Characterization of chromoshadow domain-mediated binding of heterochromatin protein 1α (HP1α) to histone H3.

The chromoshadow domain (CSD) of heterochromatin protein 1 (HP1) was recently shown to contribute to chromatin binding and transcriptional regulation through interaction with histone H3. Here, we demonstrate the structural basis of this interaction for the CSD of HP1α. This mode of H3 binding is dependent on dimerization of the CSD and recognition of a PxVxL-like motif, as for other CSD partners. NMR chemical shift mapping showed that the H3 residues that mediate the CSD interaction occur in and adjacent to the αN helix just within the nucleosome core. Access to the binding region would require some degree of unwrapping of the DNA near the nucleosomal DNA entry/exit site.

Introduction

Heterochromatin protein 1 (HP1) is a highly conserved and abundant non-histone chromosomal protein. It participates in gene silencing and in transcriptional activation (reviewed in Refs. 1–3) and is also implicated in DNA repair (4). Fluorescence recovery after photobleaching measurements on GFP-HP1 show that HP1 binding to chromatin is dynamic and that HP1 is less mobile in heterochromatin than in euchromatin (5–8).

HP1 exists in mammals as three isoforms (α, β, and γ), which have high sequence homology and similar structures. HP1α is found mostly in cytologically dense heterochromatin, whereas HP1β and HP1γ are also found in euchromatin. The isoforms are all homodimers in which the monomeric unit comprises (i) a chromodomain (CD), which specifically recognizes the methylation mark at di- and trimethylated H3K9 (9, 10); (ii) a flexible interdomain hinge region of variable sequence and length, which contains a putative KRK nuclear localization signal and has been shown to bind DNA (11); and (iii) a chromoshadow domain (CSD), which mediates dimerization of HP1 and recruitment of a variety of other nuclear chromatin-modifying proteins (12), many of which contain a PxVxL motif (13, 14). The CD and CSD have high amino acid sequence homology and very similar folds (15, 16). HP1 is generally believed to cause gene repression through chromatin compaction (17, 18), mostly in H3K9-methylated regions (19). However, it also condenses reconstituted chromatin arrays in the absence of H3K9 methylation (20), suggesting an additional methylation-independent structural role of HP1 in chromatin compaction and repression. (Dimerization of CD sticky ends on two adjacent HP1 molecules has been suggested as one possible mechanism (21).)

The CSD appears to interact directly in an H3K9 methylation-independent manner with a region at the edge of the histone H3 globular domain, which contains a PxVxL motif (see Fig. 1A), suggested to be the binding site (22). Phosphorylation of Tyr-41 within this region leads to displacement of HP1α from heterochromatin (23). The structural basis of CSD dimer binding to other proteins containing a PxVxL motif has been determined (24, 25).

FIGURE 1.

Comparison of wild-type and mutant CSDα binding to H3(1–59) and P A, sequence of H3(1–59) showing the exposed and buried regions in the context of the nucleosome and the positions of the αN helix and the PxVxL-like motif. B and C, pulldown of wild-type and mutant CSDα, respectively, with GST fusion peptides immobilized on glutathione-Sepharose beads. I, input with 25% of total proteins; B, bound proteins, left on the beads following washing; M, protein molecular mass markers (broad range, New England Biolabs).

We confirm here a role for PxVxL-mediated recognition of H3 by the CSD of HP1α (CSDα). We demonstrate a requirement for CSD dimerization, and using NMR, we identify the residues that mediate the interaction. The CSD dimer specifically interacts with the H3 residues just within the nucleosome, including (part of) the αN helix, close to the DNA entry/exit sites. This provides a second component to the interaction of HP1α with H3, in addition to the methylation-dependent interaction with di- or trimethylated H3K9 via the CD.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

Human histone H3(1–59) was expressed from plasmid pGEX2TL-H3(1–59), and human HP1 CSDα (109–185) was expressed from pGEX2T-HP1CSDα(109–185) (a gift from Dr. A. Bannister). GST-H4(1–48) was produced as described by Verreault et al. (26), GST-CAF1 p150(176–327) as described by Murzina et al. (13), and CAF1 p150(210–238) as described by Thiru et al. (24). CSDα point mutations (I165E, W174A, and E179X) and a stop codon at residue 60 in histone H3 (to generate pGEXTL-H3(1–59) from pGEXTL-H3(1–135) (26)) were introduced using a QuikChange II site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions.

