HDAC6–p97/VCP controlled polyubiquitin chain turnover

In order to better characterize HDAC6 ZnF‐UBP domain, we first selected a portion of the HDAC6 C‐terminal region spanning residues 1000–1149, which includes the entire USP homology region (Figure 1A, see also Figure 2B). The corresponding encoding sequence was cloned in an expression vector and the protein expressed and purified (data not shown).

Gel filtration (Pharmacia Superdex 75) analysis showed that the ZnF‐UBP domain eluted at a volume of 11.9 ml compatible with the ZnF‐UBP domain being an elongated monomer or forming dimers. ZnF‐UBP and ubiquitin co‐eluted at 11.4 ml, suggesting that both components associate at an equimolar ratio (Figure 1B). Sedimentation equilibrium experiments of ubiquitin, HDAC6 ZnF‐UBP domain and HDAC6 ZnF‐UBP domain/ubiquitin complex yielded masses of 11.3, 17.5 and 28.3 kDa, respectively, which is in excellent agreement with their molecular weights (His‐ubiquitin: 11.5 kDa; HDAC6 ZnF‐UBP domain: 17.0 kDa; HDAC6 ZnF‐UBP domain/ubiquitin complex: 28.5 kDa) (Figure 1C). Similar results were obtained in sedimentation velocity experiments (data not shown). Our results demonstrate that the ZnF‐UBP domain and ubiquitin are monomeric in solution and that they associate in a 1:1 complex.

The binding of ubiquitin to the HDAC6 ZnF‐UBP domain was studied quantitatively using isothermal titration calorimetry (ITC; Figure 1D). The affinity of the HDAC6 ZnF‐UBP domain to ubiquitin is high with a calculated equilibrium constant (K_D) of 60 nM and a stoichiometry of 1:1 (N=1.071+0.004). The ΔH for the binding of ubiquitin to the HDAC6 ZnF‐UBP domain is ΔH=−17.9+0.1 kcal/mol, suggesting a substantial number of non‐covalent interactions between both partners. In comparison, the USP5 ZnF‐UBP domain binds ubiquitin with a K_D of 3 μM (Reyes‐Turcu et al, 2006), which roughly falls into the range of affinities for ubiquitin observed for other known ubiquitin‐binding domains with K_D values ranging from 10 to 500 μM (Hicke et al, 2005). Interestingly, these affinities are substantially weaker than the affinity observed for the HDAC6 ZnF‐UBP domain, which binds with the highest known affinity to ubiquitin.

In total, our construct contains 11 cysteine and 10 histidine residues. The large number of conserved cysteine and histidine residues had led to the classification of this domain as a Zn‐finger‐containing domain in the Conserved Domain Database (Marchler‐Bauer et al, 2005). We used micro‐focussed beam PIXE to determine the exact number of Zn atoms present in the HDAC6 ZnF‐UBP domain. X‐ray emission of the Zn atoms was calibrated against the relative concentration of sulphur atoms present in the HDAC6 ZnF‐UBP domain (11 cysteines+5 methionines). Several independent measurements yielded a total of 2.6 (+0.3) Zn atoms, corresponding to either two fully occupied and one partially occupied Zn‐binding sites or, alternatively, to three partially occupied sites (Supplementary Table 1). We then analysed the coordination of the three Zn atoms using EXAFS (Figure 2A). EXAFS data analysis revealed an average Zn coordination by 2.9 (±0.3) cysteines and 1.1 (±0.3) histidines.

The presence of three Zn ions in the structure of the HDAC6 ZnF‐UBP domain coordinated by conserved cysteine and histidine residues is supported by the crystal structure of the USP5 ZnF‐UBP domain (Reyes‐Turcu et al, 2006). Structural alignment of 14 different ZnF‐UBP domains with the USP5 ZnF‐UBP domain as a reference reveals conservation of eight cysteines and four histidines in most ZnF‐UBP domains, but not in the USP5 ZnF‐UBP domain, where eight Cys/His residues are poorly conserved (Figure 2B). Four of these 12 conserved Cys/His residues coordinate one Zn atom in a 3Cys/1His coordination (Zn1) in the USP5 ZnF‐UBP domain structure. The other eight Cys/His residues, not conserved in USP5, cluster in two regions with 3Cys/1His and 2Cys/2His coordination and tetrahedral geometry (Figure 2C). In the case of HDAC6 ZnF‐UBP, the two additional Zn ions are predicted to stabilize the overall structure of the domain. In particular, Zn site 2 connects residues at the N‐ and C‐terminal end of the domain and therefore might be particularly important for stabilization.

