|Home | About | Journals | Submit | Contact Us | Français|
The high mobility group N (HMGN) proteins are abundant nonhistone chromosomal proteins that bind specifically to nucleosomes at two high affinity sites. Here we report that purified recombinant human HMGN1 (HMG14) and HMGN2 (HMG17) potently repress ATP-dependent chromatin remodeling by four different molecular motor proteins. In contrast, mutant HMGN proteins with double Ser to Glu mutations in their nucleosome-binding domains are unable to inhibit chromatin remodeling. The HMGN-mediated repression of chromatin remodeling is reversible and dynamic. With the ACF chromatin remodeling factor, HMGN2 does not directly inhibit the ATPase activity, but rather appears to reduce the affinity of the factor to chromatin. These findings suggest that HMGN proteins serve as a counterbalance to the action of the many ATP-dependent chromatin remodeling activities in the nucleus.
Chromatin, the natural substrate for DNA-directed processes in the nucleus, is typically viewed as a nucleoprotein complex of histones and DNA. The structure and activity of chromatin has been studied largely through the analysis of enzymes that covalently modify histones as well as chromatin remodeling factors that use the energy of ATP hydrolysis to alter the structure and positioning of nucleosomes. It is also important to consider, however, that chromatin contains abundant nonhistone proteins, such as the high mobility group (HMG) proteins, that may contribute to the diversity of chromatin function.
In this study, we examine the biochemical activity of the HMGN1 and HMGN2 (also known as HMG14 and HMG17) proteins (for reviews, see Goodwin et al., 1978; Bustin, 2001; West, 2004; Hock et al., 2007). The HMGN proteins are small (~10 kDa) highly charged proteins that bind specifically to nucleosomes at two distinct sites in a manner that is independent of the underlying DNA sequence (Albright et al., 1980; Alfonso et al., 1994; Crippa et al., 1992; Mardian et al., 1980; Sandeen et al., 1980; Shick et al., 1985; Shirakawa et al., 2000; Ueda et al., 2008). These proteins are present in vertebrates, but have not been identified in other organisms such as Drosophila or yeast. In humans, there are four HMGN proteins: HMGN1 (HMG14), HMGN2 (HMG17), HMGN3 (Trip7), and HMGN4. These proteins are most highly conserved in their nucleosome binding domains, and are less related elsewhere. HMGN proteins are a major component of chromatin. Although initial estimates of the abundance of HMGN1 and HMGN2 were approximately one molecule per 10 nucleosomes (Goodwin et al., 1978), more recent measurements suggest that there are about 0.5 to 1.5 molecules of HMGN2 per nucleosome (Kuehl et al., 1984; Boumba et al., 2004). HMGN proteins have also been found to exchange rapidly in chromatin in vitro (Landsman et al., 1986) as well as in vivo (Phair and Misteli, 2000). Hence, like histone H1, they are dynamic rather than static components of chromatin.
Do HMGN proteins affect chromatin dynamics? HMGN1 and HMGN2 have been shown to increase the stability of core particles as assessed by thermal denaturation, circular dichroism, and nuclease digestion analyses (Crippa et al., 1992; Gonzalez and Palacian, 1990; Paton et al., 1983; Sandeen et al., 1980; Yau et al., 1983). Thus, HMGN proteins could potentially alter the structure of nucleosomes so that they become more resistant to remodeling. On the other hand, a previous study found that HMGN1 did not affect ATP-dependent chromatin remodeling by the SWI/SNF complex (Hill et al., 2005). We therefore sought to investigate further the role of HMGN proteins in chromatin remodeling, and undertook the analysis of the effect of purified recombinant human HMGN1 and HMGN2 upon chromatin remodeling catalyzed by four different ATP-dependent molecular motor proteins.