All proteins were expressed as GST fusions in Escherichia coli Tuner(DE3)pLacI cells (Novagen) in 2× TY (Tryptone and yeast extract) medium at 20 °C. 15N-Labeled H3(1–59) was produced in M9 minimal medium, and uniform 15N labeling was achieved using [15N]NH4Cl as the sole nitrogen source for bacterial growth. 15N-Labeled CSDα(109–185) was expressed in MOPS minimal medium containing [15N]NH4Cl and 0.25 g/liter 15N-labeled ISOGRO® powder (Sigma-Aldrich) as nitrogen source. The proteins were affinity-purified on glutathione-Sepharose beads (GE Healthcare) using standard protocols. The GST tag was left intact for in vitro binding (pulldown) assays, but for NMR experiments, it was removed with thrombin (Novagen). 15N-Labeled H3(1–59) was further purified on a Mono S column (GE Healthcare). 15N-Labeled CSDα(109–185) was purified on Mono Q and then Superdex 75 gel filtration columns (GE Healthcare).

GST Pulldown Assays

Approximately 5 μg of protein was mixed with a GST fusion protein immobilized on glutathione-Sepharose beads in 250 μl of binding buffer (150 mm NaCl, 20 mm Tris-HCl (pH 7.5), 5 mm DTT, 0.25% Nonidet P-40, and 1× complete protease inhibitor mixture (Roche Applied Science)). 25% of the mixture was kept separately as the input sample. After incubation at 4 °C for 2 h, the beads were washed four times with 0.7 ml of binding buffer. Bound proteins were eluted from the beads with SDS-PAGE loading buffer (Invitrogen), separated on 12% NuPAGE (Invitrogen) precast gels, and stained with Instant Blue (Expedeon Protein Solutions).

NMR Spectroscopy

NMR measurements were made on 15N-labeled proteins (0.2–0.5 mm) in 10% D2O, 10 mm sodium phosphate (pH 7.0 for 15N-labeled H3(1–59) and pH 8.0 for 15N-labeled CSDα(109–185)), 1 mm EDTA, 1 mm DTT, 1× protease inhibitor mixture (0.05 mg each of leupeptin, aprotinin, and pepstatin A and 78 mg of benzamidine per 10 ml for a 50× stock), and 100 μg/ml 1-chloro-3-tosylamido-7-amino-2-heptanone. Experiments were recorded at 25 °C (for 15N-labeled CSDα) or 0 °C (for 15N-labeled H3) on a Bruker DRX600 spectrometer. Data were processed using the Azara suite of programs (version 2.8). Backbone assignments were made in three-dimensional NOESY-15N heteronuclear single quantum coherence (HSQC) and total correlation spectroscopy (TOCSY)-15N HSQC experiments (27) using CcpNmr Analysis version 2.1 (28). Chemical shift differences were calculated using Δδ = ((ΔδH)2 + (0.15 × ΔδN)2)1/2 (29).

Fluorescence Spectroscopy

The binding of increasing concentrations of the CAF1 peptide to 300 nm CSDα dimer in 10 mm sodium phosphate (pH 8.0) was measured using fluorescence emission at 350 nm (PerkinElmer LS55 spectrofluorometer) with an excitation wavelength of 280 nm, at 25 ± 0.1 °C.

Affinity Measurements

K values were calculated from fitting the NMR and fluorescence data to Δ = Δ∞(−([CSDα] + [x] + K) − √([CSDα] + [x] + K)2 − (4[CSDα][x]))/2[CSDα], where x denotes either the H3 or CAF1 peptide, and Δ is either the chemical shift difference or the change in intrinsic tryptophan fluorescence on addition of H3 or CAF1, respectively. (Representative fits are shown in supplemental Fig. 1.) The reported K was obtained by averaging the fitted values from three different residues (Ile-113, Gly-116, and Asp-149) for NMR or over three repeat measurements for fluorescence. In fitting the NMR data, [CSDα] (116 μm dimer) and [H3] (23–464 μm) were adjusted for dilution effects; dilution had a minimal effect on the CSDα intrinsic fluorescence, as the final volume increase was below 5%.

RESULTS

Dimerization of CSDα Is Required for H3(1–59) Interaction

The reported interaction between CSDα and H3 was demonstrated using a pulldown assay with a GST-tagged H3 peptide (residues 1–59) (Fig. 1A) containing the part of the H3 tail that protrudes from the nucleosome (residues 1–38), followed by the residues (residues 39–43) that pass between the DNA gyres and then the αN helix (residues 45–56), which is located within the nucleosome core (30). This region of H3 contains a variant of the PxVxL motif, 43PGTVAL48. The binding of CSDα to GST-H3(1–59) and CAF1 p150(176–327), which contains a well characterized PxVxL motif (13), was compared in a pulldown assay. The result (Fig. 1B) confirms the reported interaction of H3 with CSDα and indicates that H3 interacts less strongly than CAF1 (compare lanes 8 and 14); an H4 peptide containing residues 1–48 did not interact in this experiment (lane 11).