To further confirm the importance of the conserved Cys/His residues, we systematically mutated histidines and cysteines present in the HDAC6 ZnF‐UBP domain and generally conserved in all USPs, to alanine (Figure 2B). Wild‐type ZnF‐UBP and its mutated versions were translated in vitro in a reticulocyte lysate in the presence of [³⁵S]methionine and the in vitro‐labelled proteins were pulled down using ubiquitin‐sepharose beads. Figure 2D shows that mutating all these cysteines and histidines strongly reduces efficient ubiquitin binding. However, seven cysteines and two histidines (C1067, C1070, C1082, C1087, C1110, C1117, C1120, and H1049 and H1104) were found to play an absolutely critical role in the ZnF‐UBP–ubiquitin interaction. According to our structural model, all these residues (except C1110) are directly involved in Zn binding. Our results emphasize the functional importance of the Zn ions for the structural organization and stability of the HDAC6 ZnF‐UBP domain as a requirement for ubiquitin binding. C1110, which does not directly coordinate Zn, corresponds to a leucine in the USP5 ZnF‐UBP domain, which stabilizes the hydrophobic core. It is likely that C1110 in HDAC6 plays a similar role and mutating it into alanine would also destabilize the overall structure of the protein.

Taking into account the extraordinary affinity of HDAC6 ZnF‐UBP domain for ubiquitin, we reasoned that the HDAC6–ubiquitin complex would be unlikely to be displaced by other ubiquitin‐binding proteins.

In order to test this hypothesis, we compared HDAC6 to another polyubiquitin‐binding factor, RPN10, which is a polyubiquitin‐binding component of the 19S proteasomal subunit.

Flag‐tagged HDAC6 and RPN10 were overexpressed in Cos cells and purified (Figure 5A). Penta‐ubiquitin chains (K48 5+1) were preincubated with equal concentrations of purified HDAC6, or purified RPN10, and then treated with the same amounts of recombinant UBPY, a ubiquitin isopeptidase (Naviglio et al, 1998) presenting a remarkable in vitro activity (Hartmann‐Petersen et al, 2003). Figure 3 shows that HDAC6 efficiently hindered the action of UBPY and polyubiquitin chain degradation, whereas RPN10 only partially delayed the action of the protease. This experiment strongly suggests that the high‐affinity ubiquitin binding by HDAC6 has the potential to stabilize the pool of cellular ubiquitinated proteins.

Accordingly, we then tested the action of HDAC6 on polyubiquitin chain turnover in vivo. For this purpose, 3T3 cell lines were established from HDAC6^−/− (Zhang et al, 2003; Zhang and Matthias, in preparation) or from parental HDAC6^+/+ mice. The HDAC6^−/− cell line was also used to generate two derivatives stably re‐expressing either wild‐type HDAC6 or an HDAC6 mutant with an inactive ZnF‐UBP domain (H1094 and H1098 changed to A). This particular HDAC6 mutant was previously shown to have completely lost its ubiquitin‐binding activity (Seigneurin‐Berny et al, 2001; Bertos et al, 2004). Figure 4A shows that these cells express equivalent amounts of wild‐type and mutant HDAC6 at a level comparable to that of the HDAC6^+/+ 3T3 cell line (Figure 4A). As expected, the knockout of HDAC6 led to an accumulation of acetylated tubulin (Figure 4A, lane KO) and, interestingly, the re‐expression of wild‐type HDAC6, as well as the non ubiquitin‐binding mutant of the protein (ZnF^m), downregulated tubulin acetylation, showing that inactivation of the ZnF‐UBP domain has no effect on HDAC6 tubulin‐deacetylase activity (Figure 4A, lane ZnF^m).

