HMG-D and Histone H1 Interplay during Chromatin Assembly and Early Embryogenesis*

HMG-D is an abundant chromosomal protein associated with condensed chromatin during the first nuclear cleavage cycles of the developing Drosophila embryo. We previously suggested that HMG-D might substitute for the linker histone H1 in the preblastoderm embryo and that this substitution might result in the characteristic less compacted chromatin. We have now studied the association of HMG-D with chromatin using a cell-free system for chromatin reconstitution derived from Drosophila embryos. Association of HMG-D with chromatin, like that of histone H1, increases the nucleosome spacing indicative of binding to the linker DNA between nucleosomes. HMG-D interacts with DNA during the early phases of nucleosome assembly but is gradually displaced as chromatin matures. By contrast, purified chromatin can be loaded with stoichiometric amounts of HMG-D, and this can be displaced upon addition of histone H1. A direct physical interaction between HMG-D and histone H1 was observed in a Far Western analysis. The competitive nature of this interaction is reminis-cent

The vertebrate high mobility group proteins of the HMGB class (formerly termed HMG1/2) (1) are relatively abundant nuclear DNA-binding proteins that bend DNA substantially and appear to act primarily as architectural facilitators in the assembly of nucleoprotein complexes (recently reviewed in Refs. [2][3][4]. There is direct evidence that these proteins facilitate the binding of transcription factors to their cognate sites both in vitro and in vivo, but any involvement in the assembly of histone-DNA complexes is less well documented. Circumstantial evidence indicates that certain functions of the vertebrate HMGB proteins (5-7) may parallel those of the linker histone H1 (8). Both have been shown to bind to linker DNA sequences (7,9,10), and both stabilize and bind to four-way junctions (11)(12)(13)(14). HMG-D is one of two Drosophila proteins closely related to the vertebrate HMGB proteins but in contrast to these proteins contains only a single HMG DNA-binding domain (6,15,16). The HMG domain is followed by a region that has basic sequences similar to the C-terminal domain of histone H1 and a short C-terminal acidic stretch also seen in HMGB proteins. The NMR structure of the HMG domain from HMG-D is very similar to the previously determined structure of the B domain of HMG1, which revealed the characteristic L-shaped fold formed by three ␣-helices (17)(18)(19)(20)(21).
High mobility group proteins of the HMGB family bind DNA with little sequence specificity but recognize structural features in DNA (22)(23)(24), including cruciforms, kinks, DNA bulges, and bends (22,23,(25)(26)(27)(28). DNA footprint analysis and mutagenesis experiments suggest that HMG domain proteins bind in the minor groove (29). In addition, both crystallization and NMR studies show that HMG-D bending is achieved by the intercalation of hydrophobic residues at two different base steps (21,30).
The addition of histone H1 protein to an extended array of nucleosome core particles promotes its folding into a fiber with an approximate width of 30 nm (31)(32)(33). During early Drosophila development, histone H1 is not detectable (34 -36) until nuclear cycles 7/8 (37). Therefore, H1 is not involved in chromatin condensation in these earliest phases of Drosophila development characterized by rapid condensation-decondensation cycles. We have previously suggested that HMG-D could function as a linker protein in the absence of histone H1 in Drosophila (37). In support of this hypothesis, Wolffe and colleagues (10,38) have postulated a similar role for the Xenopus HMG B1 protein. Thus, as the maternal pool of HMG-D is titrated by the rapid replication and concomitant chromatin assembly and as histone H1 is synthesized, HMG-D would functionally be superseded by histone H1 around mid-blastula transition (37).
To test the above model and to examine the interplay between HMG-D and H1 in the process of chromatin assembly, we have used an in vitro chromatin assembly system derived from preblastoderm Drosophila embryos. We show that, in a similar manner to histone H1, incorporation of purified HMG-D into chromatin alters the nucleosome repeat length (NRL), 1 indicative of interaction with the nucleosomal linker DNA. Using bead-immobilized DNA as a substrate for HMG-D binding, we further show that H1 is able to displace HMG-D from chromatin. Finally we show, by far Western analysis, a strong physical direct interaction between H1 and HMG-D.

