Intra- and Inter-nucleosomal Protein-DNA Interactions of the Core Histone Tail Domains in a Model System*

The core histone tail domains are key regulators of eukaryotic chromatin structure and function and alterations in the tail-directed folding of chromatin fibers and higher order structures are the probable outcome of much of the post-translational modifications occurring in these domains. The functions of the tail domains are likely to involve complex intra- and inter-nucleosomal histone-DNA interactions, yet little is known about either the structures or interactions of these domains. Here we introduce a method for examining inter-nucleosome interactions of the tail domains in a model dinucleosome and determine the propensity of each of the four N-terminal tail domains to mediate such interactions in this system. Using a strong nucleosome “positioning” sequence, we reconstituted a nucleosome containing a single histone site specifically modified with a photoinducible cross-linker within the histone tail domain, and a second nucleosome containing a radiolabeled DNA template. These two nucleosomes were then ligated together and cross-linking induced by brief UV irradiation under various solution conditions. After cross-linking, the two templates were again separated so that cross-linking representing inter-nucleosomal histone-DNA interactions could be unambiguously distinguished from intra-nucleosomal cross-links. Our results show that the N-terminal tails of H2A and H2B, but not of H3 and H4, make internucleosomal histone-DNA interactions within the dinucleosome. The relative extent of intra- to inter-nucleosome interactions was not strongly dependent on ionic strength. Additionally, we find that binding of a linker histone to the dinucleosome increased the association of the H3 and H4 tails with the linker DNA region.

The core histone tail domains are key regulators of eukaryotic chromatin structure and function and alterations in the tail-directed folding of chromatin fibers and higher order structures are the probable outcome of much of the post-translational modifications occurring in these domains. The functions of the tail domains are likely to involve complex intra-and inter-nucleosomal histone-DNA interactions, yet little is known about either the structures or interactions of these domains. Here we introduce a method for examining inter-nucleosome interactions of the tail domains in a model dinucleosome and determine the propensity of each of the four N-terminal tail domains to mediate such interactions in this system. Using a strong nucleosome "positioning" sequence, we reconstituted a nucleosome containing a single histone site specifically modified with a photoinducible cross-linker within the histone tail domain, and a second nucleosome containing a radiolabeled DNA template. These two nucleosomes were then ligated together and cross-linking induced by brief UV irradiation under various solution conditions. After cross-linking, the two templates were again separated so that cross-linking representing inter-nucleosomal histone-DNA interactions could be unambiguously distinguished from intra-nucleosomal cross-links. Our results show that the N-terminal tails of H2A and H2B, but not of H3 and H4, make internucleosomal histone-DNA interactions within the dinucleosome. The relative extent of intra-to inter-nucleosome interactions was not strongly dependent on ionic strength. Additionally, we find that binding of a linker histone to the dinucleosome increased the association of the H3 and H4 tails with the linker DNA region.
Within the basic repeating subunit of eukaryotic chromatin known as the nucleosome, the nucleosome core is comprised of two copies of each of the core histones H2A, H2B, H3, and H4, and 147 bp of DNA. Approximately 75% of the core histone protein mass is organized into a "spool" onto which the nucleosome core DNA is tightly wrapped (1,2). This spool is formed primarily by the histone fold motif within each protein and additional structural elements adjacent to these domains (1,3). The structural details of these domains and their interactions with DNA have been well described (2)(3)(4)(5)(6)(7). Each nucleosome also includes a stretch of linker DNA, which joins nucleosome cores together in native chromatin, and, in most cases, a single linker histone (8). Arrays of nucleosomes are compacted into chromatin fibers of about 30 nm in diameter; these fibers are further condensed into poorly defined higher order structures (9,10).
