Ubiquitination directly enhances activity of the deubiquitinating enzyme ataxin-3

Endogenous AT3 is ubiquitinated

FLAG–AT3 becomes ubiquitinated when coexpressed with Ub in mammalian cells (Figure 1B). To determine whether endogenous AT3 is also ubiquitinated, we conducted stringent denature/renature immunopurification of AT3 from cells (Figure 1C). In neural M17 cells, inhibition of the proteasome with lactacystin led to the accumulation of higher molecular weight AT3-immunoreactive bands (Figure 1C). Immunopurification confirmed that these bands are ubiquitinated forms of AT3 (AT3-Ub), including mono-ubiquitinated AT3 (Figure 1C). Ubiquitinated AT3 is present also under non-stressed conditions, although at lower levels. Thus, a fraction of AT3 is ubiquitinated in unperturbed cells under steady-state conditions, and proteasome inhibition increases this fraction.

Ubiquitinated AT3 cleaves Ub chains more quickly than unmodified AT3

We hypothesized that AT3 ubiquitination might regulate its enzymatic activity. To investigate this, we purified unmodified or ubiquitinated AT3 from transfected cells and tested their activity in vitro (Figure 2A). Immunopurified AT3-Ub showed markedly increased DUB activity towards K63-linked hexa-Ub chains (K63-Ub6; Figure 2B). AT3-Ub also rapidly and quantitatively cleaved higher molecular weight Ub chain complexes (HMW; Figure 2B) that most likely represent longer polymers of Ub6 chains (Winborn et al, 2008). Mutating the catalytic cysteine at position 14 of AT3 to alanine renders the enzyme inactive (Burnett et al, 2003; Mao et al, 2005; Nicastro et al, 2005). Catalytically inactive, ubiquitinated AT3 (AT3(C14A)-Ub) showed no activity towards Ub chains (Figure 2B), indicating that the increased activity of AT3-Ub is not due to the spurious co-purification of another DUB. The major AT3-Ub species present in our immunopurifications is mono-ubiquitinated AT3, suggesting that a single conjugated Ub is sufficient to activate AT3 (Figure 2B).

To exclude the possibility that AT3-Ub isolated from cells co-purifies with modulators of its enzymatic activity, we performed similar reactions with recombinant AT3-Ub generated in vitro using the E2 UbcH5c and the E3 CHIP. Recombinant AT3-Ub also showed increased DUB activity (Figure 2C). Thus, ubiquitination enhances AT3 activity independent of potential cofactors/interactors or other types of post-translational modification.

Whereas AT3 does not cleave homotypic tetra-Ub (Ub4) chains efficiently, it does cleave mixed-linkage Ub4 chains (Winborn et al, 2008). Similar to unmodified AT3, AT3-Ub does not cleave K63-Ub4 chains but shows increased activity towards mixed-linkage Ub4 chains (Supplementary Figure 1). Therefore, ubiquitination enhances the activity of AT3 without altering its preference for K63-Ub6 or mixed-linkage Ub4 chains.

We next compared the kinetics of DUB activity of AT3-Ub with unmodified AT3. AT3-Ub prepared in vitro cleaved K63-Ub6 chains much more rapidly than unmodified AT3: reaction products were detectable within 10 min and went to completion by 6 h (Figure 3). The major reaction products were tetra-, tri-, and di-Ub (Figure 3) even in reactions extending for 20 h (Figure 2). The fact that both AT3 and AT3-Ub cleave longer chains into smaller chains rather than into mono-Ub (Figures 2 and 3) suggests that AT3 ubiquitination enhances activity towards K63-Ub6 chains without affecting the manner in which it cleaves.

AT3-Ub prepared in vitro comprises a ladder of ubiquitinated species (Figures 2C and 3), suggesting that AT3 mono-ubiquitination may not be the only activating post-translational event. Therefore, we compared the activity of AT3-Ub prepared with wild-type UbcH5c, which can form poly-Ub chains, with AT3-Ub prepared with UbcH5c(S22R), which only mono-ubiquitinates substrates because it cannot extend Ub chains (Brzovic et al, 2006). AT3-Ub prepared either way was more active than unmodified AT3 (Supplementary Figure 2). Together, these data show that AT3 activity is enhanced by ubiquitination and that mono-ubiquitination of AT3 is sufficient for activation.