MATERIALS AND METHODS
Chromatin Assembly and Micrococcal Nuclease Analysis-Chromatin assembly using Drosophila embryo extracts and micrococcal nucle-ase I digests were performed as described (70) with the following modifications. A standard assembly reaction (133 l) contained 40 l of embryo extract (conductivity equals 80 mM KCl), 80 l of EX buffer (10 mM Hepes-KOH, pH 7.6, 1.5 mM MgCl 2 , 0.5 mM EGTA, 10 mM ␤-glycerophosphate), 13.3 l of an energy-regenerating-system (300 mM creatine phosphate, 10 mg creatine kinase/ml, 30 mM MgCl 2 , 10 mM dithiothreitol), and 650 ng of plasmid DNA. Chromatin was assembled for 5 h at 26°C on DNA coupled to paramagnetic beads (Dynal) (71) under identical conditions, but all volumes were scaled proportionately as needed. For the experiment shown in Fig. 4, 3 g of immobilized DNA was assembled into chromatin and isolated from the reaction mixture by exposure to a magnetic field for 20 s. Chromatin beads were immediately resuspended in 100 l of EX buffer containing 80 mM KCl (EX80). 25 l of the suspension was used in each subsequent reaction. Histone H1 (ϳ1 g) or HMG-D (ϳ500 ng) were added to the beads in EX80 (total volume, 50 l) and incubated at 26°C for 30 min. After incubation the chromatin was reisolated and washed successively with 2 ϫ 50 l of EX buffers containing increasing concentration of KCl, typically, EX80, EX300, and EX500. Eluted proteins were precipitated by addition of 8 l of StrataClean resin (Stratagene). After centrifugation (12,000 ϫ g, 2 min), the proteins in the pellet and those remaining on the paramagnetic beads were analyzed by SDS-PAGE.
Coupling of Linear DNA Fragments to Magnetic Beads-Plasmid XX3.2 (71) was digested in the polylinker sequences with EcoRI and ClaI to linearize the DNA. The DNA was blunt ended in the presence of 50 M Biotin-14-dATP and 200 M each of ␣-thio-dTTP, ␣-thio-dCTP, and ␣-thio-dGTP with Klenow enzyme. Unincorporated nucleotides were removed by a gel filtration. The reaction selectively incorporates two biotinylated dATP molecules at one end of the DNA fragment, whereas ␣-thio-dCTP and ␣-thio-dGTP are added at the other end. Subsequently the DNA molecules were attached to paramagnetic beads coated with Strepavidin via the biotin moiety as described previously (40). 1 g of 6.2-kilobase DNA was coupled to 25 l of bead suspension. We routinely coupled 40 g of DNA as follows. The magnetic bead suspension (800 l) was washed in 800 l of phosphate-buffered saline containing 0.1% bovine serum albumin and 0.05% Nonidet P-40, twice in 800 l of 2ϫ wash buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 2 M NaCl), and resuspended in 660 l of 4ϫ wash buffer, 600 l of H 2 0. 40 g of linearized DNA in 60 l was added to the bead suspension, and the tube was rolled overnight at 27°C.
Purification of HMG-D-Expression of HMG-D in Escherichia coli and its purification have been described previously (20). The DNA sequence encoding the full-length protein was cloned in pET24a expression vector between the NdeI and XhoI sites. Expression of HMG-D from this clone gives rise to a mixture of truncated forms: HMG-D 109 , HMG-D 108 , HMG-D 107 , HMG-D 106 , HMG-D 105 , and HMG-D 104 (verified by mass spectrometry). To improve the yield of the full-length protein, we replaced the GAG codon by GAA codon for the Glu residues in positions 101, 102, 110, and 112 and also the GAT codon by GAC for the Asp residues in positions 108 and 109. BL21(DE3) cells transformed by the expression vectors were grown at 37°C until A 600 ϭ 0.8 and induced with isopropyl-1-thio-␤-D-galactopyranoside (250 g/ml), and growth was continued for 4 h at 25°C to overproduce HMG-D 112 . Plasmids expressing truncated HMG-D proteins HMG-D 83 and HMG-D 100 (Nterminal 83 or 100 amino acid residues) were generated by polymerase chain reaction using primer pairs, SnfdNcoI/SnHMG-D83, and snfd-NcoI/SnHMG-D100. SnHMG-D83 is 5Ј-GGAATTCAAGCTTACTATC-CACCACCGTTGGCAGCACTG-3. SnHMG-D100 is 5Ј-GGAATTCAAG-CTTACTACTTCTTGCTCTTCTTCGCCACC-3Ј. SnfdNcoI is 5Ј-GGGA-ATCAACCATGGCTTCTGATAAGCCAAAACGCCCACTCTCC-3Ј. These fragments were subsequently cloned into pKK233-2 at the NcoI-HindIII sites.
HMG-D was purified from preblastoderm embryo extract (36). The extract was heat denatured by a 10-min incubation at 75°C. Precipitated proteins were removed by centrifugation, and proteins in the supernatant were further fractionated by selective ammonium sulfate precipitation. The precipitate from a 70 -95% ammonium sulfate cut contains the majority of HMG-D. The pellet was resuspended in 2 ml of 50 mM Hepes-KOH, pH 7.6, and 50 mM KCl (containing a mixture of protease inhibitors) and dialyzed against this same buffer. The dialysate was further fractionated on a fast protein liquid chromatography Mono S column. HMG-D was loaded onto the column in buffer containing 50 mM KCl and eluted with a gradient of NaCl. It elutes at 0.3-0.35 M NaCl. Histone H1 was isolated from 0 -16 h Drosophila embryos as described previously (40).
