|Home | About | Journals | Submit | Contact Us | Français|
Defining the protein factors that directly recognize post-translational, covalent histone modifications is essential towards understanding the impact of these chromatin ‘marks’ on gene regulation. In the current study, we identify human CHD1, an ATP-dependent chromatin remodeling protein, as a factor that directly and selectively recognizes histone H3 methylated on lysine 4. In vitro binding studies identified that CHD1 recognizes di- and tri-methyl H3K4 with a dissociation constant of approximately 5 μM, whereas mono-methyl H3K4 binds CHD1 with a threefold lower affinity. Surprisingly, human CHD1 binds to methylated H3K4 in a manner that requires both of its tandem chromodomains. In vitro analyses demonstrate that unlike human CHD1, yeast Chd1 does not bind methylated H3K4. Our findings indicate that yeast and human CHD1 have diverged in their ability to discriminate covalently modified histones, and link histone modification-recognition and non-covalent chromatin remodeling activities within a single human protein.
In the nucleus of cells, chromatin represents the physiological state of DNA, where it is associated with histone and non-histone proteins. Nucleosomes are the repeating unit of chromatin, and are comprised of two copies each of the four histone proteins, H2A, H2B, H3, and H4, wrapped by 147 base pairs of DNA. Histones are subject to a variety of post-translational, covalent modifications, such as methylation, acetylation, phosphorylation, and ubiquitination (1). In particular, global changes in histone methylation and acetylation are directly associated with cancer and more importantly, may serve as a predictive indicator of clinical outcome (2). Moreover, a number of enzymes that catalyze histone lysine methylation are altered in various types of cancer and include enzymes that catalyze methylation on histone H3 lysine 4, such as MLL and SMYD3 (3,4). In eukaryotes, lysine 4 tri-methylation on histone H3 (H3K4me3) is tightly associated to the 5’ end of active genes (5,6). However, the functional and mechanistic significance of H3K4me3 is poorly understood.
Generally, silent genes correspond with methylation at H3K9, H3K27, and H4K20, whereas active genes correspond with methylation at H3K4, H3K36, and H3K79 (7,8). The chromodomain is a protein motif that specifically recognizes methylated lysine residues within histones. The chromodomain-containing proteins HP1 and Pc recognize methylated lysines that correspond with silent genes, specifically H3K9 and H3K27, respectively. It was recently reported that the yeast chromodomain protein Chd1 binds directly to di-methyl H3K4 (H3K4me2) (9). Yeast and human CHD1 encode two chromodomains in tandem, although binding of yeast Chd1 to H3K4me2 was observed to occur only through its second chromodomain (9). However, previous studies in yeast showed that the Chd1 double chromodomain does not bind to methylated H3K4 (10). CHD1 is an ATP-dependent chromatin-remodeling factor that also functions as a chromatin assembly factor in vitro (11,12). CHD1 associates with active loci and is suspected to function during elongation, as deletion of CHD1 causes hyper-resistance to a 6-azauracil phenotype (13), Furthermore, CHD1 displays physical and genetic interactions with numerous elongation factors (14–17).
Here, we report the specific and direct binding of human CHD1 to methylated H3K4. Human CHD1 was found to require both of its tandem chromodomains to interact with H3K4me3. Comparative analyses between yeast and human CHD1 using the crystal structures of chromodomain-methyl lysine interactions (18,19) predict that yeast Chd1 will not recognize methyl H3K4. Indeed, in vitro binding analyses demonstrate that yeast Chd1 fails to bind methylated H3K4. Thus, only the human CHD1 protein links covalent histone modifications with chromatin remodeling. Our results also suggest that yeast and higher organisms have diverged in their recognition and readout of modified histone tails.
Peptide-coupled resins were generated using SulfoLink coupling gel (Pierce). Histone peptides were synthesized by Global Peptide Services. CHD1 derived from HeLa nuclear extract was incubated with H3K4me0 or H3K4me3 peptides columns for 2 hours at 4° C, washed extensively (60 column volumes of 25 mM Tris, pH 8, 150 mM NaCl, 2 mM EDTA, and 0.5% NP40), and eluted with low pH buffer (100 mM glycine, pH 3.0). Various peptide-conjugated beads were incubated with yeast extracts expressing PtA tagged Chd1 or Gcn5 and analyzed as described previously (10). 10 ug of recombinant protein was incubated with various peptide columns for 2 hours at 4° C, washed extensively, and eluted with low pH buffer.