About 25% of the total mass of the core histones is contained within the N-terminal tail domains, which were initially defined by their sensitivity to proteases (8,11). These domains are highly basic and bind DNA in vitro and in native chromatin (12)(13)(14)(15). In addition, evidence suggests that these domains also participate in protein-protein interactions between histones in the fully condensed chromatin fiber (3,7,16). Despite the perception of the tails as being "unstructured," biochemical and biophysical studies indicate they adopt defined structures and make specific interactions in chromatin (17)(18)(19)(20). The tail domains are not required for assembling or maintaining the structure of the nucleosome core and removal of the tails results in only marginal changes in the hydrodynamic shape, stability, and DNA wrapping within the nucleosome (21)(22)(23). However, the tail domains are essential for folding of oligonucleosomal arrays into native 30-nm chromatin fibers and are probably required for efficient assembly of fibers into higher order structures (9, 24 -26). Subsets of tails can promote partial folding capability and intermolecular association, suggesting that the tails have overlapping functions and the H4 tail appears to be the most critical for salt-dependent folding (27)(28)(29). However, the requirement for the tail domains, and particularly the H4 tail, for complete folding of oligonucleosomes into the fully folded chromatin fiber suggests the tails mediate specific molecular interactions during chromatin folding (9,16,29). Although it has not been directly demonstrated, it is generally assumed that at least some of these interactions include inter-nucleosomal contacts (3,7,9,16). However, the complex array of structures and interactions of the tail domains in chromatin remains relatively undefined.
In higher eukaryotes native chromatin also contains approximately one linker histone (histone H1) bound to each nucleosome (8,30). Nearly all linker histones contain an ϳ80-amino acid residue protease-resistant globular domain that directs binding of these proteins to the exterior of the nucleosome (31,32). Although linker histones bind with high affinity and positive cooperativity to naked DNA, they exhibit preferential binding to reconstituted nucleosomes in vitro (33). This binding stabilizes the wrapping of DNA about the histone octamer and directs the association of an extremely basic ϳ100 residue C-terminal tail that helps neutralize electrostatic repulsion caused by packing the polyanionic DNA backbone within the chromatin fiber (31, 34 -36). Thus linker histones generally stabilize the core histone tail-dependent intrinsic ability of nucleosome arrays to fold into chromatin fibers (36). Notably, it has been shown that the binding of linker histones can directly alter some of the interactions of the core histone tail domains within individual nucleosomes (18,37).
Histone tails undergo numerous post-translational modifications, which either directly or indirectly alter chromatin structure to facilitate transcriptional activation or repression or other nuclear processes (38 -41). Thus modifications within these domains are an end point to many signal-transduction pathways directed to the nucleus of the cell (42). Whereas some modifications generate recognition sites for the binding of specific proteins such as the heterochromatin-associated protein HP1 to the tails (43,44), other modifications can directly alter chromatin structure (45,46). For example acetylation of lysine residues results in a reduction of positive charge and it has been demonstrated that mutation-based alterations in overall charge can mimic the essential function(s) of acetylation in a histone tail domain in vivo (47). This nonspecific charge effect is perhaps because of a reduced affinity or interaction of the tail with DNA in chromatin, leading to an alteration of chromatin structure. On the other hand, UV-laser cross-linking studies suggest that acetylation does not lead to a loss of tail-DNA interactions (13) and other critical post-translational modifications such as methylation do not lead to an alteration of overall charge in the tail domains. Thus some modifications are likely to directly evoke distinct changes in tail structures (48). However, a detailed understanding of the mechanism(s) by which post-translational modifications of the histone tails leads to altered functional states of chromatin must await an understanding of the basic structures and interactions of the tail domains.
Here we introduce a novel dinucleosome system for investigation of core histone tail-DNA interactions. This system allows unambiguous identification of intra-and inter-nucleosomal interactions between the histone tail domains and DNA under a variety of conditions, including those that induce folding of the linker DNA in dinucleosomes (49 -51). We find that a subset of the tail domains contact the DNA of the adjacent nucleosome although the H3 and H4 tails appear to participate in only intra-nucleosomal interactions. Interestingly, the relative extent of inter-nucleosomal interactions does not exhibit a detectable dependence on ionic strength, but binding of a linker histone increases the association of H3 and H4 tails with the linker DNA region.