The role of the H3 PxVxL-like motif in binding to CSDα was investigated further by mutagenesis of key residues, identified based on those involved in the well characterized interaction of the CSD of HP1β (CSDβ) with the CAF1 p150 PxVxL peptide. This binds to the CSD across the dimer interface (13, 16, 24); the dimeric structure is also required for the interaction of HP1 with H3 because mutation of Ile-165, a key dimerization residue in CSDα, to glutamic acid (mutant I165E) abolished the interaction with H3(1–59) (Fig. 1C). Trp-170 has previously been shown to be critical for the CSDβ interaction with PxVxL motif-containing proteins (16, 24), and the corresponding residue, Trp-174, is important in the interaction of CSDα with GST-H3(1–59) because the interaction was abolished in mutant W174A (Fig. 1C). These results therefore suggest a role in CSDα recognition for the PxVxL-like motif in H3(1–59). In addition, a CSDα C-terminal tail deletion mutant, E179X, in which Glu-179 was replaced with a stop codon, had no effect on the interaction of CSDα with GST-CAF1 p150(176–327), whereas it significantly reduced the interaction with GST-H3(1–59) (Fig. 1C), indicating that the HP1α unstructured C-terminal tail plays a role in the interaction with H3.

CSDα Interacts Similarly with H3(1–59) and a PxVxL-containing Peptide

As a prelude to NMR chemical shift perturbation mapping to identify the CSDα(109–185) residues involved in the interaction with H3(1–59), we assigned the 1HN and 15N amide resonances of 15N-labeled CSDα using three-dimensional NOESY-15N HSQC and TOCSY-15N HSQC experiments (27). All but four of the non-proline residues could be assigned in the structured region (residues 114–174); residues at the N and C termini were often broad and/or overlapped such that only very few could be assigned unambiguously. 15N HSQC spectra of 15N-labeled CSDα alone and in the presence of H3(1–59) at H3(1–59):15N-labeled CSDα dimer molar ratios of 0.6:1, 1:1, and 4:1 are shown in Fig. 2A. Some peaks did not shift upon addition of H3(1–59) (e.g. Gly-120 and Cys-160) as shown in expansions of the spectra in Fig. 2B. Others (e.g. Asp-149 and Val-151) showed significant progressive shifts, indicative of fast exchange on the chemical shift time scale. Several others shifted to a larger extent and became increasingly broad as the resonances approached the intermediate chemical exchange regime. For many of this latter set (e.g. Trp-174, Ser-132, and Gly-134), the final peak is split into two separately identifiable resonances, indicative of a breakdown in the symmetry of the CSDα dimer on binding the H3 peptide, as was seen previously in studies of the complex of the CSDβ dimer with the CAF1 peptide containing a PxVxL motif (24). For comparison, 15N HSQC experiments with CSDα in the presence of the CAF1 peptide were also recorded at a range of molar ratios (Fig. 2C). Comparison of the CSDα/H3 and CSDα/CAF1 perturbations revealed that the same CSD resonances responded to both CAF1 and H3. The main difference was that slightly larger shifts were seen for the CAF1 peptide, and saturation was achieved at a lower ratio. This suggests that the H3 and CAF1 peptides recognize the same binding interface on CSDα, but H3 has a significantly lower affinity.

FIGURE 2.

Comparison of interactions of CSDα with H3(1–59) and CAF1 p150(210–238) peptides by NMR. A, 15N HSQC spectra of CSDα (black) titrated with H3(1–59) at the input ratios of CSD dimer to peptide monomer indicated. B, expanded views of peaks that exemplify different behavior: unperturbed (panel i), shift (panel ii), shift and split (panels iii–v), and disappearance (panel vi), on addition of H3(1–59). C, spectra of 15N-labeled CSDα titrated with the CAF1 peptide showing a broadly similar response for comparison.

An average K for the H3/CSDα interaction of 58 ± 7 μm was obtained from the NMR data using the resonances that were in fast exchange (supplemental Fig. 1). This was not possible for CAF1 because most peaks showed signs of broadening due to intermediate exchange. However, a K of 0.43 ± 0.10 μm was obtained from intrinsic tryptophan fluorescence measurements (supplemental Fig. 1). (Isothermal titration calorimetry proved unsuitable due to very low enthalpies for the CAF1/CSDα interaction.)