Far Western Analysis-SDS-PAGE and electrotransfer of proteins onto nitrocellulose membranes were performed using standard conditions (72). The membrane was then subjected to denaturation in 6 M guanidine HCl in HNT (25 mM Hepes, pH 7.6, 1 mM EDTA, 150 mM NaCl, 1 mM dithiothreitol, and 0.05% Tween 20) at 4°C for 30 min. The proteins were refolded by stepwise incubation in HNT buffer containing successively decreasing concentration of guanidine HCl until 0.01 M. Each wash was for 15 min at room temperature. The membrane was then incubated in HNT containing 5% milk powder for 1 h. After washing briefly in HNT (50 mM NaCl, 1% milk powder), the filter was incubated for 2 h at room temperature in the same buffer containing radiolabeled HMG-D (ϳ10 6 cpm/ml incubation buffer). The filter was washed in HNT (50 mM NaCl) for 5 min, washed in HNT (100 mM NaCl) for 30 min, and exposed to film.
Phosphorylation of HMG-D and Histone H1-Both histone H1 and HMG-D contain serine/threonine residues that can be phosphorylated with protein kinase C. The enzyme requires a lipid mixture (Life Technologies, Inc.) for efficient incorporation of phosphate moiety. HMG-D (1 g) was incubated in 50 l of 20 mM Tris-HCl, pH 7.5, 20 mM MgCl 2 , 1 mM CaCl 2 , 1 mM dithiothreitol, 0.2% bovine serum albumin, 10 Ci of [␥-32 P]ATP, 5 l of lipids, and 0.1 milliunits of protein kinase C (Roche Molecular Biochemicals). After 30 min at 37°C, unincorporated label was removed by gel filtration (Sephacryl-200 HR, Amersham Pharmacia Biotech). The protein was used immediately or stored at Ϫ20°C in 50% glycerol. Approximately 25 ng of labeled HMG-D or H1 were added to 5 g of cold protein for the competition experiments.

Association of HMG-D with DNA during Chromatin
Assembly-To analyze the association of HMG-D with chromatin, we made use of a homologous cell-free system for chromatin reconstitution under physiological conditions in vitro (36). The core of this system is an extract from preblastoderm Drosophila embryos. Incubation of plasmid DNA in this extract in the presence of ATP and an energy-regenerating system results in the formation of nucleosomal arrays with regular spacing. These extracts are deficient in histone H1 but contain large amounts of endogenous HMG-D (37). Chromatin was reconstituted according to standard procedure (39) for 5 h and then isolated by gel filtration on a S-400 spin column. Chromatin proteins were resolved by SDS-PAGE and associated HMG-D was detected by Western blotting. Although stoichiometric amounts of the core histones were detected by Coomassie blue and silver staining on parallel gels, we were unable to detect HMG-D under these conditions (data not shown).
To exclude the possibility that HMG-D was lost during gel filtration, we repeated the experiment using linear DNA coupled to paramagnetic beads ("immobilized DNA") (40). Immobilized chromatin can be separated from the extract within seconds in a magnetic field and subjected to various extractions. Immobilized chromatin was washed sequentially with buffer containing 80 and 500 mM KCl, and the eluted proteins as well as those remaining on the beads were separated by SDS-PAGE and HMG-D detected by Western analysis (Fig.  1a). Although HMG-D in the 80 mM KCl eluted fraction represents protein sticking nonspecifically to the reaction tube and paramagnetic beads (data not shown), we consistently observed low but distinct amounts of chromatin-associated HMG-D in the protein fraction eluted with 500 mM salt wash (lane 2). We estimate that there is ϳ1 molecule of HMG-D/20 -25 nucleosomes. Relative to the core histones and histone H1, this level of HMG-D appears to be similar to what has previously been reported for other HMGB proteins (41,42). Nuclei isolated from 0 -4-h-old embryos contain substantially greater amounts of HMG-D than the 1 molecule of HMG-D/20 -25 nucleosomes (Fig. 1a, lane 4), as has been previously shown (37). We conclude from these experiments that chromatin assembled using the embryonic extract contains only little HMG-D.
To determine whether the failure to detect stoichiometric amounts of HMG-D in the assembled chromatin was due to modification of the endogenous protein (see below), we added recombinant HMG-D to the chromatin assembly and followed its association with immobilized DNA during a time course of nucleosome assembly. The samples were removed at 15, 45, 120, and 300 min, and chromatin was purified magnetically, subjected to diagnostic salt washes, and finally analyzed by Western blotting as before (Fig. 1b). Recombinant HMG-D was able to efficiently bind to naked immobilized DNA in the absence of embryo extract and was eluted entirely by 500 mM KCl ( Fig. 1b and c, lanes 1-3). During chromatin assembly, we observed the presence of substantial amounts of both exogenous and endogenous HMG-D at early times (Fig. 1b, lanes 4 -9). However, as chromatin reconstitution proceeded, the amounts of associated HMG-D decreased correspondingly; after 120 min both endogenous and exogenous HMG-D were less than 25% of the initial values. By 300 min the amounts of recombinant and endogenous HMG-D were, again, low but detectable. The antibody also cross-reacted with two additional, unknown proteins migrating at ϳ25 and ϳ32 kDa eluting at higher salt, which, like HMG-D, were abundant during the early phases of chromatin assembly but were substantially reduced or undetectable at 300 min. The exogenous HMG-D is more tightly associated with chromatin. We conclude that both the endogenous and exogenous HMG-D were associated with DNA, but the bulk of the proteins were displaced gradually as chromatin matures. A situation similar to this has been observed in vivo; embryos undergoing rapid nuclear division in the syncytial blastoderm embryo contained much more HMG-D than embryos in later stages of development (37).