EXPERIMENTAL PROCEDURES
DNA Fragments-DNA templates for nucleosomes 1 and 2 were derived from plasmid pXbs-1 containing a Xenopus borealis somatic 5 S rRNA gene (52). PCR techniques were used to add HindIII and DraIII sites to the ends of the nucleosome 1 DNA template using primers 5Ј-CAGTAAGCTTGATTCCCGGGCTTGTTTTCCTGCCTG-3Ј and 5Ј-C-ATGTCTAGAGATCCCGGGACTACACGAGGTCCCATCCAAGTAC-3Ј and DraIII and XbaI sites to nucleosome 2 using primers 5Ј-CAGTAA-GCTTGATTCACCTCGTGCTTGTTTTCCTGCCTG-3Ј and 5Ј-CATGTC-TAGAGATCCCGGGACTACGGTCTCCCATCCAAGTAC-3Ј (Fig. 1). The resulting nucleosome 1 is 205 bp, and nucleosome 2 is 209 bp. Ligation of the two templates at the asymmetric DraIII site creates a tandem repeat of 5 S sequences with a repeat of 210 bp. The two templates were cloned into pBluescript plasmids pBSXN1 and pB-SXN2. DNA fragments from each were prepared in large quantity by first digesting 10 mg of pBSXN1 or pBSXN2 with HindIII or XbaI, respectively, dephosphorylation with calf intestinal phosphatase (New England Biolabs), and filling in the 5Ј overhang with Klenow DNA polymerase in the presence of 100 M each of dATP, dGTP, dTTP, and ddCTP for pBSXN1 or dCTP, dTTP, dGTP, and ddATP for BSNX2. The plasmid was then cut with DraIII and the released template DNAs isolated on preparative 1.2% agarose gels, and recovered by electroelution. Following phenol extraction and repeated ethanol precipitation, the pellet was dissolved in 10 mM Tris (pH 8.0) and stored frozen at Ϫ20°C until needed. For each dinucleosome reconstitution radioactively labeled nucleosome templates 1 or 2 were prepared by digesting 10 mg of either parent plasmid with HindIII or XbaI, respectively. The free DNA ends were dephosphorylated as above and the DNA was stored frozen. For each labeling 100 g of either digested plasmid was incubated with Klenow in the presence of 500 Ci of [␣-32 P]dATP, and 100 M each of dGTP, dTTP, and ddCTP ([␣-32 P]dCTP, dTTP, dGTP, and ddATP for pBSXN2) for 15 min at room temperature, then 100 M cold dATP (dCTP for pBSXN2) was added for an additional 10 min. The labeled template was isolated on a 6% acrylamide gel by standard techniques and stored in TE. B, mononucleosomes were efficiently ligated into dinucleosomes. In this experiment nucleosome 1 was radioactively labeled and samples were run in a 0.7% agarose nucleoprotein gel. Lane 1, mononucleosome DNA; lane 2, gradient purified mononucleosome 1; lane 3, ligated dinucleosome DNA; lanes 4 and 5, nucleoprotein products from the ligation reaction before and after gradient purification of the dinucleosome, respectively. C, two H1s bind model dinucleosomes. Gradient-purified dinucleosomes were incubated with histone H1 and then products were resolved on a 0.7% agarose nucleoprotein gel. Lane 1, dinucleosome DNA template shown as a control; lane 2, gradient-purified dinucleosomes; lanes 3-6, dinucleosomes incubated with 4.0, 2.0, 1.5, and 1.0 mol of H1/mol of dinucleosome, respectively.