To investigate the biochemical properties of the HMGN proteins, we assembled HMGN-containing chromatin. Recombinant human HMGN1 and HMGN2 were synthesized in Escherichia coli and purified to apparent homogeneity by conventional chromatography (Figure 1A). Mass spectrometry analysis revealed that the HMGN1 and HMGN2 proteins each lack the initiating Met residue and thus have the same amino acid sequences as the native human HMGN proteins. HMGN proteins are highly charged and may potentially interact nonspecifically with other factors. Hence, to determine whether any of the observed biochemical properties of HMGN proteins are dependent upon the specific interaction of the proteins with chromatin, we generated and purified mutant versions of HMGN1 and HMGN2 that are defective for binding to nucleosomes (Figure 1A). These mutant proteins, termed HMGN1-S20, 24E and HMGN2-S24, 28E, each contain mutations in two conserved Ser residues that have been previously shown to be necessary for the binding of HMGN proteins to nucleosomes (Prymakowska-Bosak et al., 2001). We therefore used HMGN1-S20, 24E and HMGN2-S24, 28E as nucleosome binding-defective versions of their wild-type counterparts.
Chromatin was assembled with the purified ACF system in the presence or absence of wild-type or mutant HMGN proteins. The incorporation of HMGN proteins into chromatin was monitored by mononucleosome gel shift analysis (Albright et al. 1980; Mardian et al. 1980; Sandeen et al. 1980). In this assay, the binding of HMGN proteins to chromatin results in a reduction in the migration of mononucleosomes during nondenaturing gel electrophoresis. As shown in Figure 1B, wild-type but not mutant HMGN1 and HMGN2 are incorporated into their two high affinity binding sites in nucleosomes.
The assembly of HMGN-containing chromatin was additionally characterized by partial micrococcal nuclease digestion analysis (Noll and Kornberg, 1977). The incorporation of HMGN1 or HMGN2 into chromatin resulted in an increase of about 13–17 bp in the nucleosome repeat length (Figure 1C and data not shown), which is similar to the 13 bp increase seen previously with calf thymus HMGN2 (Paranjape et al., 1995). These experiments indicate that the purified recombinant HMGN1 and HMGN2 proteins are incorporated into chromatin and that the binding of HMGN proteins to chromatin increases the nucleosome repeat length.
Because HMGN proteins bind specifically to two high affinity sites on nucleosomes, we investigated their effect upon ATP-dependent chromatin remodeling. We reconstituted chromatin by using salt dialysis methodology, which involves only purified histones and DNA. In this manner, the resulting chromatin would not contain any nucleosome remodeling enzymes. We confirmed that the HMGN proteins bind to salt dialysis-reconstituted chromatin by using the nucleosome gel shift assay (Figure S1). We then proceeded to examine the effect of HMGN proteins on ATP-dependent chromatin remodeling by using the restriction enzyme accessibility assay, in which the access of a restriction enzyme (Hae III) to nucleosomal DNA is increased by the action of a chromatin remodeling factor that moves and/or disrupts nucleosomes. This assay has been used extensively for the study of different chromatin remodeling factors (see, for example: Varga-Weisz et al. 1998; Boyer et al., 2000; Shen et al. 2000; Alexiadis and Kadonaga, 2002).
These experiments revealed that both HMGN1 and HMGN2 potently inhibit chromatin remodeling by ACF, which has an ISWI ATPase subunit, as well as by BRG1, which is the ATPase subunit of some human SWI/SNF complexes (Figure 2). The extent of repression of chromatin remodeling increases progressively with the concentration of HMGN proteins (Figure S2). Importantly, the nucleosome binding-defective versions of HMGN proteins (HMGN1-S20, 24E and HMGN2-S24, 28E) do not inhibit chromatin remodeling (Figure 2). As a control, the addition of the HMGN proteins to naked DNA did not affect the digestion by the Hae III restriction enzyme (Figure 2).
To explore the generality of these effects, we further tested the ability of HMGN proteins to repress chromatin remodeling by other molecular motor proteins. We found that wild-type but not mutant HMGN proteins inhibit chromatin remodeling by Mi-2 (Figure S3) as well as by Rad54 + Rad51 (Figure S4). These findings indicate that the binding of HMGN proteins to chromatin results in the repression of chromatin remodeling by several different ATP-dependent remodeling factors.
