Identification of Specific Functional Subdomains within the Linker Histone H10 C-terminal Domain*

Linker histone binding to nucleosomal arrays in vitro causes linker DNA to form an apposed stem motif, stabilizes extensively folded secondary chromatin structures, and promotes self-association of individual nucleosomal arrays into oligomeric tertiary chromatin structures. To determine the involvement of the linker histone C-terminal domain (CTD) in each of these functions, and to test the hypothesis that the functions of this highly basic domain are mediated by neutralization of linker DNA negative charge, four truncation mutants were created that incrementally removed stretches of 24 amino acids beginning at the extreme C terminus of the mouse H10 linker histone. Native and truncated H10 proteins were assembled onto biochemically defined nucleosomal arrays and characterized in the absence and presence of salts to probe primary, secondary, and tertiary chromatin structure. Results indicate that the ability of H10 to alter linker DNA conformation and stabilize condensed chromatin structures is localized to specific C-terminal subdomains, rather than being equally distributed throughout the entire CTD. We propose that the functions of the linker histone CTD in chromatin are linked to the characteristic intrinsic disorder of this domain.

Linker histone binding to nucleosomal arrays in vitro causes linker DNA to form an apposed stem motif, stabilizes extensively folded secondary chromatin structures, and promotes self-association of individual nucleosomal arrays into oligomeric tertiary chromatin structures. To determine the involvement of the linker histone C-terminal domain (CTD) in each of these functions, and to test the hypothesis that the functions of this highly basic domain are mediated by neutralization of linker DNA negative charge, four truncation mutants were created that incrementally removed stretches of 24 amino acids beginning at the extreme C terminus of the mouse H1 0 linker histone. Native and truncated H1 0 proteins were assembled onto biochemically defined nucleosomal arrays and characterized in the absence and presence of salts to probe primary, secondary, and tertiary chromatin structure. Results indicate that the ability of H1 0 to alter linker DNA conformation and stabilize condensed chromatin structures is localized to specific C-terminal subdomains, rather than being equally distributed throughout the entire CTD. We propose that the functions of the linker histone CTD in chromatin are linked to the characteristic intrinsic disorder of this domain.
Linker histones comprise a family of small nucleosome-binding proteins that have a short N terminus, a central winged helix-like globular domain, and a long, highly basic C-terminal domain (CTD) 1 (1)(2)(3). Binding of linker histones to nucleosomal arrays in vitro influences chromatin fiber structure at multiple levels. It has been widely observed that linker histones protect an additional 20 bp of DNA from micrococcal nuclease digestion (2)(3)(4). Upon direct examination, nuclease protection appears to result from linker histone-linker DNA interactions involved in the formation of an apposed linker DNA stem motif (5), rather than from wrapping additional DNA around the nucleosome as was initially believed (2,3). Linker histones also have important functions in chromatin condensation. They stabilize locally folded secondary chromatin structures, e.g. "30-nm diameter fiber" (2,3,6,7), and facilitate the self-association of fibers into oligomeric tertiary chromatin structures thought to be relevant to global chromosomal fiber organization (6 -8). The linker histone globular domain plays a major role in protecting additional DNA from nuclease digestion (2,3). In contrast, the ability to stabilize extensively folded secondary chromatin structures lies exclusively in the linker histone CTD (9 -11). The linker histone domains that facilitate fiber self-association have yet to be defined.
The biochemical properties of the linker histone CTD are enigmatic. In most isoforms, this domain consists of ϳ100 amino acid residues. There is no CTD sequence conservation among the linker histone isoforms (2,3). However, ϳ40% of each somatic linker histone CTD consists of lysine residues that are very evenly distributed throughout the domain (12). The fact that peptides derived from the linker histone CTD have no detectable secondary structure in solution (3), together with the extensively basic character of this domain, has led to the proposal that the CTD functions in chromatin as an unstructured cationic stretch of amino acids that binds to linker DNA and neutralizes negative charges (11,12).
