HMGB1 affects the sliding activity of CHRAC and ACF

In order to test for an effect of HMGB1 on ISWI- and ACF-induced nucleosome mobility, we employed an established assay that allows visualization of single nucleosome movement, catalysed by ATP-dependent nucleosome-remodelling factors (Längst et al., 1999; Brehm et al., 2000; Clapier et al., 2001; Eberharter et al., 2001). The wrapping of short DNA fragments around histone octamers frequently leads to multiple translational positions that can be separated electrophoretically. In our case, nucleosomes occupy prominent positions on a 248 bp fragment derived from the rDNA promoter: a series of internal positions, which are not separated by the gel system, are dictated by the anisotropic flexibility of the DNA sequence, and in addition, nucleosomes can abut the fragment ends. Isolated nucleosomes positioned either at the centre or at the periphery of the DNA fragment serve as a substrate for remodelling machines. We previously showed that the remodelling ATPase ISWI is able to move nucleosomes from internal positions to fragment ends (Längst et al., 1999; Längst and Becker, 2001a). Association of ACF1 with ISWI reconstitutes the heteromeric ACF (Ito et al., 1999). The presence of two small histone fold subunits distinguishes CHRAC from ACF (Corona et al., 2000). ACF1 not only increases the efficiency of ISWI-dependent remodelling by an order of magnitude, it also leads to a qualitative change in nucleosome sliding: in contrast to isolated ISWI, ACF and CHRAC can catalyse the movement of nucleosomes from fragment ends to more central positions (Längst et al., 1999; Eberharter et al., 2001). While these simple model reactions on short linear DNA fragments are unlikely to reflect all physiological constraints of nucleosome movements in nuclei, they provide an excellent opportunity to study fundamental parameters that affect nucleosome mobility.

The first evidence for an effect of HMGB1 on nucleosome sliding was obtained in reactions catalysed by native CHRAC (Figure 2A). Stopping the reactions at different time points with competitor DNA, followed by native gel electrophoresis, measured the kinetics of nucleosome sliding as a function of fixed amounts of HMGB1. We found that nucleosomes moved significantly faster in the presence of HMGB1 (Figure 2A, compare lanes 2 and 3 with 6 and 7), but the overall level of nucleosome mobilization at the end of an extended incubation did not change (Figure 2A, compare lane 5 with 9). We next explored whether HMGB1 would also affect the kinetics of nucleosome sliding by the recombinant two-subunit ACF (Eberharter et al., 2001). In the presence of 2 pmol of HMGB1, a clear increase of nucleosome sliding, particularly at the early time points, was detectable (Figure 2B, lanes 2–4 and 8–10). A similar stimulation was observed when the HMGB1 concentrations were increased 5-fold, but raising the concentrations of remodelling factor abolished the effect (data not shown). In order to better document the kinetic effect of HMGB1 on nucleosome sliding, we quantified the degree of nucleosome sliding throughout the sliding reaction and determined the degree of stimulation by HMGB1 for each time point. The graph in Figure 2C displays the average data derived from four independent experiments. It documents a kinetic effect of HMGB1 on nucleosome sliding, which is mainly observable at the early times.

Open in a separate window

Fig. 2. HMGB1 stimulates CHRAC- and ACF-mediated nucleosome remodelling. (A) Nucleosomes positioned at the end of the DNA fragment (lane 1) were incubated with 5 fmol of native CHRAC and ATP, in the absence (lanes 2–5) or presence of 2 pmol of HMGB1 (lanes 6–9). The reactions were incubated at 16°C and stopped at the indicated time points by the addition of competitor DNA. Nucleosome positions were subsequently analysed by electrophoresis. Nucleosomes are schematized by ellipses with protruding DNA. (B) Nucleosome remodelling assay, as described in (A), using 5 fmol of recombinant ACF in the absence (lanes 1–6) or presence of 2 pmol of HMGB1 (lanes 7–12). (C) Quantification of the HMGB1 effect on nucleosome remodelling. The ratio of nucleosomes moved in the presence of HMGB1 versus nucleosomes moved in the absence of HMGB1 was calculated for four independent experiments and displayed in a graph. (D) Nucleosomes positioned at the centre of the DNA fragment were incubated with 30 fmol of ISWI and ATP, in the absence (lanes 1–5) or presence of 2 pmol of HMGB1 (lanes 6–10) and assayed as described in (A).