How might the HMGN proteins inhibit chromatin remodeling? We addressed this issue with the ACF chromatin remodeling complex, which comprises Acf1 protein and the ISWI ATPase (Ito et al., 1999). First, we found that HMGN-mediated repression of chromatin remodeling could be overcome by the addition of higher concentrations of ACF (Figure 3A). Thus, the inhibition of chromatin remodeling by HMGN proteins is not due to an irreversible inactivation of the chromatin, such as by aggregation or precipitation.
In the experiments shown thus far, the HMGN proteins were added to the chromatin before the chromatin remodeling factors. We therefore investigated whether this effect requires the binding of the HMGN proteins to the chromatin prior to the addition of ACF. To address this question, we carried out order-of-addition experiments in which HMGN2 was added to chromatin before, after, or at the same time as ACF. In these experiments, we employed concentrations of HMGN2 (2.6 μM) and ACF (5 nM) that favor repression by HMGN2 (as in Figure 3A). When HMGN2 was added before, after, or at the same time as ACF, we observed repression of chromatin remodeling by wild-type but not nucleosome-binding-defective HMGN2 (Figure 3B). Notably, the preincubation of chromatin with ACF does not block the ability of HMGN2 to repress chromatin remodeling. Therefore, the results of the titration and order-of-addition experiments indicate that there is a reversible and dynamic interaction in which the chromatin remodeling factors and HMGN proteins act in opposition to promote or to repress nucleosome mobility.
Next, we tested whether HMGN proteins affect the ATPase activity of ACF. The ATPase activity of ACF is stimulated several-fold by DNA and to a greater extent by chromatin (Fyodorov and Kadonaga, 2002). It is possible, for instance, that the HMGN proteins block chromatin remodeling by direct inhibition of the ATPase activity of ACF. In this case, both the DNA- and chromatin-stimulated ATPase activities of ACF would be inhibited by the HMGN proteins. Alternatively, the HMGN proteins might block remodeling by reducing the interaction of ACF with chromatin. In this scenario, it would be expected that the chromatin-stimulated ATPase activity but not the DNA-stimulated ATPase activity of ACF would be inhibited by HMGN proteins. To test these models, we carried out ATPase assays with ACF and wild-type or mutant HMGN proteins in the presence of DNA or chromatin (Figure 3C). These experiments revealed strong inhibition of chromatin-stimulated ATPase activity and only modest inhibition of DNA-stimulated ATPase activity by wild-type HMGN2. Wild-type HMGN2 reduces the chromatin-stimulated ATPase activity to a level that is comparable to the DNA-stimulated ATPase activity. In contrast, the nucleosome binding-defective mutant HMGN2-S24, 28E did not inhibit the ATPase activity of ACF in the presence of either DNA or chromatin. These findings suggest that the binding of HMGN proteins to nucleosomes reduces the interaction of ACF with chromatin.
To investigate further the effect of HMGN proteins on the binding of ACF to chromatin, we carried out glycerol gradient sedimentation analyses (Figure 4A). In these experiments, the conditions for the incubation of the chromatin with HMGN proteins and ACF were identical to those used in the restriction enzyme accessibility assays. Gradient fractions containing chromatin were identified by the cosedimentation of DNA and core histones, which were detected by western blot with antibodies against H2A-H2B. ACF and HMGN2 were monitored by western blot analysis with antibodies against ISWI and HMGN2, respectively.
In contrast to the findings of previous studies in which chromatin was reconstituted with crude Xenopus egg extracts (Trieschmann et al., 1995) or tandemly-repeated nucleosome positioning sequences (Hill et al., 2005), we found that the incorporation of HMGN2 into chromatin prepared with purified components and nonrepetitive DNA results in a distinct increase in the rate of sedimentation of chromatin (Figure 4A; compare ‘Chromatin’ with ‘Chromatin + HMGN2’). The binding of ACF to chromatin also leads to an increase in the rate of chromatin sedimentation (Figure 4A; compare ‘Chromatin’ with ‘Chromatin + ACF’). If both ACF and HMGN2 could simultaneously bind to chromatin, it would be expected that the resulting complex would sediment at a faster rate than either HMGN- or ACF-bound chromatin. However, when both HMGN2 and ACF are added to chromatin, the majority of ACF is not associated with chromatin, and the rate of sedimentation of the chromatin is similar to that of HMGN-containing chromatin (Figure 4A; compare ‘Chromatin + HMGN2’, ‘Chromatin + ACF’, and ‘Chromatin + HMGN2 + ACF’). These findings indicate that the binding of HMGN proteins to nucleosomes inhibits the association of ACF with chromatin. Hence, the binding of HMGN proteins to nucleosomes decreases the ability of ACF to recognize and to remodel chromatin.