The region encompassing the tandem chromodomains of human CHD1 was amplified by PCR using a HeLa cDNA library (BD Biosciences) and cloned into pGEX-6P1 (Amersham). Point mutants of CHD1 were generated by site-directed mutagenesis (Stratagene). The tandem chromodomains of yeast Chd1 in pET11a was kindly provided by S. Khorasanizadeh. Proteins were expressed in bacteria (BL21, Novagen) and purified using standard procedures. The SANT domain of human SNF2h, and the human CHD1 fragments CD1 and CD2 were kindly provided by Rui-Ming Xu (Cold Spring Harbor). Western blotting was performed using standard techniques with antibodies against GST (Amersham), PAP (Dako), and CHD1 (kind gift of Dr. R. P. Perry) (20). Silver staining was performed using standard molecular biology methods.
Fluorimetric peptide titration experiments were performed using a FluoroMax-2 spectrofluorimeter (Spex Instruments) as described (21), using 25 mM NaCl, and 25 mM Tris, pH 8.0. A constant amount of protein was used during peptide titrations (100 nM). Excitation was performed at 295-nm and the emission was detected at 340-nm. Fluorescence values were corrected for sample dilution and resulting data were fit to Equation 1.
Where Fc is observed fluorescence, fE is the fluorescence change, fb is the background fluorescence, [P] is peptide concentration, and Kd is the equilibrium dissociation constant of the peptide-protein complex.
BY4741 wild type and isogenic chd1::Kan mutant and CHD1-PtA yeast strains were grown in YPD to OD=0.6, subjected to serial dilutions and “spotted” onto complete minimal medium (SDC) containing either 0, 750 or 1000 μg/ml of 6AU. Plates were incubated for 3 days at 30º C before imaged.
In an effort to determine the functional consequences of H3K4me3 in active transcription, we sought to identify factors in HeLa nuclear extracts that selectively bind this post-translational histone modification. Towards this end, we generated an affinity column using a peptide encompassing the first eight residues of histone H3 tri-methylated on lysine 4. A number of proteins implicated in transcription post-initiation processes were identified to bind H3K4me3 by a mixture of mass spectrometry and western blot analyses (data not shown). As previously reported, SNF2h derived from nuclear extracts was observed to selectively recognize H3K4me3, but not the unmodified peptide (data not shown) (10). In addition to SNF2h, we identified the chromatin remodeling protein CHD1 as an H3K4me3 binding factor. The association of CHD1 with H3K4me3 was specific, as no association with the unmodified peptide was observed (Fig. 1A). A highly purified fraction of CHD1 also retained the ability to bind H3K4me3 (data not shown).
CHD1 encodes a number of conserved domains, including two tandem chromodomains (Fig. 1B). As the chromodomain has been demonstrated to selectively recognize methylated lysine residues, we hypothesized that CHD1 binds directly to H3K4me3 via its chromodomains. To test for direct binding, we sub-cloned the two chromodomains of human CHD1 into a GST expression vector (Fig. 1B). In vitro binding studies identified that human CHD1 binds directly and selectively to H3K4me3 (Fig. 1C). This binding was specific, as GST did not bind to H3K4me3 (Fig. 1C), and CHD1 failed to recognize H3K9me2 and H4K20me1 (Fig. 1D). Moreover, the SANT domain of SNF2h, a putative histone tail-presenting motif (22), also failed to recognize H3K4me3 (Fig. 1D).
To more quantitatively assess the binding of H3K4me3 to CHD1, we performed fluorimetric binding equilibrium titrations using the intrinsic fluorescence of CHD1. A constant amount of CHD1 protein was titrated with an increasing concentration of H3 peptides. Upon titration of the H3K4me3 peptide, we observed an increase in fluorescence of CHD1 (Fig. 2A). Titration of the H3K4me0 peptide or buffer alone yielded no change in CHD1 fluorescence, suggesting no interaction between CHD1 and the unmodified H3 tail (Fig. 2A). Titration of an unrelated tri-methylated H1K26me3 peptide (residues 20–37 of human H1b) had no effect on CHD1 fluorescence (data not shown). The calculated dissociation constants for CHD1 binding to H3K4me2 and H3K4me3 peptides were approximately 5 μM, while the H3K4me1 peptide was three fold less (~15 μM) (Fig. 2B and 2C).
