Structural insights into the interaction and activation of histone deacetylase 3 by nuclear receptor corepressors

Results

The Structure of the DAD. The region of HsSMRT encompassing the DAD (amino acids 412-480) was expressed in bacteria as a GST fusion protein, purified, and cleaved, leaving the DAD preceded by a nonnative glycine-serine dipeptide. NMR spectra of a longer protein (amino acids 389-480) were essentially identical to those of the shorter protein (additional signals were observed only in the random coil region of the spectra), suggesting that the additional residues do not adopt a stable structure. Interestingly, some preparations of the longer construct were proteolyzed during sample preparation such that the resulting fragment began with residue V422, suggesting that it is the N-terminal boundary of a structured domain. This finding was consistent with the observation that residues 412-419 have no long- or medium-range close contacts between protons as indicated by the absence of corresponding NOE interactions in NOESY spectra.

The NMR data clearly revealed four helices within the DAD, as indicated by the pattern NOE connectivities spanning helical turns, by predictions using the program talos (29), and by characteristic chemical shift differences (30, 31); details appear in Supporting Text. Structures were calculated by using 852 NOE-based restraints, and ensembles were calculated both with and without additional restraints for predicted hydrogen bonds within the helices; the latter calculations gave a larger proportion of converged structures but with somewhat higher energies. Because the only other effect of including the hydrogen bond restraints was a slight regularization of the helices, we chose to deposit only the ensemble calculated without their use, based wholly on experimentally derived restraints. A summary of the restraints and structural statistics is provided in Tables 1 and 2, and Supporting Text includes energy-ordered rmsd profiles and total energies for the calculated structures.

A superposition of the final ensemble of 28 convergent structures (calculated without using hydrogen bond restraints) is shown in Fig. 1, with a schematic diagram of the lowest energy structure. As expected, the eight amino-terminal residues show no ordered structure. The structured domain comprises four α-helices (H0-H3) packed around a largely hydrophobic central core. Because the C terminus of helix H3 protrudes somewhat from the structure it participates in relatively few long-range NOEs interactions, which in turn causes this part of the structure to be somewhat less well ordered; the backbone rmsd of residues 420-480 is 0.76 Å, but this figure improves to 0.40 Å if only residues 420-470 are considered. It may also be that there is mobility within the C terminus of helix 3 (see the rmsd plot in Fig. 6, which is published as supporting information on the PNAS web site); interestingly the loop between helices H1 and H2 is close to the C terminus of helix H3 and also shows some disorder.

Comparison of the DAD with the MYB and SANT Domains. As was predicted from sequence comparisons, the three C-terminal helices (H1, H2, and H3) pack together so as to form the characteristic architecture typical of the DNA-binding domain from MYB and related DNA-binding proteins (32). This domain architecture is also shared by the non-DNA-binding SANT domain found in components of various chromatin remodeling complexes (21, 22).

The DAD differs from both the MYB (33) and SANT (34) domains in that it has an additional helix (H0) that packs into the groove between the N terminus of helix H1 and the loop between helices H2 and H3 (Fig. 2). The helix H0 is defined by short- and medium-range NOEs (and its helical character fits well with the proton alpha chemical shift and talos secondary structure predictions). The positioning of this helix with respect to the rest of the domain is defined by long-range NOE interactions linking M420, K421, and V422 to residues at the C terminus of helix H2.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Comparison of the DAD (red) with Myb (green) (33) and SANT (blue) domains (34). The DAD has an additional amino-terminal helix (H0) that packs between the amino terminus of helix H1 and the loop between helices H2 and H3. Note that although helices H1-H3 have a similar arrangement to the MYB and SANT domains, the orientation of helix H3 in the DAD differs significantly. The alignment shown indicates the Cα atoms matched in the superposition (indicated by a boxed outline) and the rmsd between those matched Cα atoms.

Superposition of the DAD with MYB and SANT domains shows that there is a variation in the relative orientations of the three core helices (H1, H2, and H3). In particular, the orientation of helix H3 in the DAD differs from the other two domains such that there is a significantly wider angle between the amino terminus of helix H1 and the carboxyl terminus of helix H3 (these helices are almost perpendicular in the MYB and SANT domains). Given the low sequence homology, it is not straightforward to assign the cause of this difference. However, the consequence of this widening is that it produces a very noticeable groove or cleft between the C-terminal half of helix H3 and the loop between helices H1 and H2. Because both of these regions exhibit some disorder, the width of the resulting groove is somewhat ill defined, but is ≈7 Å. Such a feature is not observed in either the MYB or SANT structures.