UIMs are not necessary for increased AT3 activity by ubiquitination

The UIMs of AT3 serve at least two functions: they mediate high-affinity binding to Ub chains and restrict the types of chains that can be cleaved by AT3. Whereas normal AT3 prefers cleaving K63 linkages, AT3 with mutated UIMs cleaves K63 and K48 linkages approximately equally well (Winborn et al, 2008). Considering this important function of UIMs in modulating the ability of AT3 to bind and cleave Ub chains, we hypothesized that an intramolecular interaction between UIMs and Ub conjugated to AT3 might underlie AT3 activation by ubiquitination. We investigated this possibility with AT3 mutated at conserved residues in each UIM (A → G/S → D; denoted as AT3(UIM^*)). These mutations eliminate the ability of AT3 to bind Ub chains with high affinity (Supplementary Figure 3).

The extent of activation of AT3(UIM^*)-Ub towards K63 chains was less robust than that observed with AT3(WT)-Ub (Figure 4A). In contrast, AT3(UIM^*)-Ub showed strong enhancement of activity towards penta-Ub K48 chains (K48-Ub5; Figure 4B). Intriguingly, the pattern of activation towards K63 versus K48 chains is reversed for AT3(WT)-Ub and AT3(UIM^*)-Ub: AT3(WT)-Ub shows greater enhancement of K63 chain cleavage, whereas AT3(UIM^*)-Ub shows greater enhancement of K48 chain cleavage (Supplementary Figure 4). Thus, although the UIMs are required for preferential cleavage of homotypic K63 chains by AT3 (Winborn et al, 2008), they are not necessary for activation of AT3 by ubiquitination.

To examine which domains of AT3 are sufficient for ubiquitination-dependent activation, we studied the effect of ubiquitination on the activity of the isolated Josephin domain. The Josephin domain alone (denoted as AT3(J)) comprises amino acids 1–182 and excludes all UIMs. Although AT3(J)-Ub was not activated towards K63-Ub6 chains (Figure 5A), it cleaved K48-Ub5 chains more rapidly than unmodified Josephin domain (Figure 5B). These results support the conclusion that UIMs are not necessary for activation of AT3 by ubiquitination.

AT3 activation by ubiquitination does not occur in trans

AT3 has been reported to self-associate in cells (Todi et al, 2007a) and in vitro (Ellisdon et al, 2006). It is thus possible that activation of AT3 by ubiquitination depends on homomeric interactions between AT3 proteins. To explore this possibility, we prepared unmodified and ubiquitinated forms of catalytically inactive AT3 (AT3(C14A)) to test in DUB assays with AT3(WT). AT3(C14A) or AT3(C14A)-Ub was mixed with unmodified, wild-type AT3 at equimolar concentrations, then incubated with K63-Ub6 chains. AT3(WT) activity towards K63-Ub6 was not enhanced by the presence of AT3(C14A)-Ub, indicating that activation most likely occurs in cis (Figure 5C). Moreover, as both AT3 and AT3-Ub fractionate identically by size-exclusion chromatography (Supplementary Figure 5), ubiquitination of AT3 does not appear to lead to a major change in AT3 tertiary structure. AT3 runs as an elliptical protein (Chow et al, 2006), enabling us to use size-exclusion chromatography to detect large conformational changes. This method, however, cannot detect conformational changes occurring at a domain or subdomain level.

