Site-specific acetylation of ISWI by GCN5

Background The tight organisation of eukaryotic genomes as chromatin hinders the interaction of many DNA-binding regulators. The local accessibility of DNA is regulated by many chromatin modifying enzymes, among them the nucleosome remodelling factors. These enzymes couple the hydrolysis of ATP to disruption of histone-DNA interactions, which may lead to partial or complete disassembly of nucleosomes or their sliding on DNA. The diversity of nucleosome remodelling factors is reflected by a multitude of ATPase complexes with distinct subunit composition. Results We found further diversification of remodelling factors by posttranslational modification. The histone acetyltransferase GCN5 can acetylate the Drosophila remodelling ATPase ISWI at a single, conserved lysine, K753, in vivo and in vitro. The target sequence is strikingly similar to the N-terminus of histone H3, where the corresponding lysine, H3K14, can also be acetylated by GCN5. The acetylated form of ISWI represents a minor species presumably associated with the nucleosome remodelling factor NURF. Conclusion Acetylation of histone H3 and ISWI by GCN5 is explained by the sequence similarity between the histone and ISWI around the acetylation site. The common motif RKT/SxGx(Kac)xPR/K differs from the previously suggested GCN5/PCAF recognition motif GKxxP. This raises the possibility of co-regulation of a nucleosome remodelling factor and its nucleosome substrate through acetylation of related epitopes and suggests a direct crosstalk between two distinct nucleosome modification principles.


Background
Disruption of DNA-histone interactions by nucleosome remodelling ATPases may lead to a variety of transitions of chromatin structure, such as the partial or complete disassembly of nucleosomes, the exchange of histones, or the sliding of intact histone octamers on DNA [1][2][3][4]. In many cases their activity is focused on local disruption of the nucleosomal fibre through recruitment of DNA-binding regulators to promote access of factors further downstream in the cascade of events that leads to promoter opening [2]. However, genome-wide processes like replication, DNA damage responses or homologous recombi-nation require chromatin to be dynamic on a global scale. In addition to generating local access to nucleosomal DNA, nucleosome disruption may also have profound consequences for the folding of the nucleosomal fibre into higher order structures [5,6].
One largely unresolved issue to date is how the activity of chromatin remodelling enzymes is regulated. Established regulatory principles that govern, for example, metabolic enzymes also apply to nucleosome remodelling ATPases, but our knowledge is still anecdotal. The expression of nucleosome remodelling ATPases may be selective. For example, the fact that the mammalian ISWI isoform SNF2H is abundant in proliferating cells, whereas the related SNF2L is enriched in terminally differentiated neurons points to functional diversification of related remodelling enzymes and specialized roles in proliferation/ differentiation control [7]. Such a role has also been suggested for the SWI2/SNF2-type ATPase hBRM by the early finding that its expression is down-regulated when cells receive a mitogenic stimulus or during ras-mediated oncogenic transformation, whereas its forced expression partially reverses transformation [8].
A further regulatory strategy involves post-translational modifications of enzymes, such as their phosphorylation. Phosphorylation of hBRM and BRG-1 during mitosis correlates with dissociation of these remodellers from the chromosomes during condensation [9]. Muchardt and colleagues also showed that acetylation of hBRM correlates with a reduced inhibition of cell growth [10]. The possibility of regulating nucleosome remodelling ATPases by lysine acetylation is intriguing given that properties of their substrates, the histones, are most prominently modulated by acetylation at their exposed N-termini [11].
Here we describe another example of potential regulation of a remodelling ATPase by acetylation. We found that Drosophila ISWI, the founding member of a family of nucleosome remodelling ATPases, was preferentially acetylated by GCN5 at a single lysine within a amino acid sequence of high similarity to the N-terminus of histone H3. This acetylated lysine corresponds to lysine 14 in histone H3 (H3K14), a known target for GCN5, suggesting that this acetyltransferase may affect the chromatin structure by two distinct strategies: by acetylation of the nucleosomes and by modification of a nucleosome remodelling enzyme.