Although the DAD shares common aspects of protein architecture with both the MYB and SANT domains, a more detailed analysis of the protein shows that the domains have a very different surface character. Fig. 3 shows a comparison of the surfaces of the MYB, DAD, and SANT domains. When viewed toward the DNA-binding surface, the MYB domain is notably very basic. The equivalent view of the SANT domain reveals a rather acidic surface. The DAD is more mixed in character, with both basic and acidic regions on its surface. Interestingly, the groove between helix H3 and the H1/H2 loop does not itself have any polar side chains, but acquires a basic character caused by surrounding lysine residues.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Comparison of the architecture (Top), electrostatic (Middle), and hydrophobic (Bottom) potential surfaces of the MYB, DAD, and SANT domains. Note the basic character of the DNA-binding surface of the MYB domain and the rather acidic nature of the corresponding surface of the SANT domain. The DAD is also rather basic. Importantly, the wider angle of helix H3 in the DAD results in a groove between the amino terminus of helix H3 and the loop between helices H1 and H2. This groove does not itself have any polar side chains, but acquires a basic charge caused by surrounding lysine residues. The hydrophobic potential of the surface [calculated by using grid (35)] shows that whereas both the MYB and SANT domains have mostly polar surfaces, the groove in the DAD has a strikingly high hydrophobic potential, suggesting that it might mediate interactions with a nonpolar partner (36).

To compare further the character of the DAD, MYB, and SANT domains, we examined the domain surfaces for regions of favorable hydrophobic contact by using the program grid (35). This approach has been useful in predicting nonpolar interaction surfaces on proteins and has been termed mapping “hydrophobic potential” (36). The surfaces of both the SANT and MYB domains are both largely polar with no notable regions of high hydrophobic potential. In contrast, the DAD has a region with very high hydrophobic potential. In particular, the groove discussed above is strikingly nonpolar in character. The floor of the groove is formed from F451 and Y469, the walls are formed from F444 on one side and V467, Y470 on the other, and the back of the groove is formed from R441, M445, and Y479. The side chain of Y471 also creates a hydrophobic patch on the surface of the domain adjacent to the groove.

Interaction and Activation of HDAC3. Because the biological role of the DAD is to interact with and activate HDAC3, we wanted to explore what insight the DAD structure might provide into these functions. Previous mutagenesis studies have identified a number of residues that, when mutated to alanine, perturb the binding and activation of HDAC3 (18). Several of these (W432, F440, and L458) are completely buried within the hydrophobic core of the domain (Fig. 4A) and probably perturb the structure of the whole domain. Two residues (F451 and Y470) are at least partly exposed on the surface of the domain and form the floor and one wall of the hydrophobic groove. These results suggest that the groove may represent part of the interaction surface with HDAC3. To confirm and further explore this possibility we made a number of additional mutations in this surface. We identified an additional three residues whose mutation significantly perturbs interaction with and activation of HDAC3 by the DAD. Of these, residues F444 and Y479 form part of the walls of the groove and Y471 forms a nonpolar patch on the surface of the DAD adjacent to the groove (Fig. 4). These findings strongly implicate the groove as being an interaction surface for HDAC3. Even more striking is the finding that mutation of K449 to alanine does not greatly influence the interaction with HDAC3, but abolishes activation of the enzyme. This separation of interaction and activation suggests that the two aspects are separable and supports the role of this region of the DAD structure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The interaction between SMRT DAD and HDAC3. (A) Effect of DAD mutations on the activation of HDAC3. The gray bars indicate the percentage of HDAC3 activity in a cotransfection assay. The numbers on the bars indicate the percentage of solvent exposure of that residue based on the lowest energy NMR structure. (B) Coimmunoprecipitation assay to monitor interaction between HDAC3 and the DAD mutants. Note that the interaction with HDAC3 mirrors the activation profile with the exception of the K449A mutation, which abolishes activation yet retains the ability to interact with HDAC3, showing that the two aspects are clearly separable. (C) Location of residues that perturb HDAC3 activity >50% (red) or <50% (green) in the structure of the DAD. Note that the residues that reduce HDAC3 activity cluster together on the surface of the DAD in the vicinity of a nonpolar groove.

Given that another unique feature of the DAD (compared with MYB and SANT domains) is the additional N-terminal helix H0, we used a analogous approach to determine whether this region was important for interaction with HDAC3. Significantly, single mutations in this region (V422 and R426) appeared to have no effect. However, destabilization of this helix through the mutation to proline of two adjacent residues (D425 and R426) in the middle of the helix did abolish HDAC3 interaction and activation. These results suggest that either helix H0 plays a direct role in HDAC3 interaction or it is required for the structural integrity of the whole domain.

Point mutation of conserved residues in the disordered N terminus of the DAD (L415 and D418) did not appear to perturb either interaction with or activation of HDAC3. This result was somewhat surprising because mapping studies indicate that the N-terminal region of the domain is critical for interaction with HDAC3. Indeed, truncation of just eight residues from the amino terminus of the construct used for this NMR study was found to be enough to abolish interaction and activation of HDAC3. This finding fits with the high degree of sequence conservation in this region of the DAD from diverse species, but seems surprising given that these residues clearly do not adopt an ordered structure in the isolated domain. It seems likely therefore that these residues become ordered on interaction with HDAC3.

Discussion

The specific targeting of HDAC complexes to particular genes is emerging as an important mechanism of gene regulation (37). To avoid inappropriate, nontargeted deacetylation of histones it is important that the enzymatic activity of HDACs is carefully regulated. For example, both HDAC1 and HDAC3 are activated as a result of assembly into targeted repression complexes (the NuRD and SMRT/NCoR complexes, respectively) (18, 19, 38). Other deacetylases, such as HDAC4, have been found to be apparently inactive (39), but it seems possible that there may also be specific activating mechanisms for these enzymes.