Preparation of Native and APB-modified Histones-Recombinant wild type and mutant H2A, H2B, H2A-G2C, and H2B-2C were prepared and purified as preformed dimers as described (19,53). Coding sequences for Xenopus H3-T6C/C110A (containing cysteine and alanine codon substitutions at amino acid positions 6 and 110, respectively) and H4-G6C (containing a cysteine codon at position 6) were constructed and cloned into pET3 expression plasmids. The individual core histones and mutant proteins were expressed individually in bacteria and then H2A/H2B dimers were purified as described previously (53) and H3/H4 tetramers were prepared via the following procedure. As these proteins were present in insoluble forms in traditionally prepared lysates, we first determined the mass of proteins within each bacterial cell culture and then mixed cultures containing roughly equal amounts of H3 and H4, H3-T6C/C110A and H4, or H3-C110A and H4-G6C. A pellet containing both proteins (from one combined liter of culture) was suspended in 100 ml of P1 buffer (50 mM Tris, 10 mM EDTA, 100 g/ml RNase, pH 8.0), and then cells were lysed with an equal volume of P2Ј buffer (0.2 M NaOH, 0.4% Triton X-100) in the presence of 0.4 mM phenylmethylsulfonyl fluoride, 8 mM dithiothreitol at room temperature for an hour. The lysate was dialyzed against 2 M NaCl, TE buffer (pH 8.0, 10 mM ␤-mercaptoethanol) at 4°C for at least 16 h, with three buffer exchanges. The sample was centrifuged at 15,000 ϫ g for 30 min, and the pellet discarded. For each 10 ml of supernatant 30 ml of TE and 100 l of 1 M dithiothreitol were added, then 1 ml of exchange resin was added per ml of solution (Bio-Rex 70, 50 -100 mesh, 50% slurry). Proteins were allowed to bind the resin for 4 h at 4°C. The resin was washed extensively with 0.6 M NaCl in TE and H3/H4 tetramer was eluted with 2 M NaCl in TE. Fractions containing the tetramer were sonicated to reduce the size of contaminating nucleic acids and the Bio-Rex resin purification was repeated. The fractions containing H3/H4 tetramers were mixed with 50 l of hydroxyapatite resin per ml (Bio-Rad, 50% slurry) for 30 min at 4°C, and the resin was removed by centrifugation. This process was repeated 3 times to eliminate final traces of contaminating nucleic acids (54). The proteins were checked by SDS-PAGE and trial reconstitutions. APB modification of H2A-G2C, H2B-2C, H3-T6C, and H4-G6C was carried out as previously described (19,55).
Preparation of Model Dinucleosomes-Nucleosomes 1 and 2 were prepared separately and then ligated together. In most cases one nucleosome was reconstituted on a cold DNA template and the second reconstituted on a radioactively end-labeled template prepared as described above. Nucleosome reconstitution was carried out by a standard salt dialysis procedure (53). Typically, 6.6 g of unlabeled or ϳ3 g of labeled template fragment, respectively, mixed with 15 g of H2A/H2B, 15 g of H3/H4, and 40 g of linearized Bluescript plasmid with blunt ends (as competitor) were used for each reconstitution. Because of slight variations in different preparations of DNA fragments and APB-modified histone mutants, the ideal ratio was empirically determined. Reconstituted mononucleosomes were loaded onto 10 ml of 5-20% sucrose gradients prepared in 10 mM Tris (pH 8.0), 1 mM EDTA (Beckman 41 Ti rotor, 34 krpm, 18 h). Fractions containing purified mononucleosomes were identified by analysis in 0.7% agarose nucleoprotein gels. The nucleosome fractions were concentrated 10-fold by centrifugation through a Microcon YM-50 filter membrane. The ligation reactions typically contained 50 l (5 or 2.5 pmol of unlabeled or labeled nucleosomes, respectively) each of nucleosomes 1 and 2, 20 g of bovine serum albumin, and 3,200 units of T4 ligase (New England Biolabs) were kept at room temperature for 30 min in a total volume of 160 l, followed by the addition of 20 l of 10 mM ATP and 20 l of 20 mM Mg 2ϩ for an additional 60 min. The ligation product was purified by sucrose gradient centrifugation as above except for 15 h at 4°C.
