the bulk population of nucleosomes and the consequences for the linking-number paradox are discussed below. Relevance to Other Nucleosomes. t has been suggested that a nucleosome including 5S DNA may be an unusually stable structure and thus not representative of the bulk of nucleosomes (32). However, other results discussed below support the generality of our conclusions. The periodicities of DNase and DNase cleavage sites at the two ends of the DNA in nucleosomes containing mixed DNA sequences were close to 10.0 bp, while the spacing of those cleavage sites toward the center of the nucleosome was closer to 10.5 bp (13, 33). Photofootprinting of nucleosomes containing mixed-sequence DNA shows that thymidine dimer formation is modulated with a periodicity of about 10.0 bp on either side of a central region, which has a periodicity of about 10.5 bp. This central region, which spans the axis of dyad symmetry in the nucleosome core, offsets the phases of the outer regions by about 2-3 bp (34). Sequence analysis ofdna cloned from 177 nucleosome cores detects, near the dyad symmetry axis, a discontinuity in the periodic fluctuation of sequence content found in the rest of the nucleosomal DNA (13). Moreover, it has been noted that these modulations in sequence have a period of bases in regions on each side of the axis of dyad symmetry, which implies that the periodicity in the central region must average 10.7 bases to obtain the measured overall periodicity of bp per turn (27). The modulation patterns of sequence and thymidine dimer formation in the nucleosome thus closely parallel the structural modulations that are detected directly by the hydroxyl radical in the present work. The Change in Linking Number Upon 5S Nucleosome Formation and the Linking-Number Paradox. With the detailed information we have obtained on the helical periodicity of SS DNA in the nucleosome and in solution, we can now evaluate the structural changes that occur in SS DNA upon nucleosome formation. The linking-number paradox refers to the discrepancy in the change in linking number (ALk) observed upon nucleosome formation (-1) and expected from knowledge of the structure of the nucleosome (-2). The equation most often used to understand this process is ALk= ATw + AWr, [1] where ATw is the change in twist and AWr the change in writhe experienced by the DNA (6). The experimental determination of nuclease accessibility (or hydroxyl radical cleavage rate, as we do here) gives information on the helical repeat in the local frame of reference [i.e., with respect to the surface of the nucleosome core (6)]. The helical repeat of DNA in the local frame is related to the quantity 4), called the winding number, defined as the number of times that a vector associated with the DNA backbone rotates past the surface that is the reference frame (6). The helical repeat h, defined as the number of base pairs per unit winding number, or h = N/4), is the quantity measured in the experiments reported in this paper. 4) is not the same as the quantity Tw in Eq. 1. They are simply related, though, by the following equation (25): Tw=STw+ ), [2] where STw, the surface twist, is calculated from the geometry of the system. The corresponding equation for relating the changes in these quantities that occur upon nucleosome formation from DNA in solution is ATw = ASTw + A4). Proc. Natl. Acad. Sci. USA 87 (1990) ASTw, the change in surface twist upon forming the nucleosome, is calculated to be (6). n the experiments described in this paper, we have determined that the average helical repeat of 5S DNA bound in a nucleosome is ± 0.05 bp per turn, a value that agrees very well with the average helical periodicity of random sequence nucleosomes inferred by sequence analysis (4, 13, 28). Substituting this value into Eq. 3 along with the value for ASTw for the nucleosome and the value we determined for the helical repeat