As a step toward understanding the mechanism of activation of HDAC3 by the SMRT-DAD, we have determined the solution structure of the DAD and mapped the residues required for interaction and activation of HDAC3. As expected, the core architecture of this domain is a three-helix triangle very similar to that found in the related MYB and SANT domains. However, there are significant and important differences in the DAD that are related to the function of the domain. First, there is an additional N-terminal helix, and second, a different relative orientation of the three core helices results in the formation of a nonpolar, surface groove not present in the MYB and SANT domains.

The conservation of the 16-residue DAD-specific motif (N-terminal to the SANT-like domain) suggests that this region plays an important role in the function of the DAD. This conclusion is supported by the finding that deletion of the first 8 aa of this motif is sufficient to abolish interaction and activation of HDAC3. It was therefore surprising that these 8 aa show no ordered structure in solution. It would seem likely that these residues might play a role in interaction with HDAC3, perhaps through adopting an ordered conformation against an interaction surface on HDAC3. Somewhat surprisingly, mutation of L415 and D418 did not appear to perturb interaction and activation of HDAC3, suggesting that these side chains are not essential for the interaction.

The second portion of the DAD-specific motif forms a unique fourth helix within the DAD (here termed H0) that lies in the groove between helix H1 and the loop between helices H2 and H3. The role of this additional helix is not entirely clear because mutation of V422 (largely buried) or R426 (solvent exposed) does not affect interaction and activation of HDAC3. However, a double mutation of D425 and R426 to proline residues, designed to disrupt the helical secondary structure of H0, abolished interaction and activation of HDAC3. This finding suggests either that this helix plays a structural role in the stability of the DAD or alternatively that it plays a part in the interaction with HDAC3, but in the latter case it does not involve the side chains of V422 or R426.

A most striking feature of the DAD structure is the highly hydrophobic groove between helix H3 and the loop between helices H1 and H2. A number of residues in and around this groove perturb activation of HDAC3, delimiting a clear interaction surface.

Overall, it would seem likely that HDAC3 interacts with the DAD through two different interfaces: the DAD-specific motif and the hydrophobic groove. Probably the first residues of the DAD-specific motif (unfolded in the solution structure of the DAD alone) become structured upon binding, perhaps extending helix H0. If so, the unfolded character of this region might confer cooperativity to the interaction. It is interesting to note that the hydrophobic groove in the DAD is suggestive of a peptide interaction surface that would imply that a loop from HDAC3 might interact with this region of the DAD.

Importantly, there is ample evidence that the HDAC3-DAD assembly is a complex process requiring the TriC chaperone complex (20). This phenomenon is analogous to the MTA2-HDAC1 interaction, which requires cotranslation of the two proteins (38). However, the structure of the DAD does not itself explain why assembly with HDAC3 should not be a simple protein-protein association such as rigid docking or induced fit. A potential explanation is that HDAC3 forms homooligomers and heterooligomers (40) and it may be that these need to be actively dissociated by the chaperone before forming an active complex with the DAD. Alternatively, if both the DAD-specific motif and the hydrophobic groove contribute to the interaction, then the whole interface with HDAC3 might be quite large and require a chaperone-mediated folding event in HDAC3 for the formation of the full interface.

The finding that mutation of K449 to alanine results in a DAD that can still interact with HDAC3, but fails to activate, suggests that these two processes are separable and that simply binding to the DAD is insufficient to activate HDAC3. This observation is important and warrants further investigation in the context of the full-length corepressor, because it implies that this K449 might have a specific role in activating the enzymatic mechanism of HDAC3. Perhaps significantly this residue is invariant in all known DADs.

Interestingly, a search of translated genomes suggests that the DAD may be conserved in plants, fungi, and fission yeast (Fig. 5), although it remains unclear whether these proteins are bona fide corepressors involved in the activation of HDAC activity. However, class 1 deacetylases are ancient enzymes present in all of these species (41), and it is notable that most of the residues required for interaction and activation of HDAC3 are conserved.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Alignment of DADs from diverse species. Identical residues are shaded gray. Similar residues are boxed. Sequences from human (Hs), mouse (Mm), rat (Rn), Xenopus (Xl), zebrafish (Dr), and Drosophila (Dm) are known to be corepressor homologues. Other sequences are only putative homologues of the DAD. Mutation of residues shaded in green does not perturb HDAC3 activation. Mutation of residues shaded in red results in a >50% loss in HDAC3 activation. Note that although the eight amino-terminal residues do not adopt a stable structure and are proteolytically sensitive, deletion studies indicate that they play a role in the interaction and activation of HDAC3.

The structure and functional analysis of the DAD provides insight into the mode of interaction and activation of HDAC3. Further studies to understand the specificity of the interaction with HDAC3 will be important because it is emerging that specific targeting of the different class 1 enzymes plays a major role in substrate selection. Understanding these HDAC interactions may facilitate the design of inhibitors that target specific deacetylases.