HMG-D, Like Histone H1, Alters the Nucleosomal Repeat Length-A number of observations suggest that HMGB1 can act as a linker protein analogous to H1 (8,10,38). The incorporation of histone H1 into reconstituted chromatin can be visualized conveniently by the associated increase in the NRL (defining the nucleosome spacing) (36,43). So far histone H1 is the only protein known to increase the NRL, presumably because of its interactions with the internucleosomal linker DNA. To test whether HMG-D would also affect the NRL as expected for a linker protein, we analyzed the NRL of HMG-D containing chromatin using the classical micrococcal nuclease assay. Blank and Becker (43) have previously shown that the NRL of reconstituted chromatin is strongly affected by the cation concentrations during reconstitution. At low salt the NRL is 155-165 bp, whereas moderate increases in cation concentrations result in NRLs of 185 bp and above. The addition of recombinant HMG-D to the chromatin assembly reaction at low salt conditions results in an increase in NRL from 160 -165 to 180 -185 bp, but further addition of protein does not result in further increases (Fig. 2a). By contrast, if histone H1 is incorporated into chromatin under similar conditions, the NRL may be increased to well beyond 200 bp (43). We suggest that the  [13][14][15]. To each assembly reaction 500 g of purified HMG-D expressed in E. coli was added. After assembly for the indicated time the chromatin was purified by exposure to a magnetic field and then washed sequentially with EX80 (lanes 1, 4, 7, 10, and 13) and EX500 (lanes 2, 5, 8, 11, and 14). Lanes 3,6,9,12, and 15 contain the magnetic beads after the salt washes. Half of each eluted protein sample and that remaining on the beads was separated on a polyacrylamide gel and transferred to a nitrocellulose membrane. HMG-D was detected by Western analysis. The faster migrating band in the doublet is the exogenous E. coli expressed HMG-D, and the slower migrating band is the Drosophila HMG-D (c.f. Fig. 2b). The difference in mobility between the exogenous, E. coli expressed HMG-D and the slower migrating Drosophila HMG-D (c.f. Fig. 2b) is due to modifications that are present on the endogenous HMG-D protein. c, the remaining 50% of proteins in the eluated fractions and on the beads from experiment described in b were separated on a second polyacrylamide gel. The proteins were detected using a silver stain. The core histones are observed as four distinct bands of equal intensity in the bead fraction that represents chromatin. change in NRL results from the binding of HMG-D to the linker sequence, in analogy to but clearly distinct from histone H1. We also tested truncated HMG-D proteins and HMG-Z, a zygotic protein closely related to HMG-D (44) in this assay. Fulllength HMG-Z functioned like HMG-D, with similar changes in NRL observed upon addition of similar protein amounts (Fig.  2a). A change in NRL was also observed with a C-terminal truncation of HMG-D lacking the acidic residues (HMG-D 100 ). HMG-D 100 contained the DNA binding motif and an alaninelysine-rich region homologous to a region in histone H1. HMG-D 74 , lacking the acidic tail and the alanine-lysine rich region, failed to alter the spacing (data not shown). This truncated protein still contained the DNA-binding motif but bound DNA much less efficiently than the full-length protein (23,45,46). Together these data suggest that the acidic residues are not required for altering the spacing but that the alanine-lysine rich region is involved in binding to chromatin and in the process alters the NRL.