There are, however, several hints that the structural features of the linker histone CTD may be more complex. In addition to the large percentage of basic amino acid residues, alanine, serine, threonine, and proline residues also are frequently found in the CTD of mouse H1 0 , chicken H5, and all human somatic linker histones. By contrast, the CTDs of all linker histone isoforms are almost completely deficient in the acidic, aromatic, and highly hydrophobic amino acids. Protein domains having this distinctive amino acid composition are thought to possess "intrinsic disorder" (13)(14)(15)(16), which is characterized by molten globule-like structure in the native state. Intrinsically disordered regions frequently assume classical secondary structure when interacting with other macromolecules (13)(14)(15)(16). Consistent with this notion, the CTD peptides adopt a detectable ␣-helical structure when interacting with DNA (17,18), and also in high salt solutions and organic solvents (17, 19 -21). The CTD of all somatic linker histones also contain one or more S/TPKK DNA binding sequences (17,19,22,23), which form ␤-turn motifs that bind to the DNA minor groove (22,24) and mediate condensation of naked DNA in vitro (25)(26)(27).
In the present study we use biochemically defined model systems to probe the mechanistic basis of linker histone CTD function in chromatin. We initially compared the biochemical properties of native chicken erythrocyte H5 and recombinant mouse H1 0 . Subsequently, the structural and functional effects of incremental deletions of the mouse H1 0 CTD were determined. Our results indicate that distinct subdomains within the CTD are responsible for mediating linker histone effects on linker DNA conformation, and stabilization of condensed chromatin fiber structures. A revised mechanism for CTD function in chromatin is proposed.
Histone Purification-Core histone octamers and native chicken erythrocyte linker histone H5 were purified as described previously (30,31). To obtain full-length and mutant H1 0 proteins, pET-H1 0 -11d, pET-H1 0 C⌬24 -11d, pET-H1 0 C⌬48 -11d, pET-H1 0 C⌬72-11d, and pET-H1 0 C⌬97-11d were transformed into E. coli BL21(DE3) pLysS-competent cells. The transformed cells were grown at 37°C, harvested, and washed as described (28). The cells were sonicated, and the H1 0 proteins were purified as described (32,33)  for H1 0 , H1 0 C⌬24, H1 0 C⌬48, H1 0 C⌬72, and H1 0 C⌬97, respectively. Sonicated cells were incubated on ice for 30 min, and then pelleted. The NaCl concentration of the supernatants was decreased to 0.3 M by dilution, the supernatants were mixed with pre-hydrated CM-Sephadex C-25 (C-25) (Sigma), and the mixtures gently rocked for 3 h at 4°C to allow the binding of the H1 0 proteins to the C-25. The mixtures were centrifuged to pellet the C-25, the resin was washed in 10 mM Tris⅐HCl, 1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, 300 mM NaCl at the same pH as that of the lysis buffer, and the pelleted resin was loaded onto a C-25 column (2.5 ϫ 25 cm) pre-equilibrated with the appropriate buffer. The proteins were eluted from the C-25 column with a 0.3-1.0 M NaCl gradient in 10 mM Tris⅐HCl, 1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride at the same pH as the lysis buffer. At the completion of these steps, H1 0 , H1 0 C⌬48, and H1 0 C⌬97 required no further purification. However, we routinely observed that the fractions containing H1 0 C⌬24 and H1 0 C⌬72 were contaminated with degradation products. Consequently, the C-25 fractions containing either H1 0 C⌬24 or H1 0 C⌬72 were combined, adjusted by dilution to 10 mM Tris⅐HCl, 1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, ϳ300 mM NaCl at the same pH as that of the lysis buffer, and loaded onto a 5-ml HiTrap SP HP column (Amersham Biosciences). The mutant H1 0 proteins were then eluted with the same gradients used for the C-25 column. The purity of all H1 0 proteins was determined by SDS-PAGE and matrixassisted laser desorption ionization time-of-flight mass spectrometry. The concentrations of the purified proteins were determined using a BCA protein assay kit (Pierce).
Analytical Ultracentrifugation-Sedimentation velocity experiments were performed using either a Beckman XL-A or XL-I ultracentrifuge as described (6,7,36). Boundaries were analyzed by the method of van Holde and Weischet (37) using Ultrascan (version 5.0) software. Data were plotted as boundary fraction versus s 20,w to yield the integral distribution of sedimentation coefficients, G(s). Average sedimentation coefficients were obtained at boundary fractions equal to 0.5 of the G(s) plot. For folding experiments, arrays were mixed with an equal volume of 2ϫ MgCl 2 stock solutions prior to sedimentation to obtain the desired final MgCl 2 concentration. The final absorbance at 260 nm of all samples was 0.6 -0.8.