Activity of pathogenic AT3 is also enhanced by ubiquitination

PolyQ expansion in AT3 causes the neurodegenerative disease SCA3/MJD. Disease protein context and the normal cellular functions of the implicated protein are important factors in the selective, progressive neurodegeneration seen in polyQ diseases (reviewed by Gatchel and Zoghbi, 2005; Todi et al, 2007b; Williams and Paulson, 2008). Therefore, understanding how AT3 function is affected by polyQ expansion may provide clues to pathogenesis. Expanded AT3 binds and cleaves Ub chains similarly to normal AT3 in vitro (Burnett et al, 2003; Chai et al, 2004; Winborn et al, 2008). In cells, however, expanded AT3 (AT3(Q80)) leads to accumulation of ubiquitinated proteins, probably through an indirect mechanism (Winborn et al, 2008). Given both increased AT3 activity when it is ubiquitinated and the fact that AT3(Q80) becomes ubiquitinated when coexpressed with Ub (Todi et al, 2007a; and data not shown), we investigated the ability of AT3(Q80)-Ub to cleave K63 and K48 chains. Expanded AT3-Ub prepared in vitro showed enhanced activity towards both K63 and K48 chains (Figure 6A). This enhancement did not differ from that of wild-type AT3-Ub (Figure 6B). Thus, polyQ expansion in AT3 does not appreciably alter activation by ubiquitination in vitro.

Ubiquitination of AT3 increases with certain stressors

As AT3 is implicated in Ub-dependent protein quality control (Wang et al, 2000, 2006; Doss-Pepe et al, 2003; Chai et al, 2004; Warrick et al, 2005; Zhong and Pittman, 2006), we investigated whether perturbations in protein homoeostasis alter ubiquitination of endogenous AT3. In M17 (Figure 1C) and Cos7 cells (Figure 7A), levels of AT3-Ub increased during proteasome inhibition. Ub overexpression also led to higher levels of AT3-Ub (Figure 7A). Jointly inhibiting the proteasome and overexpressing Ub did not further increase AT3 ubiquitination, suggesting a saturation point for endogenous AT3 ubiquitination. Proteasome inhibition led to higher levels of conjugated Ub, whereas Ub overexpression increased levels of both conjugated and non-conjugated Ub (Figure 7A), suggesting that the ubiquitination status of AT3 functions as a feedback sensor of the overall levels of conjugated Ub in the cell.

Ubiquitination of endogenous AT3 also increased when the unfolded protein response (UPR) was induced by DTT, although DTT did not appreciably alter levels of conjugated Ub (Figure 7B). In contrast, acute or prolonged heat shock did not affect AT3 ubiquitination or Ub levels in Cos7 (Figure 7C), 293, or M17 cells (data not shown). Taken together, these data suggest a physiological function of AT3 ubiquitination during some, but not all, stressors.

Lastly, we investigated AT3 ubiquitination in brain lysates from transgenic mice expressing normal (AT3(Q15)) or pathogenic AT3 (AT3(Q84)) driven by the prion promoter (Cemal et al, 2002). Mice expressing AT3(Q84) show motor anomalies as early as 6 weeks of age (Cemal et al, 2002; and our unpublished data). AT3(Q84) was consistently more heavily ubiquitinated than AT3(Q15) in mice aged 4–12 months (Figure 7D). Importantly, this ubiquitinated AT3(Q84) is in the soluble fraction of brain lysates, indicating that it is not sequestered into inclusions. These data suggest that ubiquitination of expanded AT3 may be involved in SCA3/MJD pathogenesis.

Ubiquitination as a regulator of enzymatic activity

It makes biological sense that DUB activity would be regulated through feedback mechanisms tied to Ub-dependent pathways. Regulation of activity has been reported for some DUBs at various levels, including transcription (upregulation), phosphorylation (downregulation), interaction with single proteins (activation or inhibition), incorporation into complexes (upregulation or downregulation), or proteasomal degradation (inactivation) (reviewed by Nijman et al, 2005; Ventii and Wilkinson, 2008; Yao et al, 2008). In addition, USP1, a DUB that functions in DNA repair, inactivates itself through cleavage upon ultraviolet irradiation (Huang et al, 2006). Cleavage of USP7, which functions in apoptosis, is accomplished by caspase 3 (Vugmeyster et al, 2002). Here, we report for the first time that ubiquitination of a DUB directly regulates its activity.