In this study, we found that HMGN proteins repress chromatin remodeling by several ATP-dependent molecular motors. There are many different chromatin remodeling factors that facilitate a broad range of DNA-directed processes. HMGN proteins may thus function to oppose the action of chromatin remodeling factors to increase the stability of the chromatin (Figure 4B).
Repression of ATP-dependent chromatin remodeling has been observed by the Polycomb repressive complex 1 (PRC1; Shao et al., 1999). The key component in PRC1 for inhibition of chromatin remodeling was found to be posterior sex combs protein (PSC), which binds to DNA and reduces the binding of SWI/SNF complex to DNA (Francis et al., 2001). It is not known if PSC binds specifically to nucleosomes in a manner similar to the HMGN proteins. Given the low abundance and regulatory function of PRC1, it is likely that it acts in a more restricted manner than the HMGN proteins, which probably have a more general influence on chromatin stability.
The effects of histone H1 on chromatin remodeling have also been examined. Histone H1 binds to nucleosomes in the vicinity of the pseudo-dyad and the linker DNA. H1 was found to cause two- to three-fold inhibition of chromatin remodeling by the SWI/SNF complex (Hill and Imbalzano, 2000). In another work, histone H5, an H1 variant in avian and amphibian erythrocytes that binds to nucleosomes with higher affinity than H1, was found to be a strong inhibitor of chromatin remodeling by SWI/SNF, Mi-2, and ACF (Horn et al., 2002). In a third study, H1 did not to have a strong inhibitory effect on nucleosome repositioning catalyzed by the SWI/SNF complex (Ramachandran et al., 2003). Thus, except in highly repressed avian and amphibian nuclei with histone H5, it appears that H1 has a mild repressive effect on chromatin remodeling activity.
In apparent contrast to our findings, a previous study found that HMGN1 does not affect chromatin remodeling by the SWI/SNF complex (Hill et al., 2005). In the earlier work, the experiments were performed with different factors under different conditions than those used in this study. Hence, it is not possible to identify the exact basis of the different results. Some potentially important parameters include HMGN, chromatin, and remodeling factor concentrations, buffer composition, the nature of the reconstituted chromatin, and properties of the chromatin remodeling factors. A key aspect of our experiments is the use of the HMGN1-S20, 24E and HMGN2-S24, 28E versions of the HMGN proteins as nucleosome binding-defective control proteins. The inability of the mutant HMGN proteins to inhibit chromatin remodeling ensures that the repression by the wild-type proteins is due to the specific binding of the proteins to nucleosomes. The concentrations of HMGN proteins and chromatin are also important parameters. By using the nucleosome gel shift assay, as in Figure 1B, we determined the minimum concentrations of HMGN proteins that are required to achieve near complete binding of two HMGN proteins per nucleosome, and we then used the same protein and chromatin concentrations in the restriction enzyme accessibility assays. Notably, the extent of binding of HMGN proteins to nucleosomes correlates with the extent of repression of chromatin remodeling activity (as in Figure S2). Lastly, to determine the generality of the effects, we tested both wild-type and mutant versions of HMGN1 and HMGN2 with four different chromatin remodeling factors. The resulting data provide considerable evidence that HMGN1 and HMGN2 repress chromatin remodeling.