Agarose Multigel Electrophoresis-Electrophoretic mobilities, , were measured in a multigel composed of 9 individual agarose running gels ranging in concentration from 0.2 to 1.0%. The casting and running buffer was 40 mM Tris acetate (pH 7.8), 0.25 mM Na 2 EDTA. Samples were loaded and electrophoresed at 1.33 V/cm for 6 h. The temperature was 24 Ϯ 3°C. The gels were stained with ethidium bromide and the gel image was digitized. For each individual band, the migration was measured from the center of the well to the center of the band using NIH Image software and subsequently converted to . The effective macromolecular radius (R e ) and free-electrophoretic mobility ( o ) were obtained from as described (6,7).
Self-association Assay-The salt-dependent self-association of H5and H1 0 -bound nucleosomal arrays was determined using a differential centrifugation assay as described (6,36). For each salt concentration assayed, data were plotted as the percentage of the initial sample absorbance that remained in the supernatant after centrifugation at 16,000 ϫ g for 5 min in a microcentrifuge.

The Ability of Chicken Erythrocyte H5 and Recombinant
Mouse H1 0 to Bind to Nucleosomal Arrays, Alter Linker DNA Conformation, and Stabilize Condensed Secondary and Tertiary Chromatin Structures, Is Indistinguishable-The experiments described below extend our previous studies of linker histone structure-function relationships (6,7,38) by investigating how specific deletion mutations in the linker histone CTD influence the primary, secondary, and tertiary structure of biochemically defined nucleosomal arrays. The linker histone used in our earlier experiments was H5 purified from chicken erythrocytes. However, given that recombinant chicken erythrocyte H5 is expressed very poorly in bacterial cells (39), this linker histone isoform is a poor candidate for mutagenesis studies. Because the H1 0 isoform is considered to be the mammalian functional homologue of chicken H5 (40,41), and mouse H1 0 can be readily expressed and purified from E. coli (28), recombinant mouse H1 0 was chosen for all studies involving recombinant proteins.
To determine whether the in vitro properties of purified chicken H5 and recombinant mouse H1 0 were the same, we directly compared the binding of these linker histone isoforms to biochemically defined nucleosomal arrays, and characterized their respective abilities to stabilize condensed chromatin structures in salt (Fig. 1). Binding was assayed by determining the increase in average s 20,w of 208-12 nucleosomal arrays in low salt TEN buffer as a function of the molar linker histone input ratio (r LH ). Under these low salt conditions, nucleosomal arrays and chromatin fibers are unfolded, and there is no contribution to the s 20,w from higher order folding. Instead, the increase in s 20,w results from a combination of the mass of the bound linker histone, and a decreased array frictional coefficient (i.e. shortening of the overall length of the unfolded 12mer array) because of linker histone-dependent formation of an apposed linker DNA stem motif (5). Our previous studies using native chicken erythrocyte H5 showed that binding of approximately one H5 per nucleosome was achieved at r LH Х 1.3, and led to an increase in average s 20,w in low salt from 29 to 36 S (Refs. 6 and 7; also see Table I). Binding of native chicken erythrocyte H5 and recombinant mouse H1 0 to 208-12 nucleosomal arrays is shown in Fig. 1A. Consistent with our previous studies, both linker histone isoforms caused an increase in average s 20,w with increasing r LH , until a narrow plateau region was achieved at r LH ϭ 1.2-1.3. The average s 20,w in this plateau region was 36.8 Ϯ 0.7 S for chicken erythrocyte H5 and 35.9 Ϯ 0.3 S for recombinant H1 0 . We previously have shown that the plateau is narrow because the average s 20,w increases rapidly at higher r LH values because of nonspecific linker histone