The accumulation of heavily ubiquitinated cellular proteins was induced by cell treatment with the proteasome inhibitor MG132. Cells were washed (0 h) and incubated in MG132‐free medium for the indicated periods of time. Figure 4B and C shows that, compared to the HDAC6^+/+ cells, the absence of HDAC6 led to an accelerated disappearance of polyubiquitin chains. The re‐expression of wild‐type HDAC6 in the KO cells delayed the polyubiquitin chain decay and restored the parental cell line situation (Figure 4B and C). Remarkably, when HDAC6^−/− cells expressed an inactive ZnF‐UBP HDAC6 mutant, polyubiquitin chains rapidly disappeared, even faster than in the HDAC6^−/− cells (Figure 4B and C). This could be owing to a dominant negative activity of this mutant titrating out other factors involved in the protection of polyubiquitin chains.

These blots were also probed with an antibody recognizing a major histocompatibility complex (MHC) class I heavy chain whose folding and assembly is subjected to quality control mechanisms leading to the degradation of defective proteins via the proteasome–ubiquitin system (Farmery and Bulleid, 2001). Figure 4B shows that MHC heavy chains, undetectable in control cells, accumulated after MG132 treatment and their degradation paralleled that of polyubiquitin chains upon MG132 removal and was dependent on HDAC6. The use of this specific proteasome substrate further confirmed our conclusions on the role of HDAC6 as a negative regulator of polyubiquitin chain turnover.

Altogether, these results show that HDAC6, through its ubiquitin‐binding activity, controls the stability of the cellular pool of ubiquitinated proteins.

We reasoned that, owing to the high‐affinity binding of HDAC6 to ubiquitin, specific cellular factors/chaperones could be required to mediate the release of HDAC6 from ubiquitin and regulate ubiquitin‐dependent functions of HDAC6. The chaperone p97/VCP appeared as a very good candidate for several reasons: it interacts with HDAC6 (Seigneurin‐Berny et al, 2001), it possesses a segregase activity, it is a positive regulator of ubiquitin/proteasome‐dependent protein degradation and it negatively regulates excessive polyubiquitin chain assembly (Braun et al, 2002; Wang et al, 2003; Richly et al, 2005).

In order to test this hypothesis, Flag‐tagged HDAC6, p97/VCP and RPN10 were expressed in Cos cells and purified as described above (Figure 5A). His‐tagged penta‐ubiquitin chains (K48 5+1) were immobilized on Ni beads and first preincubated with an excess molar ratio of purified HDAC6. After the removal of unbound HDAC6, the beads were incubated with equal amounts of purified RPN10, p97/VCP or both (Figure 5B, input panel). The incubations were carried out in the presence or absence of ATP (Figure 5B, indicated). Following the pull‐down of the beads, interactions between different proteins were detected by Western blotting. This experiments confirmed that the three studied components, HDAC6, p97/VCP and RPN10, are individually capable of polyubiquitin chain binding (lanes 2, 3 and 4). In the absence or presence of ATP, RPN10 binding was dominant over that of p97/VCP (lanes 6 and 11) and preincubation with HDAC6 prevented the binding of RNP10 to polyubiquitin (lanes 7 and 12). In the absence of ATP, ubiquitin‐bound HDAC6 could also retain p97/VCP but did not allow the interaction of RNP10 with ubiquitin (compare lanes 5–8). Interestingly, in the presence of ATP, p97/VCP displaced HDAC6 and remained associated with polyubiquitin (lane 10). The addition of RNP10 to this mixture allowed the release of p97/VCP (lane 13), which formed a complex with HDAC6 in solution (Figure 5C).

These experiments suggest a mechanism through which cells may regulate the inhibitory effect of HDAC6 on polyubiquitin chain turnover. p97/VCP allows the removal of HDAC6 from polyubiquitin chains and can itself be removed by RPN10. By releasing p97/VCP from polyubiquitin chains, RPN10 actually allows the capture of p97/VCP by free HDAC6 (Figure 5C), and this could allow the recognition of ubiquitinated protein substrates by polyubiquitin‐binding proteins such as RPN10.