ISWI is acetylated in vivo
In order to explore whether ISWI was acetylated in Drosophila cell lines we immunoprecipitated the ATPase from extracts of SF4 cells ( Figure 1A, lanes 1-3). Probing the precipitate with a pan-acetyl-lysine antibody (αAcLysine) we detected a labelled protein migrating at the position of ISWI ( Figure 1A, lanes 4, 5). To facilitate detection of  acetylated ISWI we treated Kc cells with the histone  deacetylase inhibitor Trichostatin A (TSA), prepared whole cell extracts and monitored ISWI levels ( Figure 1B,  lanes 1, 2). We then immunoprecipitated ISWI or proteins containing acetylated lysines from these extracts. In the absence of TSA ISWI was barely detectable in the αAcLysine precipitate, which may be due to the inefficiency of the antibody and/or the small amounts of acetylated ISWI present in Kc cells ( Figure 1B, lanes 3-5). However, upon overnight TSA treatment the levels of acetylated ISWI increased significantly ( Figure 1B, lanes 6-8). Taken together these results suggested that a minor fraction of ISWI was acetylated in Kc and SF4 cells. In order to confirm this notion by an independent experiment we metabolically labelled SF4 and Kc cells by addition of [ 3 H]acetic acid for 3 hrs to the growth medium, prepared extracts and determined the ISWI levels as before ( Figure  1C, lanes 9, 10). ISWI was immunoprecipitated from these extracts ( Figure 1C, lanes 5-8) and acetylated proteins in the precipitate were detected by gel electrophoresis and autoradiography. A band migrating at the position of ISWI was only detectable in the αISWI precipitate, but was absent in the control ( Figure 1C, lanes 1-4). A second labelled band points to an acetylated ISWI-associated protein of unknown identity (asterisk in Figure 1C). Collectively, these data demonstrate that a relatively minor fraction of ISWI is acetylated in Drosophila tissue culture cells.

Histone acetyltransferases with substrate preference for the histone H3 N-terminus acetylate ISWI in vitro
We wished to identify acetyltransferases able to acetylate ISWI. Towards this end we expressed and purified GCN5, p300 and MOF and confirmed their activity and specificity on histone substrates. In agreement with published work [12][13][14], GCN5 only acetylated histone H3, p300 mainly acetylated H3 and to a lesser extent H4 and MOF preferred histone H4 as a substrate (Figure 2A). Replacing the histone substrate with recombinant ISWI we found that GCN5 and p300 were also able to acetylate ISWI, whereas MOF did not ( Figure 2B). The activity of all HATs in these assays is conveniently controlled by their autoacetylation (bands labelled with asterisks in Figure  2B). Interestingly, the ability to acetylate ISWI correlated with the enzymes' preference for H3 among the histones. Because acetylation of ISWI by GCN5 was particularly robust, we employed GCN5 in the subsequent analyses.

GCN5 acetylates ISWI at lysine 753
We next wished to identify the site(s) at which ISWI was acetylated by GCN5 in vitro. The ATPase domain of ISWI is located in the central part of the protein ( Figure 3A). While more N-terminal sequences of the protein have no discernable features, the structure and function of the Cterminal third are better known. A region defined by HAND, SANT and SLIDE domains harbours a prominent nucleosome recognition module [15]. At the very C-terminus an ACF1-interaction determinant (AID) has been described [16]. A panel of different ISWI fragments fused to hexahistidine-(His 6 -) or GST-tags ( Figure 3) were expressed in E. coli and used as substrates for GCN5dependent acetylation. Acetylation of the first 697 N-terminal amino acids (aa) of ISWI containing the ATPase domain and sequences up to the HAND domain was barely detectable ( Figure 3B, lane 1). By contrast, the remaining C-terminal part (aa 691-1027) was strongly labelled ( Figure 3B, lane 2). Trimming this fragment from the C-terminus to aa 962 did not affect its acetylation, however, deleting amino acids between 691 and 796 abolished the acetylation ( Figure 3B, lanes 4-7). We concluded that a prominent acceptor site for acetylation must reside within the HAND fold [15]. This region of ISWI contains 10 lysines that could potentially be acetylated. Upon closer inspection we found the region between aa 747 and 756 to be remarkably similar to the N-terminus of histone H3 ( Figure 4A). In this alignment ISWI residues 748 and 753 (K748 and K753, respectively) correspond to lysines 9 and 14 of H3, which are frequently acetylated by GCN5. These ISWI residues and their sequence context are conserved in Xenopus laevis (xISWI) and the human ISWI Figure 1 ISWI is acetylated in vivo. (A) ISWI was immunoprecipitated from whole cell extracts of Drosophila SF4 cells. A parallel immunoprecipitation (IP) with rabbit IgG antibodies served as specificity control. Enrichment of ISWI was documented by Western blotting (WB) using an αISWI antibody (lanes 1-3). A pan-acetyl-lysine (αAcLysine) antibody detects a protein comigrating with ISWI (lanes 4, 5). (B) Kc cells were cultured for 17 hrs in the presence or absence (control) of 200 ng/ml trichostatin A (TSA). Whole cell extracts contained equal amounts of ISWI (lanes 1, 2). Proteins were immunoprecipitated from these extracts with αISWI or αAcLysine antibody as indicated. ISWI was detected in the precipitate by Western blot analysis. (C) Drosophila SF4 and Kc cells were metabolically labelled with [ 3 H]-acetic acid and extracts prepared. Extract ISWI levels were visualised by Western blotting (lanes 9, 10). ISWI was immunoprecipitated from these extracts and visualised by Western blotting (lanes 5-8). An equivalent gel was then autoradiographed (lanes 1-4) in order to detect acetylated protein. The arrow points to the position of ISWI, the asterisk indicates an unknown acetylated protein.