Among the nearly 100 DUBs expressed in humans, at least six are known to be ubiquitinated: USP4, USP6, USP7, UCH-L1, AT3, and Josephin 1 (Shen et al, 2005; Wada and Kamitani, 2006; Fernandez-Montalvan et al, 2007; Meray and Lansbury, 2007; Todi et al, 2007a; and our unpublished data). At least five additional DUBs contain UIMs, a Ub-binding domain that often promotes ubiquitination of the host protein (Klapisz et al, 2002; Oldham et al, 2002; Polo et al, 2002; Shekhtman and Cowburn, 2002; Miller et al, 2004). Therefore, we anticipate that the ability of cells to control DUB function through ubiquitination will extend beyond AT3 to other DUBs. Modulation of DUB activity most likely will prove to be complex and tailored to the pathways in which the particular DUB participates.

Physiological function of ubiquitinated AT3

AT3 has been implicated in diverse cellular pathways of protein quality control. In cooperation with Valosin Containing Protein/p97, AT3 regulates ERAD substrate trafficking to the proteasome. In Drosophila, AT3 rescues neurodegeneration from expanded polyQ proteins in an activity- and proteasome-dependent manner (Wang et al, 2000, 2006; Doss-Pepe et al, 2003; Chai et al, 2004; Warrick et al, 2005; Zhong and Pittman, 2006). Building off these reports, we investigated changes in the ubiquitination levels of endogenous AT3 when the proteasome is impaired, when excess Ub is present, or when the UPR is induced. In each case, we observed an increase in levels of AT3-Ub, suggesting a physiological function of ubiquitinated AT3 during some stressors. AT3 binds both K48- and K63-linked Ub chains, but it preferentially cleaves K63-linked Ub chains, and is efficient at cleaving K63 linkages within mixed-linkage chains (Winborn et al, 2008). AT3 ubiquitination increases its activity, but does not alter the preference of AT3 for K63 linkages. Therefore, during certain types of stress, particularly those associated with perturbations in Ub chain homoeostasis, we propose that AT3 activity is upregulated through ubiquitination, thus helping to restore homoeostasis quickly. The fact that AT3-Ub levels do not increase during heat shock argues that this post-translational modification may not be part of a general response to proteotoxic stress.

We favour the view that increased levels of AT3-Ub when the proteasome is inhibited do not simply reflect AT3-Ub being a substrate for proteasomal degradation. When AT3-Ub is incubated in vitro with proteasome fractions, it is rapidly deubiquitinated by proteasome-associated DUBs, but the now unmodified AT3 is only slowly degraded (Todi et al, 2007a). This suggests that although AT3-Ub is readily deubiquitinated by proteasome-associated DUBs, it is not a favoured substrate for proteasomal degradation. Thus, it is possible that proteasome perturbation leads to accumulation of ubiquitinated proteins in cells, exceeding the capacity of proteasome-associated DUBs to rapidly deubiquitinate proteins, including AT3-Ub. The resultant increase in AT3-Ub (and correspondingly increased DUB activity) may help to restore Ub-dependent protein homoeostasis. Once homoeostasis is reestablished, AT3-Ub levels would be lowered by proteasome-associated DUBs.

Ubiquitination of AT3 could also regulate its functional associations with Ub ligases. AT3 interacts with at least three Ub ligases: E4B (Matsumoto et al, 2004), Hrd1 (Wang et al, 2006), and CHIP (KMS and HLP, manuscript in preparation). The interaction of AT3 with Ub ligases probably serves a function beyond merely targeting AT3 for degradation. Indeed, AT3 may exert an effect as a Ub chain editor within Ub ligase complexes. As AT3 is ubiquitinated by CHIP during ubiquitination cycles of model substrates (KMS and HLP, manuscript in preparation), we propose that AT3 ubiquitination serves in part to upregulate its DUB activity so as to facilitate Ub chain editing when Ub is conjugated to proteins.