The perturbations in the NMR spectra of CSDα/CAF1 closely resembled those recorded for CSDβ in the presence of the CAF1 peptide (24), suggesting identical interaction of the α- and β-isoforms with CAF1 p150. In many cases, peak doubling or peak broadening/disappearance prevented the precise quantification of chemical shift changes. Therefore, a qualitative assessment of each peak was made for both the CAF1 and H3(1–59) titrations, and where possible, peaks were classed according to one of four specific behaviors: no shift, small shift, large shift and/or doubling, and disappearance/ambiguous. The peak behavior for each residue of CSDα upon titration with either H3(1–59) or the CAF1 peptide was mapped onto the structure of highly homologous CSDβ, as the only available CSDα structure (CBX5; Protein Data Bank code 3I3C) lacks the peptide substrate, and the interacting region is disordered (Fig. 3). Where it was possible to assign peaks and their shifts, it was clear that binding of H3(1–59) to CSDα produced a pattern of peak shifts/peak doubling similar to that for binding of the CAF1 peptide, confirming that the binding interfaces for H3 and CAF1 on CSDα overlap. A few residues farther away from the PxVxL-binding site were affected by CAF1 binding, but not by H3 binding.

FIGURE 3.

Comparison of H3(1–59)- and CAF1 p150(210–238)-binding sites on CSD. A, qualitative comparison of the H3(1–59)- and CAF1 peptide-binding sites mapped onto a structure of the homologous CSDβ dimer-CAF1 peptide complex (Protein Data Bank code 1S4Z (24)), with CSDβ in purple ribbons and CAF1 in gray. Residues displaying a large chemical shift change in both the H3(1–59) and CAF1 complexes are shown in red (space filling). Residues that shift to a lesser extent in both complexes are shown in yellow; residues that shift with the CAF1 peptide only are shown in green. B, same as A after a 90° rotation about the horizontal axis. Structural representations were prepared using MolScript (31) and Raster3D (32).

Residues 36–58 of H3(1–59) Interact with CSDα

To map precisely the binding site for CSDα on H3(1–59) using 15N HSQC chemical shift perturbation experiments, all non-proline 1HN and 15N amide resonances of 15N-labeled H3(1–59) were first assigned using three-dimensional NOESY-15N HSQC and TOCSY-15N HSQC experiments (27). Elevated {1H}15N heteronuclear NOE values measured for free H3(1–59) indicated that residues 39–55 are partially structured in the free peptide (Fig. 4C). These residues coincide with the αN helix seen in the context of the nucleosome, but also extend ∼6 residues to the N-terminal side of this region. Spectra of 15N-labeled H3(1–59) alone, and in the presence of CSDα at molar ratios of 0.5:1, 1:1, and 3.6:1 CSDα dimer:15N-labeled H3(1–59), are shown in Fig. 4A. Although some perturbed residues were in the fast exchange regime and therefore trackable by their progressive changes in chemical shift, the peaks corresponding to residues 37–56 disappeared entirely upon the first addition of CSDα (0.5:1 molar ratio) and did not reappear with further addition, even at a 3.6-fold excess of the CSDα dimer to the H3 peptide. Expanded views (Fig. 4B) highlight residues that disappeared entirely (e.g. Arg-42) or shifted progressively (e.g. Lys-36 and Thr-58) on addition of CSDα. The 1H and 15N chemical shift differences were scaled appropriately, combined (as Δδ) (29), and plotted in Fig. 4C. This shows that residues that disappeared, shown as “quenched peaks,” are flanked on either side by the residues that show the largest trackable perturbations, suggesting that the whole of the region comprising H3(36–58) is involved in the interaction with CSDα. Identical results were obtained when analogous titrations were performed with CSDβ and 15N-labeled H3(1–59) (data not shown). The interacting region of H3 (residues 36–58) encompasses the segment that exits the nucleosome through the aligned minor grooves of two DNA gyres and the adjacent αN helix (30) and includes Tyr-41, a residue shown to regulate HP1α binding to heterochromatin (23). The stretch of residues (positions 36–58) forming the interaction interface is shown in Fig. 5 mapped onto a ribbon diagram of the nucleosome (30). There are no significant perturbations elsewhere in the H3(1–59) sequence.

FIGURE 4.

Identification of CSDα-binding site on H3(1–59). A, 15N HSQC spectra of 15N-labeled H3(1–59) alone and in the presence of CSDα at varying input ratios as indicated. B, expansions of spectra in A showing examples of different responses (peak disappearance (panel i) and peak shift (panels ii and iii)). C, chemical shift perturbation map showing the residues in the interacting region, with significantly shifting peaks (yellow) and peaks that broaden and disappear (red asterisks) highlighted. The H3(1–59) amino acid sequence is indicated below the x axis; residues known to make contacts with DNA (30) are shown in light blue. 15N{1H} heteronuclear NOE values are shown above to indicate the residues (positions 39–55) that display reduced flexibility/secondary structure in the free peptide; residues 45–56 (α-helix) constitute the αN helix in nucleosomal H3.