ISWI is acetylated in vivo
isoforms hSNF2H and hSNF2L ( Figure 4A). We therefore focused our attention on this part of ISWI.
A peptide corresponding to aa 1-19 of histone H3 is acetylated rather well by GCN5 and p300 in a standard HAT reaction, but less by MOF ( Figure 4B, left panel). By direct comparison, a peptide spanning the corresponding 19 aa of ISWI ( Figure 4A) is well acetylated by GCN5, but not at all by MOF ( Figure 4B, right panel). Maldi-TOF mass spectrometry and nano-electrospray sequencing documented that the peptide was mono-acetylated (Figure 4C) exclusively at K753 ( Figure 4D). In order to test for the contribution of K753 to ISWI acetylation in the context of the entire protein we expressed recombinant ISWI derivatives that had either K748 or K753 or both lysines replaced by alanines (A) or arginines (R). Equivalent protein amounts ( Figure 4E, lower panel) were used as substrate for GCN5-dependent acetylation reactions. The ISWI derivative bearing the K748A mutation was acetylated as the wildtype protein, whereas replacement of lysine 753 by alanine (K753A) led to a significantly reduced acetylation ( Figure 4E, compare lanes 4, 5). Collectively the data identify K753 of ISWI, which appears related by sequence context to H3K14, as the major target of acetylation by GCN5 in vitro.
We were not able to evaluate the effect of acetylation on ISWI ATPase activity due to the small fraction of ISWI modified in these in vitro reactions. However, replacement of either lysine 753 or lysine 748 (which corresponds to lysine 9 in H3) by arginine or alanine did not affect the ATPase activity in response to DNA or nucleosomes (Figure 4F).

ISWI is acetylated at K753 by GCN5 in vivo
In order to monitor the distribution of ISWI K753ac in vivo we raised an antibody against an ISWI peptide acetylated at K753. The antibody recognised ISWI only after acetylation by GCN5 ( Figure 5A, lanes 5, 6) and did not crossreact with H3 even when the histone was acetylated to much higher levels ( Figure 5A, B; compare lanes 4 and 6). The antibody precipitated ISWI from Drosophila embryo extracts ( Figure 5C) as well as a protein of the size of ISWI that was stainable with the ISWI K753ac antibody ( Figure  5D).
The antibody also recognised an epitope in nuclei of Drosophila SL2 cells ( Figure 6B, upper left panel). To confirm that the antibody reactivity was specific in vivo we performed RNA interference on SL2 cells with dsRNA corresponding to ISWI or GCN5 sequences and monitored the success of the knockdown by Western blotting of whole cells extracts ( Figure 6A) and immunofluorescence (Figure 6B). Knocking down either ISWI or GCN5 slowed cell proliferation effectively increasing the doubling time two fold. Ablation of ISWI removed not only the signal obtained with the ISWI antibody, but also the ISWI K753acspecific signal ( Figure 6B). Remarkably, knockdown of GCN5 also greatly reduced the signal obtained with the ISWI K753ac antibody, but did not affect ISWI in general ( Figure 6B). Taken together, these data document the specificity of the antibody for ISWI K753ac in immunofluorescence applications. They also show that ISWI is acetylated at K753 by GCN5 in SL2 cells.