Restriction Enzyme Digestion and Hydroxyl Radical Footprinting of Nucleosomes-Sucrose gradient-purified mononucleosomes and dinucleosomes were subjected to restriction enzyme digestion as described (56). Dinucleosomes used for hydroxyl radical footprinting were first subjected to buffer exchange to 10 mM Tris (pH 8.0), 1 mM EDTA on a Microcon YM-50 concentrator to eliminate sucrose.
UV-induced Cross-linking and Identification of Cross-linking Bands-Gradient-purified dinucleosomes were subjected to buffer exchange as above, and then 32 l (ϳ0.1 pmol) was rapidly mixed with 4 l of bovine serum albumin (1 g/l), and 4 l of concentrated NaCl solutions. The mixture was placed into the bottom of a Falcon 5-ml polystyrene tube, which was inside a 15-ml Pyrex 9820 glass tube. The sample was irradiated with 365 nm UV light generated by a VMR LM-20E light box for 1 min. The sample was then mixed with 5 l of 1 M dithiothreitol, 2 l of 10% SDS, 2 l of 1.5 g/l calf thymus DNA, and incubated at 37°C for 10 min, followed by adding 85 l of ddH 2 O and standard ethanol precipitation. The pellet was dissolved in 30 l of 10 mM Tris (pH 8.0) and then 13 l of the sample was digested with DraIII, in a total volume of 30 l at 37°C for 3 h. Another 13-l portion was diluted to 30 l with ddH 2 O and mock-digested at 37°C for the same time. The samples then mixed with SDS and loaded onto a 6% acrylamide gel containing Tris-glycine-SDS buffer and run at 100 V for 14 h before gel drying and PhosphorImager analysis.

RESULTS
Assembly of the Dinucleosome-To assess potential internucleosomal interactions of the core histone tail domains, we assembled a model dinucleosome in which one nucleosome contained native histones and a radiolabeled DNA fragment while the second contained one histone with a photoactivatable cross-linking probe located within the tail domain. To assemble the dinucleosome, the two individually reconstituted nucleosomes were ligated together (Fig. 1). To ensure proper orientation upon ligation, we placed an asymmetric recognition sequence for the restriction enzyme DraIII at appropriate ends of the DNA templates (Fig. 1) and polished the opposite ends (filled in with a terminal dideoxynucleotide and dephosphorylated) to prevent self-ligation (see "Experimental Procedures"; Fig. 2A). We found that efficient ligation of the radioactive nucleosome required this nucleosome to be present with at least 10 nM concentration and a 2-3-fold excess of the unlabeled nucleosome (Fig. 2). Ligated dinucleosomes were isolated by centrifugation through sucrose gradients to remove any unli- Characterization of the Model Dinucleosome-Purified ligated dinucleosomes migrate through sucrose gradients and in agarose gels as expected based on migration of mononucleosomes and native dinucleosome controls ( Fig. 2B and data not  shown). Also, similar to other dinucleosomes (57), our ligated dinucleosome showed a two-step shift when bound by increasing amounts of histone H1 (Fig. 2C). To further check integrity, gradient-purified dinucleosomes were subjected to hydroxyl radical footprinting and restriction enzyme digestion assays. By labeling different ends of the dinucleosome, we compared the hydroxyl radical footprinting patterns of each end of the template with the respective mononucleosomes. As shown in Fig. 3, the cleavage patterns from nucleosomes 1 and 2 in the model dinucleosome were very similar to corresponding mononucleosomes, indicating that the ligation process did not alter nucleosome rotational orientation or position or select for minor populations of mononucleosomes with alternative positioning. Mapping of nucleosome positions in the mono and model dinucleosome via restriction enzyme site protection assays substantiated the footprinting results (Fig. 4). Again, the extents of protection at each site in the mono-and dinucleosomes were similar, indicating the mononucleosome positions were not significantly altered when ligated into dinucleosomes.