According to our data, a molar excess of HDAC6 or a decrease in p97/VCP concentration should favour protein polyubiquitination. CFTR‐ΔF508, a mutant form of CFTR, which is prone to misfolding and ubiquitination, provides an appropriate model to monitor misfolded proteins aggregation (Johnston et al, 1998). To better visualize high molecular weight ubiquitinated forms of the protein, a truncated form of CFTR‐ΔF508 (CFTR3M) was generated and its ubiquitination monitored after 6 histidine‐tagged ubiquitin coexpression and capture of ubiquitinated proteins on Ni beads. Figure 6 shows that the expression of HDAC6 significantly increases the accumulation of high molecular weight CFTR3M derivatives (lane 5). Interestingly, the downregulation of p97/VCP using specific siRNA led to the same observation (lane 4). An additive effect was observed in cells treated with p97/VCP siRNA and overexpressing HDAC6 (lane 6). Ni capture approach showed that these high molecular weight CFTR3M‐related proteins are indeed ubiquitinated forms. In the absence of 6 histidine‐tagged ubiquitin expression, no CFTR3M could be captured by Ni beads (Figure 6B).

It had previously been shown that, owing to its ubiquitin‐binding activity and its interaction with dynein motors, HDAC6 transports misfolded and ubiquitinated CFTR aggregates to aggresomes (Kawaguchi et al, 2003). We reasoned that by interfering with HDAC6–ubiquitin binding activity, p97/VCP should also decrease the efficiency of HDAC6 in inducing CFTR aggresome formation.

In order to test this hypothesis, HA‐CFTR3M was expressed in cells alone or together with p97/VCP and the % of CFTR3M‐expressing cells presenting aggresomes was determined. Ubiquitin and vimentin, known markers of aggresomes (Kawaguchi et al, 2003), were also detected to better identify these structures (Figure 7A). As previously reported (Kawaguchi et al, 2003), the expression of HDAC6 led to a marked increase in the proportion of cells forming aggresomes and the coexpression of p97/VCP significantly decreased the efficiency of aggresome formation (Figure 7B). In our hands, the treatment of cells with p97/VCP siRNA also increased the efficiency of CFTR aggresome formation (not shown). The effect of p97/VCP downregulation on aggresome formation should however be considered with caution, as, at least in yeast, Cdc48 (p97/VCP homologue) is required for normal microtubule organization (Moir et al, 1982) and p97/VCP downregulation may indirectly affect aggresome formation through its action on microtubules.

Altogether, our data strongly suggest that the steady‐state pool of cellular polyubiquitin chains depends on a finely tuned equilibrium between the concentration of HDAC6 and p97/VCP. An imbalance of HDAC6–p97/VCP molar ratio in favour of HDAC6 would enhance the formation of ubiquitinated protein aggregates and ultimately aggresome formation.

Mice with invalidated HDAC6 gene were generated (Zhang et al, 2003; Zhang and Matthias, data not shown) and mouse embryo fibroblasts were isolated from E13.5 embryos. 3T3 cell lines were established following a standard 3T3 protocol. Wild‐type and mutant HDAC6 cDNAs were cloned into a pMSCV.EGFP vector. The retrovirus was made from Phoenix cells following standard protocols. The HDAC6^−/− 3T3 cells were infected and kept in culture for 2 weeks and single GFP‐positive cells were sorted into 96‐well plates.

Cells were maintained in DMEM supplemented with 10% fetal calf serum and seeded in appropriate plates and grown to 60–70% confluence on the day of transfection. Transfections were performed with Lipofectamine 2000 (Invitrogen), using 5 μl Lipofectamine and various quantities of expression vectors or empty vectors for a total of 5 μg DNA.

p97/VCP siRNAs were designed as previously described (Wojcik et al, 2004) and purchased from Eurogentec. Control siRNAs were inactive siRNAs against HAT1 (not shown). HeLa cells were seeded in six‐well plates at 10⁵ cells/well 24 h before siRNA treatment. In situ immunofluorescence analysis was carried out as described previously (Matsuyama et al, 2002).