Acetylated ISWI may reside in the NURF complex, but not in CHRAC/ACF
So far ISWI is known to reside in two kinds of nucleosome remodelling complexes in Drosophila: CHRAC/ACF, which are characterised by the ACF1 subunit, and NURF, which contains NURF301 as a defining feature [17]. In order to explore whether ISWI K753ac was preferentially associated with one of these complexes we ablated the signature subunits ACF1 and NURF301 by RNA interference in SL2 cells ( Figure 6A) and found that the ISWI K753ac epitope was significantly reduced in the absence of NURF301, but remained unperturbed if ACF1 was knocked down ( Figure 6B). In agreement with the notion that ISWI K753ac is not a subunit of CHRAC/ACF we found that the expression pattern of the acetylated ISWI form was unperturbed in acf1 null flies (data not shown).
In vitro, ISWI is specifically acetylated by HATs able to modify histone H3 Figure 2 In vitro, ISWI is specifically acetylated by HATs able to modify histone H3. (A) The histone acetyltransferase activities of recombinant GCN5, p300 and MOF were assayed in reactions containing all four histones. Incorporation of [ 3 H]-acetyl-coenzyme A leads to labelled proteins, which are separated by SDS-PAGE and detected by autoradiography of the gel. Bands corresponding to acetylated histones are marked by arrows. (B) Recombinant ISWI was incubated in acetylation reactions with GCN5, p300 or MOF. Acetylated ISWI (arrow) was detected as in (A). All other bands (asterisks) are due to auto-acetylation of the corresponding enzyme.
ISWI K753ac staining was unperturbed in flies heterozygous for the NURF301 allele [E(bx) ry122 ] [18]. Homozygous mutants are lethal, but survive until the end of the larval stage due to maternal contribution of NURF301 and were not analysed. Immunoprecipitation of NURF301 from extracts of SL2 cells and 8-12 hr embryos co-precipitated acetylated ISWI, which was not the case if ACF1 was precipitated ( Figure 6C, D). We conclude that in SL2 cells ISWI may be acetylated in the context of NURF but is not acetylated in association with ACF1. The data do not exclude the existence of additional complexes harbouring the modification.

Discussion
Many acetyltransferases were first identified following their histone acetylation function and hence named histone acetyltransferases (HATs). However, most of them have now been shown to acetylate non-histone proteins as well, and these modifications can have profound effects on the structure and function [19]. For example, the effects of acetylation on the tumour suppressor protein p53 have been heavily studied [20]. The current work, together with the earlier observation that the effect of the BRM ATPase on cell proliferation can be regulated by acetylation [10] highlight the potential for direct regulatory cross-talk between different chromatin modifying principles [21].
The fact that ISWI can be acetylated in vitro and in tissue culture cells by GCN5, which is also known to target the H3 N-terminus, is explained by the sequence similarity between the histone and ISWI around the K753 acetylation site. The common motif RK T / S xGx(K ac )xP R / K differs from the previously suggested GCN5/PCAF recognition motif GKxxP [19]. This raises the possibility of specific coregulation of a nucleosome remodelling factor and its nucleosome substrate through acetylation of related epitopes. Strikingly, the related motif DKGKGKKRP also occurs in the C-terminus of hBRM, overlapping with potential acetylation sites [10]. Acetylation at one or more C-terminal lysines of BRM correlates with reduced interference with cell cycle progression in mammalian cells and mutation of several of these lysines into arginines impairs BRM function. The data suggest that site-specific acetylation of the ATPase by, for example P/CAF, may adversely affect its function, possibly by interfering with targeting interactions [10].
The fact that the acetylated ISWI epitope is conserved during evolution suggests functional significance. However, so far we were unable to document an effect of acetylation on the catalytic properties or protein interactions of ISWI, due to the small proportion of ISWI either acetylated in vitro or naturally present in vivo. Lysine 753 of ISWI resides in the HAND domain, a fold of unknown function situated between the ATPase domain and the C-terminal SANT/SLIDE domains that mediate nucleosome interactions [15]. Exploring the role of the HAND domain in ISWI may lead to novel approaches towards evaluating the impact of K753 acetylation. Applying the K753ac-specific antibody to developmental stages of Drosophila revealed that the acetylation of ISWI is a developmentally regulated process, which occurs during oogenesis and early embryogenesis, but is absent from later developmental stages (CC and PBB, unpublished observations). We also obtained preliminary support for the idea that the acetylation of ISWI marks a functionally distinct subpopulation: the acetylated epitope was dramatically enriched on the condensed metaphase chromosomes in early embryos (CC and PBB, unpublished observations). A causal role between the acetylation and the unusual ability of the ATPase to interact with compact chromatin remains to be explored.