The H2A N-terminal Tail Makes Intra-nucleosomal DNA Contacts in a Dinucleosome-To assess contacts to DNA made by the H2A tail in the model dinucleosome, we attached a photoactivatable cross-linking probe to a position near the tip of H2A N-terminal tail (H2A-G2C-APB) in nucleosome 2, whereas the HindIII end of nucleosome 1 was radioactively labeled. Cross-linking was initiated as described under "Experimental Procedures" and the extent of intra-versus inter-nucleosomal contacts to DNA by the H2A tail measured as the relative abundance of cross-linked species before and after DraIII digestion. Before DraIII digestion, a strong cross-linking band was observed above the full-length dinucleosome template band, dependent on UV irradiation (Fig. 5, lanes 1 and 2). This band represents both intra-and inter-nucleosomal crosslinking to DNA by the tail of H2A (Fig. 5, graphic). Digestion with DraIII reveals that ϳ15-20% of the total cross-linking is because of inter-nucleosomal DNA binding by the H2A Nterminal tail (Fig. 5, lane 10).
To substantiate the above results we placed both the radioactive label and H2A-G2C-APB within the same nucleosome (Fig. 6, schematic). Again, UV irradiation results in the generation of a band corresponding to the cross-linked species (Fig. 6,  lane 2). After DraIII digestion, the majority (ϳ80%) of the cross-linked band is converted to a slower migrating crosslinked species (Fig. 6, lane 9). In addition, we note that increases in salt concentration did not significantly alter the fraction of the total cross-links that were inter-nucleosomal, but did reduce the overall efficiency of cross-linking (Figs. 5 and 6, lanes 10 -16 and 9 -14, respectively). Although the basis for this reduction has not been completely defined, it is due in part to a moderate effect of salt on the efficiency of the crosslinking reaction as a similar effect of ionic strength on crosslinking is observed with mononucleosomes. 1 APB H2AA12C Does Not Participate in Inter-nucleosomal Interactions-We next determined the extent to which a position near the first ␣-helix in H2A is able to mediate internucleosome interactions. This position is located at the junction of the H2A N-terminal tail and the histone fold domain, near the inside of the superhelical DNA gyre, and therefore is less likely to participate in inter-nucleosomal DNA binding (3,(5)(6)(7)58). Previous experiments have shown that irradiation of mononucleosomes containing H2A-A12C-APB results in crosslinks to DNA located ϳ40 bp to either side of the nucleosome dyad, consistent with the known structure of this complex (3,19). Similar to the previous experiment the HindIII end of nucleosome 1 in the dinucleosome was radiolabled, and H2A-A12C-APB was incorporated into nucleosome 2. Upon UV irradiation a band indicating cross-linking between H2A-A12C-APB and DNA was generated but DraIII digestion showed that the observed cross-links were virtually in their entirety intranucleosomal (Fig. 7, compare lanes 2 and 9). As previously observed, the total extent of cross-linking was reduced as the NaCl concentration was raised to 100 mM (Fig. 7, lanes 2-7). To confirm this result, in a second experiment we put both the radioactive label and H2A-A12C-APB in the same nucleosome. In this case, after DraIII digestion, nearly 100% of the crosslinking was converted to the slower migrating species (Fig. 7,  lanes 15-18) The H2B N-terminal Tail Makes Inter-nucleosomal Contacts with DNA-To probe potential internucleosomal interactions by the N-terminal tail domain of H2B, we radiolabeled the HindIII end of nucleosome 1, and placed H2B-2C-APB in nucleosome 2. Protein-DNA cross-linking bands were generated upon UV irradiation under different salt concentrations (Fig. 8,  lanes 2-8). Upon separation of the two nucleosome templates by DraIII digestion, protein cross-linked to mononucleosome DNA was observed, indicating that the H2B N-terminal tail participates in inter-nucleosomal interactions (Fig. 8, lanes  10 -16). Again, we noticed that the relative proportion of H2B inter-nucleosomal interactions did not change upon increasing ionic strength of the solution.