In the absence of exogenous HMG-D the NRL is 160 -165 bp, yet the extract contains high levels of endogenous HMG-D. Why, then, does the endogenous protein not cause a change in the NRL? To address this issue we purified HMG-D from em-FIG. 2. HMG-D binds as a linker protein and changes the NRL on addition to the assembly reaction. a, chromatin was assembled on plasmid DNA by incubation with an early embryonic extract in the presence of an ATP regenerating system. Lanes 1, 5, and 9 contain no E. coli expressed HMG proteins. Lanes 2-4 contain 100 ng, 500 ng, and 1 g, respectively, of HMG-D 112 added to the assembly reaction. Lanes 6 -8 contain 100 ng, 500 ng, and 1 g, respectively, of HMG-Z 111 added to the assembly reaction. Lane 10 contains 1 g of HMG-D 100 added to the assembly reaction. After 5 h, each assembly reaction was divided into to two, and the DNA was subjected to a digestion with micrococcal nuclease I for either  1 with lane 3). c, electron spray mass spectrometry profile of endogenous HMG-D purified from 0 -16-h-old Drosophila embryos. The predicted molecular mass of unmodified HMG-D is 12,297.0. The endogenous protein shows two molecular peaks corresponding to molecular mass of 12,324.0 and 12,404.6, which represent acetylated HMG-D (peak 1) and acetylated ϩ phosphorylated HMG-D (peak 2). bryo extracts (Fig. 2b) and added it into a chromatin assembly reaction. However, we did not observe any change in the NRL (data not shown). HMG-D isolated from early embryo extracts ran differently from the recombinant protein (Fig. 2b), suggesting that the endogenous protein is extensively modified, as also inferred by Renner et al. (47). Different fractions from the Mono S column were heterogeneous, suggesting the presence of isoforms (Fig. 2b, lanes 1 and 2). By electron spray mass spectrometry we have shown that all the endogenous protein was modified by acetylation and that a large proportion of the endogenous protein was also phosphorylated at a single site (Fig. 2c). The affinity of the endogenous protein for DNA is similar to that of the isolated HMG domain (data not shown), suggesting that the modifications may antagonize the function of the basic domain required for efficient nucleosome spacing.
HMG-D Binds to Preassembled Chromatin-Although exogenous HMG-D increased the NRL, arguing for a stoichiometric interaction with nucleosomes, only relatively low protein levels were found to associate in a stable manner with purified chromatin. The association with chromatin may, therefore, be transient and dynamic in the presence of the crude embryonic extract. We next tested whether HMG-D was able to bind chromatin in the absence of the assembly system. Chromatin was reconstituted for 5 h on immobilized DNA and isolated on a magnet. Under these conditions little to no endogenous HMG-D was associated with the chromatin. We then incubated the chromatin with recombinant HMG-D. To facilitate the analysis of the chromatin-associated HMG-D, we "spiked" the HMG-D with a small amount of marked protein that had been radiolabeled with protein kinase C and [␥-32 P]ATP. Half of this reaction was analyzed with a micrococcal nuclease digest and yielded a regular nucleosomal ladder having the shorter spacing characteristic of linear chromatin (48). The other half was subjected to sequential washes with increasing salt concentrations that allowed us to test the affinity of bound HMG-D to chromatin. Fig. 3b shows that HMG-D associated with chromatin could be eluted with 80 and 160 mM KCl in the wash buffer. Under these conditions relatively high levels of HMG-D can be incorporated on to the chromatin, such that it can be detected by silver staining of the protein gels (Fig. 3c). We estimate that ϳ0.5-1 molecules of HMG-D/histone octamer can associate with bead chromatin.
In a manner analogous to HMG-D, we also were able to incorporate histone H1 into preassembled chromatin (Fig. 3d). Chromatin was assembled on immobilized DNA, isolated, and incubated with histone H1 containing a tracer of radiolabeled linker histone. H1 protein associated with immobilized chromatin was eluted with salt washes and analyzed by SDS-PAG electrophoresis. The presence of H1 in the various fractions was detected by autoradiography (Fig. 3d) and by silver staining (Fig. 3e). Majority of the H1 was observed in the 500 mM salt wash fraction as well as a small fraction that remained associated with the magnetic bead fraction containing the core histones. Similar to our observations with HMG-D, near equivalent levels of H1 relative to the core histone could be incorporated into the preassembled bead chromatin.
Competition between HMG-D and H1-Having established conditions that allow the selective incorporation of H1 or HMG-D into immobilized chromatin, we asked whether H1 and HMG-D could displace each other competitively. Bead chromatin was prepared as described under "Materials and Methods" on immobilized DNA, and HMG-D incorporated as described in Fig. 3b. H1 was added subsequently to compete for the binding to the chromatin. To determine the relative levels of H1 and HMG-D present on the chromatin, the bead chromatin was washed with sequential salt concentrations. Chromatin-bound HMG-D was clearly displaced by H1 (Fig. 4, lanes 9 -16) in a concentration-dependent manner. Two types of binding of HMG-D to chromatin were seen (Fig. 4): a low affinity binding disrupted by 80 mM salt concentration and a high affinity binding which was disrupted by a high salt wash. Strikingly, H1 disrupted the higher affinity binding species first, presumably corresponding to the linker-associated protein. At high concentrations of H1 both types of HMG-D binding were disrupted. However, if the reaction contained a high level of HMG-D (ϫ10 relative to the H1 concentration) H1 was unable to displace the HMG-D (lanes 17-20). This experiment clearly demonstrated the competition between H1 and HMG-D in vitro, which has previously been suggested to take place during Drosophila embryonic development based on our cytological observation (37).