The observation that HMGN proteins repress chromatin remodeling may suggest a role of these proteins in transcriptional repression, such as that seen for HMGN1 with Sox9 and N-cadherin genes (Furusawa et al., 2006; Rubinstein et al., 2005). However, though it is often presumed that nucleosome mobility correlates with gene activity, there may not necessarily be a strict linkage between the two phenomena. Chromatin remodeling could be used to facilitate gene activation or repression. For instance, the NuRD (also known as NURD, NRD, and Mi-2) chromatin remodeling complex contains histone deacetylases that appear to be involved in transcriptional repression (Tyler and Kadonaga, 1999). Inertness or resistance to chromatin remodeling could be used to maintain an active or repressive state of a gene. Thus, the ability of HMGN proteins to inhibit chromatin remodeling does not necessarily indicate that these nucleosome-binding proteins promote or repress gene activity.
Although HMGN proteins repress ACF-mediated chromatin remodeling, they do not inhibit the ability of ACF to assemble chromatin (Figures 1B and 1C and data not shown). This result is consistent with the finding that HMGN proteins do not inhibit the DNA-stimulated ATPase activity of ACF (Figure 3C). Thus, prior to the formation of nucleosomes, HMGN proteins do not impede ACF-catalyzed chromatin assembly.
Our observations regarding HMGN1 and HMGN2 and chromatin remodeling do not preclude other potential functions of HMGN proteins, such as in DNA repair (Birger et al., 2003) or regulation of transcription factor activity (Zhu and Hansen, 2007). In the latter study, HMGN1 was found to repress estrogen-stimulated transcription via specific interactions between HMGN1 and the estrogen receptor αor SRF transcription factors. It is interesting to consider that there may be a link between the binding of HMGN1 to the DNA-binding transcription factors and the repression of chromatin remodeling by HMGN1.
Toward the longer term goal of understanding the functions of the abundant nonhistone components of chromatin, we have found that the HMGN chromosomal proteins, which bind specifically to two high affinity sites on nucleosomes, function to counteract chromatin remodeling by ATP-dependent molecular motor proteins. The interaction between the factors is reversible and dynamic, and the HMGN proteins may thus serve as a counterbalance to the action of the many chromatin remodeling activities in the nucleus.
Chromatin was reconstituted either by using the purified ACF assembly system or by salt dialysis techniques. Human (HeLa cell) core histones were used in all experiments shown in the main text and Supplemental Data, except for those shown in Figure 1 and Figure 4A, in which chromatin was assembled with Drosophila core histones. ACF-mediated chromatin assembly was performed as described (Fyodorov and Kadonaga, 2003). A standard reaction contained core histones (0.35 μg), NAP-1 (1.4 μg), ACF (3 nM), MgCl2 (5 mM), ATP (3 mM), an ATP regeneration system (3 mM phosphoenolpyruvate, 15 U/mL pyruvate kinase) and relaxed circular DNA plasmid pGIE-0 (3.2 kb; 0.35 μg) in a final volume of 70 μL. Recombinant wild-type or mutant HMGN proteins were added in a 10:1 molar ratio to histone octamers, unless otherwise specified. The final concentration of KCl was 80 mM. Salt dialysis reconstitution of chromatin was performed with purified core histones and plasmid DNA by the method of Jeong et al. (1991). The resulting chromatin was purified by 15 to 50% sucrose gradient sedimentation. Micrococcal nuclease digestion analyses were performed as previously described (Pazin et al., 1997).
Restriction enzyme accessibility assays were performed with salt dialysis-reconstituted chromatin as described previously (Alexiadis and Kadonaga, 2002). In reactions containing HMGN proteins, chromatin (or naked DNA control) was pre-incubated with wild-type or mutant HMGN protein at 27°C for 20 minutes prior to the addition of the chromatin remodeling factor, unless specified otherwise.
We thank Jer-Yuan Hsu, Tammy Juven-Gershon, Debra Urwin, Alexandra Lusser, Joshua Theisen, and Sharon Torigoe for critical reading of the manuscript. We also thank Alexandra Lusser (Medical University of Innsbruck) for the generous gift of purified Mi-2. B.P.R. (F32GM74527) and T.Y. (F32GM76936) were supported in part by NRSA fellowships from the National Institutes of Health. This work was supported by a grant from the National Institutes of Health (GM058272) to J.T.K.
The Supplemental Data for this article include additional details regarding the experimental procedures, and four figures.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.