ISWI is acetylated in the HAND domain
The ATPase ISWI is the catalytic moiety of several nucleosome remodelling complexes. Better known and characterized are the factors CHRAC, ACF and NURF (for review, see [22]. The available data suggest that the acetylated form of ISWI may mark a subpopulation of the NURF complex. NURF is involved in the transcription of selected sets of Drosophila genes [18,23] but flies mutant for the large, defining NURF301 subunit also show defects of chromosome structure [23]. Assessment of which of these functions may be modulated by ISWI acetylation requires a genetic approach.
The acetylation of the amino terminus of histone H3 is an important determinant of active chromatin, but chromatin structure can also be modulated by other modifications of the histone 'tail', notable its methylation or phosphorylation. Whether the analogy between the ISWI and the H3 N-terminus can also be extended to include phosphorylation and methylation remains an interesting aspect for future research.
Characterisation of an antibody specific forISWI K753ac Western blotting on such a substrate using αISWI K753ac did not reveal any cross-reactivity. (C) ISWI was immunoprecipitated from embryonic nuclear extract using antibodies against ISWI or ISWI K753ac and detected by Western blotting. (D) αISWI K753ac immunoprecipitates were analysed by Western blotting for the presence of acetylated ISWI.

Conclusion
We identified and characterised a site-specific acetylation of the remodelling ATPase ISWI by GCN5. The modification occurs predominantly in the context of the remodelling factor NURF, however, due to the fact that the acetylation marks a minor subpopulation of the remodeller a functional evaluation of the modification has not yet been achieved.

Cell culture, RNA interference and immunostaining
Drosophila SF4 cells (an SL2 clone sorted for diploidy and adherence) were provided by D. Arndt-Jovin (Göttingen). SF4, Kc and SL2 cells were maintained at 26°C in Schneider's Drosophila medium (Invitrogen) supplemented with 10% fetal calf serum (FCS), antibiotics and glutamine. When needed, trichostatine A (TSA) was added at a final concentration of 200 ng/ml to the medium for 17 hrs before harvesting the cells. RNAi experiments were performed as described [27]. The primers used to generate the double-stranded RNA are listed in the supplemental materials. Briefly, 1-2 × 106 SL2 cells were seeded into 6-wells plates in 1 ml of medium without FCS just before adding 10 µg of dsRNA. Plates were then placed onto a shaking platform for 10 min and then for 50 min at 26°C. 2 ml of medium complemented with FCS were then added to the cells and incubated at 26°C. Cells were collected 10 or 12 days after dsRNA treatment. Protein extractions for Western blot analysis after RNAi, 1-2 × 106 cells were pelleted and lysed in 50 µl of urea buffer (8 M urea, 5% SDS, 200 mM Tris-Cl pH 6.8, 0.1 mM EDTA, 100 mM DTT) and incubated at 65°C for 15 min. Immunostainings were performed as described [28]. For immunostaining following RNAi, cells were collected 12 days after dsRNA treatment. After fixation, permeabilisation and blocking, cells were incubated for 1 hr with antibody against ISWI, diluted 1:600 in blocking solution (2% BSA and 5% goat serum in PBS), or aISWIK753ac, diluted 1:300. After washes, cells were incubated for 1 hr with Cy3-conjugated secondary antibody (Jackson Immunoresearch Laboratories) diluted in blocking buffer. Cells were washed four times in PBS. DNA was counterstained with 1 µg/ml bisbenzimide (Hoechst 33258). Slides were mounted using 1.5% n-propyl gallate, 50% glycerol in PBS. Images were acquired using a Zeiss Axiophot microscope coupled to a Retiga Exi CCD Camera (Qimaging, Burnaby, Canada). Images were cropped and levels adjusted in Photoshop.

Immunoprecipitation
Total cell extracts were prepared as described [29] from 4 × 10 7 Drosophila cells. All steps were performed at 4°C. Cells were washed with PBS, resuspended in 1 volume of lysis buffer (50 mM Tris-Cl pH 8.0, 300 mM NaCl, 10 mM MgCl2, 0.4% NP40 and proteases inhibitors) and incubated for 15 min. The supernatant was cleared by centrifugation and mixed with 1 volume of dilution buffer (50 mM Tris-Cl pH 8.0, 0.4% NP40). Diluted extracts were pre-cleared with protein A/protein G Sepharose beads (Amersham Biosciences) and incubated with αISWI, αAcLysine or irrelevant antibodies, and immunoprecipitated with protein A/protein G Sepharose beads. Following extensive washing with a 1:1 mix of lysis and dilution buffer, beads were resuspended in Laemmli buffer and proteins were analyzed by Western blot analysis.