A Physical Interaction between H1 and HMG-D-Is the displacement of HMG-D by H1 observed because of a direct physical interaction of H1 with HMG-D, or does it involve other chromatin-associated factors? We tested for a direct physical interaction between the two-linker proteins by a Far Western analysis using radiolabeled HMG-D as the probe (Fig. 5a). We observed a strong signal corresponding to the direct binding of labeled HMG-D to purified histone H1 (lane 5) or H1 in the context of total nuclear protein (lane 2). In addition there was a very weak signal corresponding to binding of HMG-D to histone H2A (lane 2). We repeated this experiment using labeled H1 as the probe (Fig. 5b). H1 strongly interacted with itself and weakly interacted with HMG-D 100 (a truncated HMG-D lacking the C-terminal 12 amino acids), but full-length HMG-D was not detected by H1. Although a definite explanation for this result is still lacking at present, one reason could be the incorrect folding of HMG-D on the nitrocellulose membrane. We also found that H1 interacted with H2A in a mixture of purified core histones or in the context of a crude embryo extract (lanes 2 and 3). DISCUSSION A Role for HMG-D as a Linker Protein-Histone H1 and HMGB1 proteins could influence chromatin structure in a similar manner by binding to linker DNA sequence. Histone H1 associates with linker DNA sequences and organizes nucleosomal arrays into higher order chromatin structures, such as the 30-nm chromatin fiber (31,33). However, little is known about how HMGB1 interacts with the nucleosome and about the consequences in structure and function. H1 and HMGB1 share important features; both protect linker DNA sequences from nuclease digestion (10,14,38,49), and both bind four-way junctions (12,13). Jackson et al. (8) provided evidence for an interaction of HMGB1 with the nucleosome and suggested that it might replace histone H1. Consistent with these observations, we observed during very early stages of Drosophila embryogenesis that histone H1 is absent, but the high mobility group protein D (HMG-D) is present in vast excess. Based on the similarities between HMG-D and H1, a role for HMG-D as a linker protein compatible with and perhaps required for the fast condensation-decondensation cycles associated with the very rapid nuclear division cycles found in preblastoderm embryos was suggested (37). An analogous role has now been proposed for the Xenopus HMGB1 and B4 proteins; both proteins have been demonstrated to bind di-nucleosomal DNA (10,38).
The fact that recombinant HMG-D increases the NRL in a cell-free chromatin assembly system strongly supports this hypothesis. The NRL is strongly dependent on the ionic environment such that polycations are particularly effective in increasing the average separation between adjacent nucleosomes (43). In accordance with these findings our data impli-cate the polycationic basic region (residues 85-99; net charge, ϩ10) of HMG-D in this function. However, the HMG-D-dependent increase in NRL is mediated both by the full-length protein and by HMG-D 100 . These forms differ substantially in net charge ϩ7 (for HMG-D) and ϩ17 (for HMG-D 100 ), suggesting that the chromatin DNA can compete effectively with the polyanionic acidic tail of HMG-D. Histone H1 and the HMGN1 and HMGN2 proteins (formerly HMG-14 and HMG-17) are the only other proteins reported to cause such a change (50,51), in the case of H1 presumably by binding to the linker DNA. The binding of H1 to the linker sequence appears to differ from that of HMG-D. Increasing concentrations of histone H1 added to the assembly reaction will continue to increase the NRL to well over 220 bp before the regular nucleosomal array is lost. HMG-D, on the other hand, increases the NRL to only ϳ180 bp. This may reflect the stoichiometry of binding to the linker sequence. Ura et al. (10) showed that di-nucleosomal DNA reconstituted by dialysis was able to bind two molecules of H1 but only a single molecule of HMGB1. Although the exact nature of the binding remains unknown, HMG-D binds ϳ14 bp of DNA (23), and consequently 1-2 molecules of HMG-D could potentially occupy the linker space.
Previously, a tight correlation between nucleosome spacing and the folding of the nucleosomal fiber into a 30-nm fiber was observed (43,48), which led to the suggestion that different NRLs would correspond to particular fiber geometries and, therefore, compaction states. Accordingly, increased nucleosome spacing is indicative of more compacted chromatin. The observation that HMG-D does not increase the NRL beyond 185 bp as H1 may indicate that HMG-D-containing chromatin is folded but is less compacted.