binding and aggregation (Fig. 1A) (6, 7). The data in Fig.  1A demonstrate that, within experimental error, stoichiometric binding of native chicken erythrocyte H5 and recombinant mouse H1 0 to 208-12 nucleosomal arrays was achieved at the same molar linker histone input ratio, and led to the same increase in sedimentation coefficient under low salt conditions. We next examined the salt-dependent condensation of H5and H1 0 -bound nucleosomal arrays. In vitro condensation consists of intramolecular folding in the range of 0.1-0.5 mM MgCl 2 , followed by cooperative array self-association at higher MgCl 2 concentrations (6, 7). Folding of the biochemically defined model systems was characterized by sedimentation velocity, and the data were analyzed by the method of van Holde and Weischet (37) to yield the integral distribution of sedimentation coefficients, G(s), of the entire sample (6,7,42). The G(s) plots of H5 and H1 0 arrays in TEN Ϯ 0.25 or 0.5 mM MgCl 2 are shown in Fig. 1B. In TEN, ϳ50% of both the H5-and H1 0bound arrays sedimented at 36 -37 S (boundary fraction Ն50%), indicating that this fraction consisted of 208-12 DNA templates stoichiometrically bound with 12 histone octamers/ template and 1 linker histone per nucleosome (see Fig. 1A) (6,7). The remainder of the templates (boundary fraction Յ50%) contained substoichiometric amounts of histone octamers and (or) linker histones, rendering this fraction of the samples uninformative for fiber stability studies (6,7). The G(s) plots of the saturated H5-and H1 0 -bound arrays in the presence of 0.25 and 0.5 mM MgCl 2 were essentially identical (Fig. 1B). Notably, in 0.5 mM MgCl 2 the saturated fraction of both the H5 and H1 0 arrays formed homogeneous populations of extensively condensed ϳ55 S particles, as indicated by the nearly vertical G(s) plots. These data indicate that native chicken erythrocyte H5 and recombinant mouse H1 0 were equally effective at stabilizing the highly folded 55 S secondary chromatin structures formed by 12-mer nucleosomal arrays in 0.5 mM MgCl 2 . We next analyzed the cooperative self-association of nucleosomal arrays bound to chicken H5 and mouse H1 0 . Self-association occurs independently of folding (43,44) and has been proposed to be an in vitro manifestation of long-range fiberfiber interactions that occur in intact chromosomes (8,44). It has been shown previously that nucleosomal arrays bound to chicken H5 self-associate at a significantly lower MgCl 2 concentration than nucleosomal arrays alone, i.e. 0.5 versus 2.0 mM, respectively (6). We found that the salt dependence of self-association of H5-and H1 0 -bound arrays was identical within experimental error (Fig. 1C). Taken together, the data  , ࡗ; H1 0 (TEN), q; H1 0 (0.25 mM MgCl 2 ), ƒ; H1 0 (0.5 mM MgCl 2 ), ‚. C, self-association of H5 (f) and H1 0 (E) bound 208-12 nucleosomal arrays. Self-association was analyzed by differential centrifugation as described under "Experimental Procedures." Each data point represents the mean Ϯ S.D. of two to three experiments.
in Fig. 1, B and C, demonstrate that native chicken erythrocyte H5 and recombinant mouse H1 0 were equally effective at stabilizing salt-dependent secondary and tertiary chromatin structures in vitro.
Binding of H1 0 C-terminal Deletion Mutants to Nucleosomal Arrays-The mouse H1 0 C-terminal domain consists of 97 amino acid residues. To determine the extent to which the CTD functions through nonspecific DNA charge neutralization, four truncation mutants were created that incrementally deleted an approximately equal number of basic amino acid residues from the H1 0 CTD. Specifically, 24, 48, 72, or 97 amino acid residues were deleted beginning from the C-terminal end of H1 0 , yielding mutants termed H1 0 C⌬24, H1 0 C⌬48, H1 0 C⌬72, and H1 0 C⌬97 ( Fig. 2A). After purification, the full-length H1 0 and H1 0 truncation mutants were judged to be homogeneous by SDS-PAGE (Fig. 2B).