During the course of our studies, it was reported that yeast Chd1 binds directly to H3K4me2 (9), although previous in vitro binding studies with yeast Chd1 contradict these results (10). Nevertheless, a fragment of yeast Chd1 only containing the second chromodomain was observed to bind methyl H3K4 based on peptide pull-down experiments (9). Based on these observations, the authors suggested that the second chromodomain of Chd1 defines a new sub-class of methyl binding motifs (KWxxLxY)(9). This motif is precisely conserved within the second chromodomain in human CHD1 (Fig. 3); thus, we tested the H3K4me3 binding ability of the individually expressed chromodomains of human CHD1. In contrast to the previously reported findings, we did not observe H3K4me3 association with either of the separately expressed chromodomains from CHD1 (Fig. 4A). Next, we generated point mutations at conserved aromatic residues within the first (CD1) and second (CD2) chromodomains of human CHD1. These residues are important for HP1 recognition of the methyl-lysine within H3K9me2 (18,19) (Fig. 3 and and4B).4B). Surprisingly, when a single aromatic residue in either CD1 or CD2 was mutated, CHD1 failed to bind to H3K4me3 (Fig. 4B). Mutant CHD1 binding to H3K4me3 was also analyzed by fluorimetric titrations. Although the wild type protein bound H3K4me3 as described previously, neither the mutant-CD1 nor mutant-CD2 proteins interacted with the H3K4me3 peptide (Fig. 4C). Collectively, these results are in stark contrast to those reported for yeast Chd1, and suggest that human CHD1 requires both chromodomains to interact with methylated H3K4.
The crystal structure of human CHD1 associated with methylated H3K4 confirms that both chromodomains are required to stabilize this interaction, although methyl-coordination occurs through aromatic residues within the first chromodomain (S. Khorasanizadeh, personal communication). Importantly, comparison of the chromodomain structure of HP1 (18,19) with yeast and human CHD1, indicates that the first chromodomain (CD1) of yeast Chd1 lacks an important aromatic residue essential for methyl-lysine binding by HP1 and human CHD1 (Fig. 3). In the description of yeast Chd1-H3K4me2 binding, a His to Tyr mutation within the first chromodomain restored a putative KWxxLxY motif that facilitated binding of the yeast CD1 to H3K4me2 (9) (Fig. 3, purple residues). The crystal structure of human CHD1 bound to methyl-H3K4 does not support these findings and clearly demonstrates the requirement of a KWxxW motif. These collective findings predict that yeast Chd1 lacks the structural determinants to bind methyl-H3K4. To directly test this prediction, the methyl-H3K4 binding ability of yeast Chd1 was analyzed. Yeast extracts expressing full-length tagged Chd1 or Gcn5 were incubated with peptide-coupled beads. While Gcn5 was observed to bind to each H3 tail regardless of its modification state, Chd1 failed to bind any of the H3-tail peptides, including methylated H3K4 (Fig. 4D). To test if the tagged Chd1 used in Fig. 4D is functional in vivo, we utilized a CHD1 deletion strain that displays resistance to 6-azauracil, a phenotype indicative of deficiencies in transcript elongation (13). Indeed, Chd1-PtA complemented this mutation in a manner identical to the wild type protein (Fig. 4E). Moreover, a fragment of yeast Chd1 encompassing its tandem chromodomains was analyzed for direct binding to methylated H3K4. Human CHD1 strongly recognized H3K4me2 peptides, while in contrast, yeast Chd1 failed to bind this methyl peptide under identical conditions (Fig. 4F). Fluorimetric titration experiments failed to detect any interaction between yeast Chd1 and mono-, di-, or tri-methyl H3K4 peptides, although human CHD1 recognized these methyl peptides under identical conditions (Fig. 2 and and4G).4G). Our results clearly indicate that contrary to human CHD1, yeast Chd1 does not bind methyl-H3K4. Our results are also supported by genetic studies, which demonstrate that yeast Chd1 localization on active genes is unaffected by deletion of Set1, the enzyme that catalyzes H3K4 methylation (J. Mellor, personal communication). Based on the requirement of the KWxxW motif, it is unlikely that CHD3 or CHD4 bind to methylated-lysine residues (Fig. 3). Consistent with this notion, previous studies identified a preference of the NuRD complex, which contains CHD3 and CHD4, for the unmodified H3 tail (23,24).
Why does human CHD1 bind to methyl-H3K4, but yeast Chd1 does not? Illuminating the functional role of the human CHD1/methyl-H3K4 interaction will likely provide some answers, although the differences in the size and architecture of human genes compared to yeast may be a clue. Human genes are far larger and more complex than those found in yeast, and the binding of human CHD1 to methyl H3K4 might provide additional stability in its association, and the association of its interaction partners, with active genes.
We would like to thank Drs. R.P. Perry and D. Stokes for the gift of CHD1 antibodies, R.M. Xu for the gift of the recombinant SANT domain of SNF2h and human CHD1 fragments CD1 and CD2, G. Q. Tang for helpful assistance, K. Cabane for technical assistance, S. Khorasanizadeh and J. Mellor for communicating results prior to publication, and S. Khorasanizadeh for providing the expression vector of yeast Chd1. We thank Drs. P. Trojer, B. Lewis, and S. Gamblin for helpful discussions. These studies were supported by the HHMI (to DR), and by grants from NIH (GM-37120 to DR, GM-71166 to RJS, and GM-51966 to SSP).