H3 and H4 Tail Interactions in the Model Dinucleosome-We probed for internucleosomal interactions by the N-terminal tail domains of H3 and H4 to DNA in a manner identical to that used for H2A and H2B. Strong protein-DNA cross-linking to the dinucleosome template was observed when the cross-linking probe was placed near the tip of either the H3 or H4 tail in nucleosome 2 and the template of nucleosome 1 contained a radioactive label. However, when the individual nucleosome templates were separated via DraIII digestion, no protein cross-linking to the DNA of nucleosome 1 was observed, regardless of the ionic strength tested (Figs. 9 and 10, lanes 6 -7 and  8 -11, respectively). These results suggest that H3 and H4 tails do not interact with the DNA of the neighboring nucleosome in our model dinucleosome system.
Effect of Linker Histone on Tail Interactions-Previously we and others showed that the binding of linker histones affects the interactions of some of the histone tail domains with DNA in mononucleosomes (18,37). To investigate potential effects of H1 binding on internucleosome interactions in our model dinucleosome, we incubated templates with H1 and repeated the cross-linking experiments. H1 was carefully titrated with respect to dinucleosome concentration up to two molecules of H1 per dinucleosome template (Fig. 2C). The binding of linker histone did not induce the formation of bands corresponding to internucleosomal cross-linking by the H3 and H4 tail domains (Fig. 9, lanes 8 -10; Fig. 10, lanes 12-14). However, H1 binding did induce the formation of cross-linked species refractory to digestion by DraIII (Fig. 9, lanes 8 -10, Fig. 10, lane 12). Such bands are likely because of the formation of cross-links in and around the cognate DNA site for DraIII in the linker DNA. Interestingly, the binding of H1 had little or no apparent effect on the total extent of internucleosomal or linker DNA interactions by the H2A or H2B N-terminal tail domains (data not shown). We note, however, that because DraIII did not cut the cross-linked dinucleosome band completely even in the absence of linker histone in the H2A experiments (Fig. 5, lanes 10 -16), it is possible that minor H1-dependent increases in cross-linking would not be detected. In this case, presumably some crosslinking of the H2A tail to the linker DNA in the absence of H1 binding prevents DraIII cutting. DISCUSSION Our results support the emerging view that different histone tails have different structural roles in chromatin (9, 59). We find that in our model dinucleosome the N-terminal tails of H2A and H2B make detectable interactions with the DNA of a neighboring nucleosome, whereas the tails of H3 and H4 apparently make exclusively intra-nucleosomal histone-DNA interactions. If interactions within our model system resemble those found within native chromatin, our results imply that dislocation of H2A/H2B dimers, such as has been suggested to occur in transcriptionally active chromatin (60,61) would lead to a loss of inter-nucleosomal interactions and a drastic loosening of compact chromatin structure (62). It has been shown that loss of only one or two sets of such contacts within a much larger array of nucleosomes significantly destabilizes folding of the array into a condensed higher order structure (63). This may be significant in light of the relatively high rate of transcription-coupled turnover of the H2A/H2B dimer in vivo, compared with the (H3/H4) 2 tetramer (64 -66).