Like other HMG domain proteins such as LEF-1 and SRY FIG. 3. a, preparation of chromatin on DNA immobilized on paramagnetic beads. Any linear DNA sequence can be attached to the paramagnetic beads after cleavage with appropriate restriction enzymes and incorporation of biotinylated nucleotides at one end of the fragment. The paramagnetic beads are coated with Strepavidin that has extremely high affinity for biotin, and thus the DNA can be attached to the beads by a short incubation step. The bead DNA is used as the template in the chromatin assembly reaction. The assembled chromatin complex is purified simply by exposure of the assembly reaction to a magnetic field for 20 s. The chromatin is then subjected to various washing conditions to remove components that interact nonspecifically. b, incorporation of HMG-D into preassembled chromatin. Chromatin was prepared on DNA attached to paramagnetic beads, isolated, and incubated with 500 ng of E. coli purified HMG-D, which was spiked with radiolabeled HMG-D. To determine the affinity of HMG-D binding to the chromatin, we subjected the chromatin to sequentially increasing salt concentrations (80, 160, 250, and 500 mM KCl, lanes 2-5, respectively). There is bimodal binding of HMG-D to chromatin: a low affinity binding that is disrupted with 80 mM KCl (lane 2) and a higher affinity binding that is disrupted by 160 mM KCl (lane 3). sup refers to the supernatant fraction removed from the paramagnetic beads after incubation of HMG-D. It represents the fraction of HMG-D that is unincorporated into the bead chromatin. Lane 6 is loaded with the paramagnetic beads after the final 500 mM KCl wash. It contains the core histones and proteins very tightly associated to the chromatin. c, silver-stained polyacrylamide gel showing that similar amounts of HMG-D relative to the core histones can be incorporated into chromatin preassembled onto paramagnetic beads. Lanes 1 and 2, protein fractions eluted with 80 and 500 mM KCl salt washes from bead chromatin after it had been incubated with 1 g of recombinant HMG-D. Lane 3, chromatin proteins remaining associated with the beads after the 500 mM KCl salt wash. The core histone proteins are the prominent proteins that remain on the bead chromatin after the salt washes. d, incorporation of H1 into preassembled chromatin. Bead chromatin was incubated with either 0.5 (lanes 2-5) or 1 (lanes 6 -9) g of histone H1. The chromatin was reisolated and subjected to washes containing 80 mM KCl (lanes 2 and 7) and 500 mM KCl (lanes 4 and 8). Lanes 5 and 9 show the proteins still associated with the bead chromatin after the 500 mM KCl salt wash. The core histones and a residual amount of H1 remain associated with the DNA. Lane 1 shows the input radiolabeled histone H1. e, silver-stained polyacrylamide gel showing that similar amounts of H1 relative to the core histones can be incorporated into chromatin preassembled onto paramagnetic beads. Lane 1, ϳ200 ng of histone H1. Lanes 2-4, protein fractions eluted with 80, 250, and 500 mM KCl salt washes, respectively, from bead chromatin after it had been incubated with 1 g of histone H1. Lane 5, chromatin proteins remaining associated with the beads after the 500 mM KCl salt wash. (52,53), HMG-D can introduce sharp bends or kinks into DNA. Our current estimates of the magnitude of the DNA kinks induced by HMG-D range from 100 -120°for the full-length protein to 60 to Ͼ90°for HMG-D 100 (27). These values are substantially greater than the average curvature of DNA wrapped around the histone octamer (54,55) and indicate that HMG-D bound DNA is not smoothly curved. In the context of linker DNA, such a state would be consistent with both the lack of UV-induced thymine dimer formation in the linker (56) and also, with evidence from electric dichroism studies, that the trajectory of linker DNA differs from that of DNA bound to the core histones (57). Of particular relevance are the observations of Hamiche et al. (58), who showed that, in the presence of histone H1 derivatives containing a major proportion of the basic C-terminal domain, the linker DNA enters and leaves a single chromatosome as a straight rod approximately perpendicular to the superhelical axis. A similar structure was subsequently observed in chromatin fibers (59). This organization implies that the DNA must bend sharply as it enters and leaves the chromatosome. A possible role for HMG-D would be to induce such sharp bends by kinking the DNA and thereby promoting a higher level of chromatin folding.
Jackson et al. (8) provided evidence for an interaction of HMGB1 with the nucleosome and suggested that it might replace histone H1 in the nucleosome. Similarly Carballo et al. (60) and Kohlstaedt and Cole (61) have provided evidence for interactions between histone H1 and HMGB1. Our results are consistent with these observations and those of Wolffe and co-workers (10,62). Firstly, in a Far Western analysis H1 was the predominant protein identified when labeled HMG-D was used as a probe. Secondly, using chromatin assembled on DNA attached to paramagnetic beads and preloaded with HMG-D protein, we showed that HMG-D is displaced upon titration of histone H1. We note that the full-length HMG-D and HMG-D 100 both interact with H1 in a Far Western analysis. The alanine-lysine-rich region (amino acids 84 -100, AKKRAKPA-KKVAKKSKK) is very similar to a region found in histone H1. Our Far Western analysis suggests that this region, or possibly the region immediately preceding glycine-rich linker, is interacting with H1. In HMG-D this sequence contains a serine residue that is phosphorylated by casein kinase II (63).