In order to establish whether ISWI by itself would also profit from the presence of HMGB1, we added the factor to reactions in which ISWI mobilized a centrally positioned histone octamer to slide to a fragment end (Figure 2D). Surprisingly, ≥2 pmol amounts (data not shown) of HMGB1 did not affect the kinetics of ISWI-dependent nucleosome sliding, suggesting a critical contribution of the ACF1 moiety for the functional synergism with HMGB1.

HMGB1 promotes ACF binding to the nucleosome

The stimulatory effect could be due to interactions of HMGB1 with ACF, with the nucleosomal linker DNA or both. We have used mouse HMGB1 in the context of a nucleosome remodelling machinery from Drosophila to minimize the probability of direct interactions, and, accordingly, we were so far unable to detect any significant interactions between HMGB1 and ACF in biochemical pull-down experiments (data not shown). Therefore we next analysed the effect of HMGB1 on the interaction of the remodelling factors with the nucleosomes. Binding of ACF to the nucleosome can be visualized by EMSA as a single prominent band if no competitor DNA is added prior to gel loading (Figure 3A, lanes 4–6). As before, HMGB1 shifts the nucleosome into a position close to, but distinct from, the ACF–nucleosome complex (Figure 3A, lanes 13 and 14). Importantly, the presence of 0.6 pmol of HMGB1 does not shift the nucleosomal band significantly (Figure 3A, lane 10). However, if this amount of HMGB1 was included in a nucleosome-binding reaction with ACF, 4-fold lower concentrations of ACF were required to shift the nucleosome (Figure 3A, compare lanes 1–6 with 15–20). At higher, saturating concentrations of ACF, the degree of nucleosome binding was unaffected by HMGB1, in close correspondence with the nucleosome sliding profiles. Under these conditions, no evidence for the existence of a ternary complex of ACF, HMGB1 and the nucleosome was detected using specific antibodies in ‘supershift’ strategies (data not shown). Thus, HMGB1 may function as a DNA chaperone that facilitates ACF binding to the nucleosome at suboptimal conditions without being a stable component of the complex.

Open in a separate window

Fig. 3. HMGB1 promotes the formation of a ACF–nucleosome complex. (A) Nucleosomes preferentially positioned at the centre of the DNA fragment were incubated with increasing amounts of ACF (lanes 1–7; 15, 30, 60, 120, 240, 480 and 960 fmol), HMGB1 (lanes 9–14; 0.3, 0.6, 1.2, 2.4, 4.8 and 9.6 pmol) and increasing amounts of ACF in the presence of 0.6 pmol HMGB1 (lanes 15–21). Complexes were analysed in EMSAs, as before. Nucleosome–ACF complexes are marked by a triangle and nucleosome–HMGB1 complexes are indicated by a circle. (B) Positioned nucleosomes were incubated with increasing amounts of ISWI (lanes 1–5; 60, 120, 240, 480 and 960 fmol), HMGB1 (lanes 7–12, 0.3, 0.6, 1.2, 2.4, 4.8 and 9.6 pmol) and ISWI with fixed amounts of HMGB1 (lanes 13–17, 0.3 pmol HMGB1; lanes 18–22, 0.6 pmol HMGB1). Nucleosome– ISWI complexes are marked by triangles and the nucleosome– HMGB1 complex is indicated by a circle.

A similar experiment, in which we assayed for an effect of HMGB1 on the ISWI-nucleosome interaction, revealed no stimulation (Figure 3B). We conclude that the potential of HMGB1 to facilitate nucleosome sliding correlates with its ability to lead to enhanced interaction of the remodelling machine with the nucleosomal substrate. The data identify the ACF1 moiety of ACF as crucial for this effect.