Previously, both site-directed and general cross-linking techniques have been used to map interactions of the core histone tail domains (67)(68)(69)(70). However, much of this work has been done with either single nucleosomes or with native chromatin where inter-nucleosomal tail interactions were difficult or impossible to quantify. Our system allows unequivocal identification of inter-nucleosomal interactions. Radiolabeled mononucleosome template covalently cross-linked to histone protein can only arise from internucleosome interaction of a tail domain with DNA of the adjoining nucleosome. Such interactions, representing communication between adjacent nucleosomes, were not detected for histone H3 or H4 N-terminal tails, yet these tails, and particularly the H4 tail, are clearly required for formation of the fully condensed chromatin fiber (27)(28)(29). Interestingly, recent results suggest that the H3 tail does not contribute significantly to the stability of the folded chromatin fiber and that the predominant contribution of the H4 tail appears to coincide with a region thought to interact with an acidic patch on the surface of the H2A/H2B dimer, perhaps in an adjacent nucleosome (3,29). These results are consistent with our observation suggesting that these tails do not make contacts to the DNA of an adjacent nucleosome in chromatin. In addition, it is unclear if H3 or H4 tails would make contacts between successive nucleosomes along the DNA strand (N to N ϩ 1) or if these tails would interact preferentially with nucleosomes beyond the nearest neighbor (N to N ϩ 2, N ϩ 3 . . . ). Indeed, dependent on the actual structure of the chromatin fiber, nucleosome N may be physically closest to nucleosomes N ϩ 2 or N ϩ 3 (71). These observations are also consistent with the idea that the H3/H4 tails are the primary mediators of intra-nucleosomal DNA accessibility (56,72).
It has been well established that raising the monovalent cation concentration in a solution can effect condensation or compaction of an extended array of nucleosomes into a form resembling a contacting zig-zag or "10-nm" structure (8,9,(73)(74)(75). Furthermore, even in the absence of linker histones, addition of small amounts of divalent cations (i.e. Mg 2ϩ ) can induce further compaction of oligonucleosomes to a structure with the same hydrodynamic shape as a fully condensed ϳ30 nM chromatin fiber (9,73). In addition, chromatin fragments as small as dinucleosomes can undergo some amount of salt-dependent condensation, perhaps related to the condensation of FIG. 9. Binding of H1 increased H3 tail-linker DNA interactions in the dinucleosome. Dinucleosomes were prepared with a radiolabel at the HindIII end of nucleosome 1 and H3-T6C-APB in nucleosome 2, irradiated, and cross-links were analyzed as described. All samples shown underwent buffer exchange to 10 mM Tris (pH 8.0), 1 mM EDTA after gradient purification. Samples in lanes 1 and 6 contain no NaCl; lanes 2-5 and 7-10 contain 50 mM NaCl; lanes 3-5 and 8 -10 contain 2.0, 1.5, and 1.0 mol of H1 per mol of dinucleosome, respectively; lanes 6 -10 were digested with DraIII after UV irradiation. Scheme (right) is as described in the legend to Fig. 5. oligonucleosomes (49 -51). Given the role of the histone tail domains in fiber compaction (24,25,27), we hypothesized that one or more of the tails may exhibit an alteration in the extent of inter-nucleosome interactions upon transition to more condensed states. However, in our system the relative proportion of inter-nucleosome interactions did not increase with increasing NaCl (or upon the addition of Mg 2ϩ , data not shown) at concentrations reported to cause condensation of the dinucleosome (49 -51). It is possible that although the conformation of the dinucleosome changes, the interactions of histone tails with DNA do not rearrange in such a way resulting in increasing inter-nucleosome tail-DNA interactions. Alternatively, it is possible that adjacent nucleosomes do not come in close proximity in native chromatin, as has been proposed for some models of the 30-nm chromatin fiber (see above) (71). In addition, some of the tail domains may participate in inter-nucleosomal protein-protein interactions in condensed chromatin (9,76). In this regard, we note that two grooves in the nucleosome protein surface have been proposed to bind histone tails from adjacent nucleosome in native chromatin, based on interactions observed in x-ray crystal structures of nucleosome cores or clusters of histone mutations interrupting silencing (3,7,76). Experiments assessing inter-nucleosomal histone tail-protein interactions will address these issues.