Binding of the H1 0 truncation mutants to biochemically defined nucleosomal arrays was monitored by sedimentation velocity as in Fig. 1A, and agarose multigel electrophoresis. The latter technique determines the effective macromolecular radius (R e ), which provides an independent assay for structural changes because of linker histone binding (6, 7, 42). As seen with native H5 and full-length H1 0 (Fig. 1A), binding of both the H1 0 C⌬24 and H1 0 C⌬48 mutants to 12-mer nucleosomal arrays also resulted in a narrow 36 S plateau with increasing r LH (Fig. 3). Under the same conditions we observed a decrease in R e from 27.5 to ϳ22 nm. The R e of 12-mer nucleosomal arrays containing approximately one bound H5 per nucleosome is 21-22 nm (Fig. 1A) (6, 7). Binding of the H1 0 C⌬72 mutant also significantly shifted the array s 20,w and R e with increasing r LH (Fig. 3A), although the extent of change appeared to be slightly less than for full-length H1 0 , H1 0 C⌬24, and H1 0 C⌬48. In marked contrast, the plateau s 20,w observed upon binding of the H1 0 C⌬97 mutant (i.e. the H1 0 N-terminal and globular domains only) was 32 S rather than 36 S (Fig. 3), whereas the plateau R e was only ϳ25 nm rather than ϳ22 nm. In summary, the data in Fig. 3 indicate that removal of the C-terminal most 48 H1 0 amino acid residues had no measurable effect on formation of the 36 S/22 nm conformation, removal of the Cterminal most 72 residues had a small effect at best, whereas removal of the entire 97-residue CTD completely abolished the ability of H1 0 to form the 36 S/22 nm conformation.
Importantly, the r LH required to achieve the plateau s 20,w and R e values increased as the H1 0 CTD was incrementally deleted; the middle of the plateau region for full-length H1 0 , H1 0 C⌬24, H1 0 C⌬48, H1 0 C⌬72, and H1 0 C⌬97 was ϳ1.3, 1.7, 1.9, 2.2, and 2.2, respectively. These results suggest either that deletion of the CTD decreases the relative binding affinity of linker histones for nucleosomal arrays, or that there are substantive differences in linker histone stoichiometry at the respective plateau regions. Because the o term obtained from agarose multigel electrophoresis is directly proportional to macromolecular surface charge density, linker histone stoichiometries can be estimated with reasonable accuracy using the measured o values, and the total macromolecular charge of the respective linker histones and parent nucleosomal arrays (6,7,42). The stoichiometries calculated for the full-length and mutant arrays at their respective plateau regions are shown in Table I. Whereas there is some scatter in the data, there is no evidence that full-length H1 0 or the H1 0 mutant CTD have a markedly greater stoichiometry at their plateau regions compared with native H5. The H1 0 C⌬97 mutant had the greatest stoichiometry (ϳ1.7 linker histones per nucleosome), but nevertheless, was unable to direct formation of the 36 S/22 nm Recombinant full-length and mutant H1 0 s were purified as described under "Experimental Procedures." One microgram of each protein was subjected to SDS-PAGE, and the gel was visualized after staining with Coomassie Blue .   FIG. 3. Binding of H1 0 CTD mutants to nucleosomal arrays. 208-12 nucleosomal arrays were incubated with increasing linker histone to nucleosome molar ratio (r LH ), and the average s 20,w was determined by analytical ultracentrifugation as described under "Experimental Procedures." Each data point represents the mean Ϯ S.D. of two to three experiments. The inset shows the R e of the same samples as a function of increasing r LH , as determined by agarose multigel electrophoresis. Symbols: H1 0 C⌬24, f; H1 0 C⌬48, E; H1 0 C⌬72, OE; H1 0 C⌬97, ƒ.
conformation. We therefore conclude that the affinity of linker histones for nucleosomal arrays is dictated in part by the CTD.
Salt-dependent Folding-Sedimentation velocity experiments in salt were used to assay the extent to which trunca-tions in the CTD influenced the ability of H1 0 to stabilize folded secondary chromatin structures (6, 7). The G(s) plot in TEN of the samples used for these experiments is shown in Fig. 4A. Folding subsequently was determined in TEN containing 0.25 or 0.5 mM MgCl 2 (Fig. 4, B and C). The lower MgCl 2 concentration induces intermediate extents of folding, whereas the higher concentration stabilizes the saturated fraction of H1 0bound 208-12 nucleosomal arrays in the maximally folded 55 S conformation (Fig. 1B) (6). Assuming a charge neutralizationbased mechanism of CTD function (2, 3, 12), our expectation was that incremental deletions in the CTD would cause an incrementally reduced extent of folding at both MgCl 2 concentrations. This would be indicated by successively left-shifted G(s) plots relative to those of H1 0 -bound arrays. However, the G(s) plots of saturated full-length H1 0 and H1 0 C⌬24-bound arrays were identical in TEN buffer containing 0.25 (Fig. 4B) and 0.5 mM MgCl 2 (Fig. 4C). Furthermore, whereas the extent of folding of H1 0 C⌬48-bound arrays was significantly reduced in both salt concentrations, the G(s) plots of the H1 0 C⌬72-and H1 0 C⌬48-bound arrays were superimposed. Finally, the G(s) plots of H1 0 C⌬97-bound arrays in 0.25 and 0.5 mM MgCl 2 were essentially identical to those of the parent 208-12 nucleosomal arrays, indicating complete loss of linker histone-dependent stabilization of salt-dependent chromatin folding. The inability of the N terminus and globular domain alone to stabilize folded chromatin fibers has been observed previously (10,11). The data in Fig. 4 demonstrate that the ability of the H1 0 CTD to stabilize folded secondary chromatin structures is not spread evenly throughout the entire domain. Instead, the sequences that mediate folding are localized to two discrete 24-amino acid stretches within the 97-amino acid domain (see Fig. 6).