of SS DNA bound to a precipitate of calcium phosphate, and considering that there are 146 bp of DNA in the nucleosome core, we obtain ATw = ASTw + A(N/h) = (146/ /10.49) = [4] The change in writhe calculated for DNA assembling onto the nucleosome is (25), so, using Eq. 1, the change in linking number for formation of the 5S nucleosome is ± 0.13 (assuming an error of +0.1 in the calculated change in writhe). Although ALk for formation of a nucleosome on Xenopus 5S DNA has not been experimentally measured, the value we calculate above may be compared with the experimental value of ± 0.08 per nucleosome observed for Lytechinus variegatus 5S nucleosome arrays (35). t would thus seem that the observed alteration in the helical periodicity of SS DNA upon nucleosome formation does not, by itself, completely explain the linking-number paradox as currently formulated (6). This is entirely due to the difference in the helical periodicity of nucleosomal DNA of 10.0 bp per helical turn assumed in theoretical work (6) and the actual helical periodicity of bp per turn measured directly in this work and inferred from sequence analysis (4, 13). One reason for this discrepancy may be that in nucleosomal arrays (35) other sources of writhe and surface twist than the nucleosome exist. Moreover, it has been suggested that the nucleosome itself might not be best represented as a simple cylinder and that alternative paths of DNA may contribute to the experimentally observed change in linking number (36). The calculation above considers only the length of DNA found in the nucleosome core particle as isolated by micrococcal nuclease digestion of chromatin. More than 146 bp of DNA may be in contact with the histone core in nucleosome arrays. nspection of the hydroxyl radical footprints (Figs. 1 and 2) reveals a modulation of cleavage extending more than 90 bp from the center of dyad symmetry of the nucleosome (14), whereas the modulation of DNase cleavage only extends as far as position +70. The Positioning of Nucleosomes on DNA. Considerable evidence exists in support of sequence-directed positioning of nucleosomes on DNA (13, 14, 18, 37). Chicken, frog, and yeast histones recognize the same structural features of a 5S RNA gene, leading to the same nucleosome position. However, it should be noted that whether or not a nucleosome is positioned on the X. borealis somatic SS gene in vivo is not known. Mutagenesis experiments using sea urchin 5S DNA indicated that a region comprising bp to either side of the center of dyad symmetry of the nucleosomal DNA contained the major elements necessary for positioning (38). n addition, Ramsay (39) located two regions, centered about 1.5 helical turns to either side ofthe center ofdyad symmetry, that are important for positioning of a bacterial DNA fragment on a histone core. The hydroxyl radical detects variations in minor groove width (26). Such variations are apparent in the hydroxyl radical cleavage pattern of naked DNA containing the 5S RNA gene from X. borealis (Fig. 2). Quantitative analysis indicates a periodic reduction in minor groove width every bp (Fig. 2, "Free DNA" trace) and therefore raises

5 iochemistry: Hayes et al. the possibility that this stretch of DNA bends toward the narrowed grooves. More importantly, the features present in naked DNA are retained and exaggerated in the hydroxyl radical cleavage pattern of the nucleosome (Fig. 2, compare "DNA in Nucleosome" to "Free DNA'' data sets). positions of reduced minor groove width in the naked DNA, Thus, located where sequence-directed positioning elements are expected, are also places where the minor groove is narrowed when the DNA is bent around the histone core. Comparison of sequences found at these positions reveals few similarities, except for some A+T-rich character. Our results imply that these different sequences can adopt similar secondary structural conformations. We