HMGB1, a hallmark of dynamic chromatin

HMGB1 diffuses throughout the nucleus with the highest rates, indicating a rather dynamic association with chromatin (Scaffadi et al., 2002). Since HMGB1 binds nucleosomes at sites overlapping those recognized by the linker histone H1, HMGB1-like proteins can compete with histone H1 for nucleosomal-binding sites (Nightingale et al., 1996; Ner et al., 2001). Association of histone H1 prevents spontaneous nucleosome sliding at elevated salt and temperature (Pennings et al., 1994) and inhibits the remodelling action of the SWI/SNF complex on mononucleosome substrates (Hill and Imbalzano, 2000) and nucleosomal arrays (Horn et al., 2002). Whether histone H1 affects nucleosome sliding or other ATP-dependent functions of ISWI has not been determined in a defined system yet (but see Varga-Weisz et al., 1995). Given that the interaction of H1 with linker DNA (Widom, 1998) is likely to interfere with ISWI binding, some kind of inhibition is expected. Therefore, HMGB1 may facilitate nucleosome remodelling in two ways: (i) by ‘unlocking’ the nucleosome through displacing linker histones, and (ii) by stimulating the activity of ACF/CHRAC as shown here and discussed below.

Early experiments indicated that a fraction of chromatin exists containing reduced levels of histone H1 and corresponding enrichment of HMGB1/2 to stoichiometric levels (with respect to nucleosomes) (Jackson et al., 1979). Our findings suggest that chromatin of this type should be characterized by high levels of nucleosome mobility. A further example for chromatin with unusual dynamic properties is found in the preblastula nuclei of Drosophila embryos, which lack histone H1 but contain large amounts of the Drosophila HMGB1-like protein, HMG-D, which also interacts with the nucleosomal linker DNA (Ner et al., 2001). We imagine the resulting chromatin to be extraordinarily dynamic, consistent with the high replication and chromatin assembly rates observed before cellularization (Ner and Travers, 1994).

HMGB1 and nucleosome sliding: mechanistic considerations

Prevalent models assume that the critical step in nucleosome mobilization is the distortion of a DNA segment at the edge of the nucleosome by the remodelling ATPase, and the introduction of a segment of ‘bulged’ DNA on the surface of the histone octamer. Propagation of this ‘bulge’ over the histone surface eventually would lead to displacement of the DNA relative to hallmarks on the octamer surface (Havas et al., 2000; Längst and Becker, 2001b; Narlikar et al., 2001; for review, see Becker and Hörz, 2002). In the context of this model, HMGB1 could, in principle, facilitate at least two steps: the initial distortion of DNA or the propagation of the distortion over the nucleosomal surface (Figure 5). Several arguments lead us to favour the first mechanism. (i) The band-shift data shown in Figure 3 demonstrate an improved binding of ACF to nucleosomal DNA in the presence of HMGB1. The magnitude of this effect correlates with the kinetic acceleration of nucleosome sliding. (ii) Conceivably, the initial distortion of a relatively short segment of DNA must be rate-limiting the overall reaction, while the propagation of the ‘bulge’, once formed, is presumed to require relatively little further energy input (Längst and Becker, 2001a; Becker and Hörz, 2002). The ‘HMG-effect’ is particularly prominent at early times during the course of the experiment, pointing to an effect on initiation. (iii) Consistent with these ideas, we recently observed that single-stranded nicks in DNA close to the entry into the nucleosome—the site of ISWI interaction—facilitated ATP-dependent nucleosome sliding (Längst and Becker, 2001a). We hypothesized that the presence of these nicks might enhance the flexibility of DNA at this site, such that it could be more easily bent or otherwise distorted. (iv) HMGB1 affected nucleosome sliding by ACF and CHRAC, but not by recombinant ISWI alone. Previous results already indicated profound differences in the initiation of nucleosome mobilization by these factors (Eberharter et al., 2001). Apparently, ACF and ISWI approach the nucleosomal substrate somehow differently, since ISWI only triggers the sliding of nucleosomes to fragment ends whereas ACF is able to move nucleosomes from ends to more central positions on DNA. At the same time, ACF mobilizes nucleosomes by an order of magnitude more efficiently than ISWI alone, which may be due to the ability of ACF1 to contribute to substrate recognition.

Open in a separate window

Fig. 5. Model showing the stimulatory effect of HMGB1 on ACF-mediated nucleosome remodelling. The wrapping of DNA (black line) around a histone octamer (grey sphere) is shown sideways. See text for details.