Salt-dependent Self-association-The self-association of 208-12 nucleosomal arrays bound to full-length and mutant H1 0 s is shown in Fig. 5. As with folding, the ability of the H1 0 FIG. 4. Salt-dependent formation of folded, secondary chromatin structure. 208-12 nucleosomal arrays containing bound H1 0 proteins were analyzed by sedimentation velocity in TEN alone (A) or TEN containing 0.25 (B) or 0.5 mM (C) MgCl 2 . Shown are the G(s) plots obtained from the sedimentation velocity data. Symbols in panels A-C: nucleosomal arrays, q; full-length H1 0 , dashed line; H1 0 C⌬24, f; H1 0 C⌬48, E; H1 0 C⌬72, OE; H1 0 C⌬97, ƒ. The r LH of the samples used in these experiments was the same as indicated in Table I. FIG. 5. Salt-dependent formation of oligomeric, tertiary chromatin structure. 208-12 nucleosomal arrays containing bound H1 0 proteins were analyzed by differential centrifugation as described under "Experimental Procedures." Each data point represents the mean Ϯ S.D. of two to three experiments. Symbols: nucleosomal arrays (NA), q; full-length H1 0 , line; H1 0 C⌬24, f; H1 0 C⌬48, E; H1 0 C⌬72, OE; H1 0 C⌬97, ƒ. CTD to facilitate self-association was not spread evenly throughout the domain. However, the CTD regions that influence chromatin fiber folding and self-association were distinct. There was no appreciable difference in the salt dependence of self-association of arrays bound to full-length H1 0 , H1 0 C⌬24, and H1 0 C⌬48. The H1 0 C⌬72-bound arrays did require a slightly greater amount of MgCl 2 to achieve 50% self-association, although whether this small difference is significant is difficult to discern because of the highly cooperative nature of the self-association transition (Fig. 5). In contrast, 50% selfassociation of H1 0 C⌬97-bound arrays occurred at 2.5 mM MgCl 2 , which is an intermediate concentration relative to that observed for arrays bound to full-length H1 0 (1.5 mM MgCl 2 ), and nucleosomal arrays alone (3.3 mM MgCl 2 ) (Fig. 5). Thus, the ability of H1 0 to promote self-association resides in both the globular domain and the first 25 amino acids C-terminal to the globular domain. In contrast, the 72 most H1 0 C-terminal residues do not contribute to this process. DISCUSSION Our results have provided new insight into the molecular mechanism of action of the linker histone CTD in chromatin. Key findings are summarized in Fig. 6. Historically, there is a deeply ingrained perception that the CTD behaves as an unstructured polycationic domain that binds linker DNA and neutralizes negative charge, thereby permitting close approach of neighboring nucleosomes and stabilizing condensed chromatin structures (2,11,12,45). However, evidence in favor of this model is largely theoretical (12,45). By constructing specific deletion mutants of the H1 0 CTD and assembling the mutants into biochemically defined model systems, we have performed the first direct experimental test of the charge neutralization hypothesis in a bona fide chromatin environment. A charge neutralization-based mechanism predicts that the functions mediated by the H1 0 CTD would be equally distributed throughout the entire domain, as are the Lys/Arg residues themselves. Such a mechanism would manifest experimentally as an incremental loss of function with incremental deletion of C-terminal amino acids. However, we did not obtain this result for any of the in vitro functions mediated by the CTD, i.e. alteration of linker DNA structure and chromatin fiber conden-sation. Rather, we found that these functions were clustered into two discrete subdomains (Fig. 6). We also observed that the in vitro functions mediated by the C-terminal domain are not dependent on the Lys:Arg ratio, as has been proposed previously (2). Nucleosomal arrays containing chicken H5 and mouse H1 0 behaved identically in our folding and self-association assays (Fig. 1), despite the fact that H5 and H1 0 have a substantially different Lys:Arg composition (13 Arg/32 Lys and 2 Arg/40 Lys, respectively).