A Role for HMG-D as a Nucleosome Assembly Factor-Although it is possible to argue for a structural role for HMG-D and HMGB1 in early embryonic chromatin (37,62), our in vitro observations show that in the absence of H1 HMG-D, although initially present at high levels, is displaced to below 1 molecule/ 10 -20 nucleosomes as the reaction proceeds and the chromatin matures. This would argue against a purely structural role for HMG-D and suggest that the protein may fulfill a different FIG. 4. Demonstration that histone H1 can displace HMG-D in a concentration dependent manner. Bead chromatin was prepared as described under "Materials and Methods" on immobilized DNA, and HMG-D incorporated as described in Fig. 3b. H1 was added subsequently to compete for the binding to the chromatin. To determine the relative levels of H1 and HMG-D present on the chromatin, the bead chromatin was washed with the salt concentrations indicated. The concentrations of histone H1 were 0.1 (lanes 5-8), 0.5 (lanes 9 -12), and 2 g (lanes 13-16) with a constant amount of HMG-D (0.5 g) expect in lanes 17-20 where 2 g of HMG-D was incubated with 0.5 g of H1. Note: both H1 and HMG-D were spiked with small amounts of labeled proteins for detection purposes. The HMG-D with the higher affinity for chromatin is displaced first (compare lanes 3, 7, 11, and 15). At high concentrations of H1 low affinity binding of HMG-D is disrupted (com- pare lanes 2, 6, 10, and 14). If high concentrations of HMG-D are incubated relative to H1, H1 is prevented from binding to chromatin (lanes [17][18][19][20]. sup refers to the supernatant fraction removed from the paramagnetic beads after incubation with H1 and HMG-D and represents the input fraction. The lanes labeled Beads are loaded with the paramagnetic beads after the final 500 mM KCl wash. They contain the core histones and proteins very tightly associated to the chromatin. FIG. 5. a, Far Western analysis to demonstrate a physical interaction of HMG-D with histone H1. After SDS-PAGE the proteins were transferred to nitrocellulose and probed with radiolabeled HMG-D. Lane 1, ϳ100 g of 0-90-min Drosophila chromatin assembly extract. Lane 2, ϳ 100 g of 0-16-h Drosophila embryonic nuclei. Lane 3, 2 g of E. coli expressed and purified HMG-D. Lane 4, blank. Lane 5, 2 g of histone H1 purified from 0 -16-h Drosophila embryos. Lane 6, ϳ5 g of purified and partially trypsinized core histone proteins. Lane 7, ϳ5 g of fully trypsinized core histones. Lane 8, ϳ5 g of core histones purified from 0 -16-h Drosophila embryos. Histone H1 is the main band that is detected with HMG-D. On longer exposure a band corresponding to H2A appears. b, Far Western probed with radiolabeled H1 protein. Similar amounts of protein were loaded on the gel as indicated for panel a. Also loaded is ϳ 2 g of HMG-D 100 (lane 6) and HMG-D 74 (lane 8). Histone H1 detects itself (lanes 2 and 5) but also H2A (lanes 2 and 3) and faintly HMG-D 100 . role. One possibility is that HMG-D functions as a chaperone molecule (64,65) and preconfigures the DNA to facilitate the chromatin assembly process. HMG-D could participate to bend the DNA at the exit and entry points to the nucleosome, and this bend is then stabilized by histone H1 to generate a structure similar to those observed by Hamiche et al. (58) and Bednar et al. (59). Under such a scenario, as chromatin assembly proceeds and the core histones are recruited, HMG-D molecules are displaced. The linker sequences would be the only locations where the protein would persist for longer duration. However, this too would be displaced on the addition of other chromatin-associated proteins (transcription factors, assembly factors). Such a mechanism would be very similar to that proposed for the recruitment of transcription factors (4). Similarly the displacement and competition with histone H1 can be envisaged as part of a process in which the DNA is kinked by HMG-D, and then the binding of the linker histone stabilizes this kink.
A Role for HMG-D during Preblastoderm Development-Preblastoderm embryonic chromatin clearly differs profoundly from post-blastoderm chromatin. Early syncytial nuclei are much larger and contain chromatin that is less compacted than later nuclei. In the early embryo HMG-D is highly abundant, although not all molecules are necessarily available for DNA binding. It is deposited in the egg by the mother but thereafter is maintained at an approximately constant level per embryo. Consequently, with each nuclear division the average number of HMG-D molecules per nucleus falls, although during nuclear cycles 7-14 the amount of H1 rapidly increases (37). Only during cycle 7 does the size of the nuclei begin to decrease. By cycles 10 -12 a sufficient amount of histone H1 has accumulated to allow the reorganization of chromatin to a transcriptionally active state. Subsequently, increased zygotic transcription elevates histone H1 levels further. This exponential increase of histone H1 together with the increasing number of nuclei rapidly deplete HMG-D protein to levels that cannot have global effects on chromatin structure. What could be the physiological significance of different linker proteins? As discussed above, HMG-D-or H1-containing chromatin may differ profoundly in the degree or mode of compaction. The looser structure formed in the absence of H1 could facilitate the rapid condensation and decondensation required during the very short early cleavage cycles.
The switch from HMG-D-to H1-containing chromatin correlates with the acquisition of global transcriptional competence. Similar observations have been described in the Xenopus system in which B4, an H1 variant, and HMGB1 disappear during mid-blastula transition (66), again correlating with a change in the accessibility of embryonic chromatin to class III transcriptional machinery (10,(67)(68)(69). The cell-free system employed in this study may facilitate the detailed analysis of this major switch in genome function during embryonic development.