In summary, we hypothesize that HMGB1 acts as a DNA chaperone (Crothers, 1993; Travers et al., 1994) to flexibilize or distort DNA (or to stabilize an intrinsic bend) at the nucleosomal edge (Figure 5). This flexibilization and/or distortion might then enhance the ability of ACF to interact with this site. Since HMGB1 does not form ternary complexes with ACF and the nucleosome under these conditions, we speculate that ACF1 might enable the remodelling factor to recognize DNA structure within the nucleosomal linker and hence sensitize it for pre-bent DNA. Remarkably, ACF1-like molecules appear to be obligatory subunits of ISWI-containing remodelling ATPases (Längst and Becker, 2001b; Xiao et al., 2001), and therefore seem to be tightly involved in the remodelling process itself. Although a direct allosteric effect on ACF cannot be excluded in light of the fact that HMGB1 is able to interact with several regulators of chromatin structure (Dintilhac and Bernues, 2001), this appears not to be the basis of our current findings since we have been unable to detect a direct interaction of HMGB1 with ISWI and ACF in the absence of nucleosomal DNA.

The observation that catalytic concentrations of HMGB1 accelerate the rate of nucleosome sliding, but not a mutant HMGB1 with 100-fold increased DNA binding, highlights the requirement for a dynamic association of HMGB1 with nucleosomes. Due to its higher affinity for the nucleosomal linker DNA, HMGB1ΔC might resist displacement by ACF and hence continue to ‘lock’ the nucleosome in analogy to histone H1. In agreement with previous findings (Lee and Thomas, 2000; Muller et al., 2001), the acidic C-terminus of HMGB1 is responsible for its dynamic interaction with nucleosomal DNA. Differential calorimetry experiments, both in the presence (Muller et al., 2001) or the absence of DNA (Ramstein et al., 1999), indicated that the acidic tail binds to one of the HMG boxes of HMGB1 (most likely box A). By dynamic competition with DNA this intramolecular binding might render the HMG box–DNA interaction reversible and transient. The DNA bulge released from box A could then be recognized by ACF.

Proteins

Histones were extracted from Drosophila embryos and purified by chromatography on a hydroxylapatite column as described (Simon and Felsenfeld, 1979). ISWI, ACF and CHRAC were prepared according to Eberharter et al. (2001).

HMGB1 and HMGB1ΔC protein expression and purification was described in Muller et al. (2001); HMGB1ΔC was identified there as the HMGB1 AB truncation.

Nucleosome assembly and purification

Nucleosomes were assembled using purified histones and polyglutamic acid (PGA; Sigma P4886) according to Stein et al. (1979). Histones and PGA were mixed in a 2:1 ratio in 0.15 M NaCl and incubated for 1 h at room temperature. Precipitates were removed by centrifugation at room temperature (RT) for 4 min at 11 kr.p.m. (Eppendorf, 5415R), the supernatant (HP-mix) was stored at –20°C. Different ratios of DNA and HP-mix were incubated for 3 h at 37°C and analysed by electromobility shift to reveal optimal conditions for nucleosome assembly. Translational positions of the nucleosomes were separated by electrophoresis on 4.5% polyacrylamide gels in 0.4× TBE and isolated as described previously (Längst et al., 1999).

Nucleosome mobility assay

Sixty femtomoles of nucleosomes with defined translational positions were incubated with ISWI (30 fmol), ACF (5 fmol), CHRAC (5 fmol), HMGB1 and HMGB1ΔC for 0–90 min at 16°C. Reaction mixes contained EX70 buffer (70 mM KCl, 10 mM Tris, 1 mM EDTA, 1.5 mM MgCl₂ and 10% glycerol) supplemented with 1 mM ATP, 200 ng/µl chicken egg albumin (CEA; Sigma) and 1 mM dithiotheritol (DTT) in a final volume of 10 µl. Reactions were stopped by the addition of 300 ng competitor DNA and separated on 4.5% polyacrylamide gels in 0.4× TBE. Gels were scanned with a phosphoimager (Fuji) and the individual bands were quantified with Aida software.

Electromobility shift assays

Sixty femtomoles of purified nucleosomes were incubated with increasing amounts of ISWI, ACF, HMGB1 and HMGB1ΔC in EX80 buffer containing 200 ng/µl CEA and 1 mM DTT. Reactions were incubated for 10 min at RT and then separated on 4.5% polyacrylamide gels containing 0.4× TBE. ACF-/HMGB1–nucleosome complexes were supershifted with the specific HMGB1 (Pharmingen) and anti-Flag (Sigma) antibodies.