We did find, however, that binding of H1 0 to nucleosomal arrays appeared to have an electrostatic component mediated by the CTD. As the CTD was progressively deleted, the amount of H1 0 needed to achieve stoichiometric binding increased in nearly equivalent increments (Fig. 3). Taken together, our results suggest that the CTD binds to linker DNA through a charge neutralization-based mechanism, but once bound, specific CTD subdomains act through alternative mechanisms while functioning in chromatin. Inherent in this hypothesis is that binding of the CTD to linker DNA induces formation of specific, localized protein structures within the CTD that are responsible for mediating linker DNA stem motif formation, stabilizing locally folded chromatin structures, and facilitating array self-association.
We hypothesize that the molecular basis of linker histone CTD function originates from its unique protein chemistry. The CTD sequences of linker histone isoforms are not conserved, yet these domains have a remarkably similar amino acid composition (Table II). Furthermore, the amino acid composition is distinctive, and has been found in many other proteins (13)(14)(15)(16). The ϳ100-amino acid residue CTD of somatic linker histone isoforms have a preponderance of Lys, Ala, and Pro, lesser amounts of Val, Thr, Ser, and Gly, and are almost completely devoid of the other 12 common amino acids (Table II). This composition is characteristic of protein regions that possess what has been termed intrinsic disorder (13)(14)(15)(16). Importantly, intrinsic disorder does not automatically equate to unstructured polypeptide coils, but is thought to often reflect stretches of polypeptides that have molten globule-like properties in their native state (13)(14)(15)(16). One of the most common characteristics of intrinsically disordered regions is that they become The total number of each amino acid residue in the CTDs of chicken linker histone H5 (cH5), mouse H1 0 (mH1 0 ), human H1 0 (hH1 0 ), and somatic human linker histones (46) is shown. The amino acids that promote intrinsic disorder, promote order, and are neutral (13)(14)(15)(16) are indicated as D, O, and N, respectively. structured upon binding to their macromolecular partners (13)(14)(15)(16). Thus, the properties associated with intrinsic disorder are consistent with our empirical evidence suggesting that regions of the CTD become structured upon binding to linker DNA, and that these subdomains subsequently mediate alteration of linker DNA structure and chromatin condensation. A primary functional advantage of intrinsic disorder stems from its conformational malleability (13)(14)(15)(16), which in principle allows a single protein domain to mediate several different functions by participating in different specific macromolecular interactions. Ultimately, our results suggest that a closer examination of the role of intrinsic disorder and interaction-induced secondary structure is warranted, and is likely to reveal important relationships between the structure and function of the linker histone CTD.
From previous results, one possible explanation for the clustering of function (Fig. 6) is that the subdomains identified in our studies contain S/TPKK motifs. These sequences form ␤Ϫturns that bind the minor groove of DNA (22,24), and are known to mediate CTD-dependent condensation of naked DNA (25)(26)(27). The CTD of both chicken H5 and mouse H1 0 contains canonical S/TPKK motifs. Interestingly, the two H1 0 S/TPKK sequences are found in each of the C-terminal subdomains (residues 97-121 and 146 -169, respectively) needed to stabilize extensively folded secondary chromatin structures (Fig. 6). Whereas it is attractive to propose a functional role for the S/TPKK motifs, we note that chicken H5 has three separate S/TPKK sequences, and yet the H5 CTD performs identical functions as the H1 0 CTD in vitro. Thus, the involvement of S/TPKK sequences in CTD function in intact chromatin remains to be determined.
In summary, the studies reported here have led to new insights into linker histone CTD function in chromatin, and have focused attention on the intrinsic disorder of this domain. They have also laid the foundation for future studies of biochemically defined chromatin model systems assembled from recombinant core and linker histone mutants and isoforms. Through such experiments, it will be possible to dissect the complex macromolecular events involved in chromatin condensation.