HDAC6 expression vectors were previously described (Verdel et al, 2000; Seigneurin‐Berny et al, 2001). HA‐ and Flag‐p97/VCP were cloned in pcDNA.3 vectors after PCR amplification of its coding sequence from a mouse embryo bank (Clonthec). GFP‐CFTR‐ΔF508 expression vector was kindly provided by Dr Kopito. Flag‐RPN10 expression vector was a kind gift of Dr Minoru Yoshida. pSG His‐ubiquitin was a gift of Dr Bohman. pGEXKG‐UBPY plasmid was a kind gift of Dr Hartmann‐Petersen. CFTR‐ΔF508 mutant lacked its C‐terminal ATPase domain (deletion of 1057–1480 region). GST‐Cterminal part of HDAC6 was cloned in pGEX4‐T‐1 vector (Amersham) after a PCR amplification of the corresponding sequence with the following primers: 5′ccc ctt cgg cga cct cca gtc ctg tac ttg3′ and 5′gtg tga gtg ggg cat gtc ctc ccc aaa3′. p97/VCP antibody was a gift of Dr Tonks. Anti‐HDAC6 was described previously (Seigneurin‐Berny et al, 2001). Anti‐MHC class I heavy chain was a monoclonal antibody against haplotype H2Dd (Cedarlane Labs, CL9009B).

GST‐Cter was expressed in BL21 (DE3)pLysS strain (Stratagene), and after purification on glutathione sepharose resin (Sigma), the GST tag was removed with thrombin (Sigma) and the HDAC6 part was purified to homogeneity after phenyl sepharose column and S75 gel filtration (Amersham). The ubiquitin sequence was amplified by PCR and cloned in pETM13, as His‐tagged fusion and expressed in BL21(DE3) strain (Stratagene) and purified on nickel probond resin (Invitrogen) and purified to homogeneity by S75 gel filtration (Amersham).

HDAC6 C‐terminal part cloned in pET28a+ (Novagen) and mutagenized using the QuickChange site‐directed PCR mutagenesis (Stratagene) and the resulting mutations were controlled after sequencing. The corresponding ³⁵S‐labelled proteins were generated in a reticulocyte lysate (Promega) and pull‐downs were performed as described previously (Seigneurin‐Berny et al, 2001).

Flag‐tagged HDAC6, p97/VCP and RPN10 were overexpressed in Cos7 cells after massive transfection and the tagged proteins were immunoprecipitated using M2 coupled beads and purified using a Flag‐peptide elution procedure (Sigma). Proteins were eluted in ubiquitin‐binding buffer and then concentrated using Centricon concentrator (Millipore).

Penta‐ubiquitin binding assays were performed as follows: 1 μM of purified His‐tagged penta‐ubiquitin (ubiquitin₅⁺¹ from Affinity) was immobilized on Ni beads and preincubated with 6 μM Flag‐HDAC6 for 30 min at room temperature. A 6 μM portion of Flag‐p97/VCP or Flag‐RPN10 or both was then added. Binding reactions were performed in 50 mM Tris pH 7.4, 150 mM NaCl, 1 mM β‐mercaptoethanol, 2% of glycerol, anti‐protease inhibitor (Roche) with or without 200 μM ATP and 200 μM MgCl₂, during 30 min at room temperature.

Unbound proteins were removed by centrifugation. Proteins retained were eluted by electrophoresis sample buffer. Unbound HDAC6–p97/VCP complex, after the ATP‐dependent action of p97/VCP, was immunoprecipitated using an anti‐HDAC6 antibody (Seigneurin‐Berny et al, 2001) and the proteins detected by Western.

A 1 μM portion of ubiquitin chains in ubiquitin‐binding buffer was preincubated at 37°C with 6 μM of purified Flag‐HDAC6 or 6 μM of purified Flag‐RPN10. GST‐UBPY (6 μM) was then added for different times. Reactions were blocked by SDS–PAGE loading buffer, and ubiquitin, GST and Flag‐tagged proteins were detected.

Supplementary data are available at The EMBO Journal Online.