FIGURE 5.

CSDα-binding site on H3(1–59) shown in the context of the nucleosome. A, space-filling representation of the H3 residues that are significantly shifted (yellow) or “quenched” (red) on addition of CSDα, mapped onto the nucleosome (Protein Data Bank code 1AOI (30)). Histone-fold helices of H3 are shown as blue ribbons, and H3Y41 is indicated in green space-filling representation. B, same as A after a 90° rotation about the vertical axis. C, bifunctional binding of an HP1 dimer to a single nucleosome, shown as in A except that the entering/exiting DNA strand lying over the CSDα-binding site is shown “unwrapped” to illustrate how conformational changes resulting from nucleosome dynamics may expose these residues. HP1 is shown schematically in gray with the CSD (Protein Data Bank code 1S4Z (24)) in space-filling representation. The H3 tail is extended beyond the density detected in the x-ray structure to include the methylated Lys-9 site (K9Me), which recruits the CD. Structural representations were prepared using MolScript (31) and Raster3D (32).

DISCUSSION

Our results confirm and extend previous reports (22, 23) that CSDα plays a role in histone H3 recognition. This interaction is distinct from the well studied CD-mediated interaction of HP1 with methylated H3K9. We have shown by NMR and CSDα mutagenesis that H3(1–59) interacts with CSDα through a run of residues, H3(36–58), containing a PxVxL-related motif, 43PGTVAL48, in a manner very similar to that of CAF1 p150, a known PxVxL-containing family member (24). Mutation of Val-46 in this motif to alanine significantly decreases the strength of the interaction with HP1α (22). Structural studies of the interaction of the highly homologous CSDβ with the CAF1 PxVxL peptide suggest that an exact consensus is not required for interaction (24, 25). The deviation from the consensus in the H3 PxVxL-like sequence probably accounts for the reduced affinity of the H3 peptide (K 58 μm) relative to the CAF1 peptide (K 0.43 μm). The CSDα interaction with H3(36–58) provides a rationale for the in vivo observations that phosphorylation of Tyr-41 by JAK2 specifically displaces HP1α from chromatin and that HP1α binding to chromatin can be competed in permeabilized cells by the H3(31–56) peptide (23).

The region of H3 responsible for binding to CSDα, comprising residues 36–58, includes the αN helix, which is located near the nucleosomal DNA exit and entry point, and is an important regulator of nucleosome dynamics and stability (33). In the conventional representation of the static nucleosome, H3 residues 36–58 would be relatively inaccessible and confined within the DNA gyres (Fig. 5, A and B). Although a window of opportunity for HP1 binding in vivo may, of course, arise during DNA replication (34) or transcription, nucleosomes can also undergo spontaneous conformational changes that result in transient unwrapping of the DNA from the octamer surface (Fig. 5C) (35–37). Unwrapping may also be caused by the activity of remodeling complexes (38), and remodeling complexes (Brg1 (22) and ACF1 (39)) have been reported to facilitate HP1 loading onto chromatin. In addition, unwrapping might result from the action of a protein such as HMGB1, which binds to linker DNA adjacent to the nucleosome (and interacts with the H3 tail (40) and data not shown) and could cause unwrapping of DNA by torsion transmitted from bending of the linker DNA (41). Moreover, the αN helical region of H3 in Mg2+-condensed chromatin has been shown by hydrogen/deuterium exchange studies to be relatively exposed (42), possibly as a result of structural rearrangement on chromatin folding. In contrast, residues 1–37, which are relatively exposed and flexible in soluble arrays, are protected to the same degree as the folded H3 “core” within the H3/H4 tetramer, suggesting that the N-terminal tail (and the H3K9 CD-binding site) is buried in the folded structure. Thus, at least in this form of condensed chromatin, the αN helical region might remain accessible to CSD-mediated interaction with HP1. The DNA that is unpeeled to allow access could then be captured by the basic interdomain HP1 linker, thereby stabilizing the complex. HP1 could thus prevent further spontaneous unwrapping while also preventing access of remodeling complexes (22, 38).

We suggest that HP1 contributes to repressive condensed chromatin structure in at least two ways: 1) CD tethering through methylated H3K9 with consequent recruitment of silencing complexes and 2) CSD binding to the H3 αN helix.