conclude that the hydroxyl radical provides a simple means to detect structural elements in free DNA that are likely to be functional in the positioning of a nucleosome. CONCLUSONS We have used hydroxyl radical footprinting to obtain highresolution data concerning the structure of a particular sequence of DNA in a nucleosome. From our results we conclude that the helical periodicity of 5S DNA is changed from ± 0.05 bp per turn in solution to ± 0.05 bp per turn over extensive regions on either side of the center of dyad symmetry of the nucleosome. ecause a shorter region around the center of dyad symmetry changes its helical periodicity in the other direction, to 10.7 bp per turn, the average helical periodicity of DNA in the 5S nucleosome is ± 0.05 bp per turn. Calculation ofthe contribution these changes in helical periodicity make to the change in linking number expected upon formation of the 5S nucleosome demonstrates that changes in helical periodicity do not completely explain the linking-number paradox as currently formulated (6). Detection by the hydroxyl radical of structural features in naked DNA that are maintained in nucleosomal DNA supports the assertion that the capacity ofdna to bend is a major determinant of nucleosome positioning. These studies demonstrate that the hydroxyl radical footprinting technique will prove a powerful analytical tool in determining the structural details of DNA in large nucleoprotein assemblies. We thank D. D. rown, M. E. A. Churchill, D. J. Clark, H. R. Drew, G. Felsenfeld, M. Gellert, A. Klug, R. Morse, D. Rhodes, R. T. Simpson, and A. A. Travers for critical reading of the manuscript. We also thank T. Flatley for the Fourier algorithm. We are grateful to V. Thuy for the preparation of the manuscript. 1. Kornberg, R. (1977) Annu. Rev. iochem. 46, Felsenfeld, G. (1978) Nature (London) 271, Proc. Natl. Acad. Sci. USA 87 (1990) Richmond, T. J., Finch, J. T., Rushton,., Rhodes, D. & Klug, A. (1984) Nature (London) 311, Drew, H. R. & Travers, A. A. (1985) J. Mol. iol. 186, Klug, A. & Lutter, L. C. (1981) Nucleic Acids Res. 9, White, J. H. & auer, W. R. (1989) Cell 56, Wang, J. C. (1982) Cell 29, Morse, R. H. & Simpson, R. T. (1988) Cell 54, Goulet,., Zivanovic, Y., Prunell, A. & Revet,. (1988) J. Mol. iol. 200, Zivanovic, Y., Goulet,., Revet,., Le ret, M. & Prunell, A. (1988) J. Mol. iol. 200, Klug, A. & Travers, A. A. (1989) Cell 56, Lutter, L. C. (1978) J. Mol. iol. 124, Satchwell, S. C., Drew, H. R. & Travers, A. A. (1986) J. Mol. iol. 191, Rhodes, D. (1985) EMO J. 4, Tullius, T. D. & Dombroski,. A. (1985) Science 230, Tullius, T. D., Dombroski,. A., Churchill, M. E. A. & Kam, L. (1987) Methods Enzymol. 155, Wolffe, A. P., Jordan, E. & rown, D. D. (1986) Cell 44, Losa, R. & rown, D. D. (1987) Cell 50, Peterson, R. C., Doering, J. L. & rown, D. D. (1980) Cell 20, ogenhagen, D. F. & rown, D. D. (1981) Cell 24, Camerini-Otero, R. D., Sollner-Webb,. & Felsenfeld, G. (1976) Cell 8, Simon, R. H. & Felsenfeld, G. (1979) Nucleic Acids Res. 6, Stein, A. (1979) J. Mol. iol. 130, Wolffe, A. P. (1988) EMO J. 7, White, J. H., Cozzarelli, N. R. & auer, W. R. (1988) Science 241, urkhoff, A. M. & Tullius, T. D. (1987) Cell 48, Travers, A. A. & Klug, A. (1987) Philos. Trans. R. Soc. London Ser. 317, Rhodes, D. & Klug, A. (1980) Nature (London) 286, ehe, M. & Felsenfeld, G. (1981) Proc. Natl. Acad. Sci. USA 78, Evans, T. & Efstratiadis, A. (1986) J. iol. Chem. 261, Horowitz, D. S. & Wang, J. C. (1984) J. Mol. iol. 173, Lorch, Y., LaPointe, J. & Kornberg, R. D. (1988) Cell 55, Lutter, L. C. (1979) Nucleic Acids Res. 6, Gale, J. M. & Smerdon, M. J. (1988) J. Mol. iol. 204, Simpson, R. T., Thoma, F. & rubaker, J. M. (1985) Cell 42, White, J. H., Gallo, R. & auer, W. R. (1989) J. Mol. iol. 207, Simpson, R. T. (1990) Nature (London) 343, FitzGerald, P. C. & Simpson, R. T. (1985) J. iol. Chem. 260, Ramsay, N. (1986) J. Mol. iol. 189,