The detrimental effects of expanded (i.e., pathogenic) AT3 on the nervous system were established before the cellular functions of this disease protein became recognized (Kawaguchi et al, 1994; Stevanin et al, 1995a, 1995b). As with other polyQ disease proteins, the mechanism by which pathogenic AT3 causes neurodegeneration is still poorly understood. What has become increasingly clear is the importance of protein context to polyQ disease pathogenesis. Studies of ataxin-1 (the disease protein in Spinocerebellar Ataxia Type 1) have shown that the relative abundance of specific ataxin-1 macromolecular complexes differs in normal and disease states (Lam et al, 2006; Bowman et al, 2007; Lim et al, 2008). How polyQ expansion affects protein–protein interactions for AT3 is still unknown. Thus far, expanded AT3 has behaved similarly to non-pathogenic AT3 in vitro (Burnett et al, 2003; Chai et al, 2004; Winborn et al, 2008). Likewise, in this study, polyQ expansion in AT3 did not alter activation by ubiquitination. We should stress, however, that although we do not notice differences in these in vitro assays, expanded AT3 could differ in its activity towards endogenous substrates. Indeed, pathogenic AT3 was shown to be less able than wild-type AT3 to reduce conjugated Ub levels in cells overexpressing Ub (Winborn et al, 2008).

We also observed increased ubiquitination of pathogenic AT3 in a mouse model of SCA3/MJD, suggesting that expanded AT3 ubiquitination is linked to pathogenesis in SCA3/MJD. Catalytically active AT3 is neuroprotective against toxic proteins in flies, and expanded AT3 retains this neuroprotective function (Warrick et al, 2005). As ubiquitinated AT3 has enhanced catalytic activity, it is possible that in SCA3/MJD mice, expanded AT3 ubiquitination contributes to neuronal protection. Future studies investigating whether AT3 ubiquitination is necessary for maintenance of cellular homoeostasis will address this and other possibilities.

AT3 regulation by ubiquitination

Our results indicate that ubiquitination of the isolated Josephin domain is sufficient to enhance its activity. Ubiquitination could enhance AT3 activity through at least two possibilities. First, if unmodified AT3 normally interacts with Ub chains in a manner that aligns chains so that the catalytic site has poor access to isopeptide bonds, AT3 ubiquitination could improve Ub chain presentation to the catalytic site. Ubiquitination could also directly alter the conformation of the catalytic site on the Josephin domain, effectively increasing its ability to cleave isopeptide bonds. The results of our size-exclusion chromatography, which argue against large conformational changes in AT3 upon ubiquitination, do not exclude smaller conformational changes occurring at a sub-domain level. The solution structure of the Josephin domain has been resolved (Mao et al, 2005; Nicastro et al, 2005). Therefore, identifying where AT3 is ubiquitinated in cells and in vitro, and which of its 15 lysines is important for activation, should offer structural insight into how AT3 is activated.

Although UIMs are dispensable for ubiquitination-dependent activation of AT3, they do confer upon AT3 a preference for K63 linkages. Unlike wild-type AT3, AT3 with mutated UIMs cleaves K48 and K63 linkages approximately equally well (Winborn et al, 2008). In the current studies, however, AT3-Ub with mutated UIMs showed more robust activation towards K48 than K63 linkages. Perhaps the catalytic site has an inherent preference for K48 linkages, which is altered by the UIMs in the full protein. In this case, ubiquitination of AT3 in the absence of UIMs would be expected to enhance its activity more significantly towards this preferred substrate. This model is supported by our results with the isolated Josephin domain, which when ubiquitinated showed activation towards K48- but not K63-linked Ub.

The enhanced activity of ubiquitinated Josephin domain is not as pronounced as that of ubiquitinated wild-type AT3 (e.g., compare Figures 3 and 5). Although we have ruled out the UIMs as necessary for ubiquitination-dependent activation of AT3, this by no means excludes their importance to the overall activity of AT3. Indeed, the marked activation of ubiquitinated wild-type AT3 may reflect the manner in which UIMs present isopeptide bonds to the catalytic site. It is also possible that full-length AT3 and the isolated Josephin domain are ubiquitinated at different lysines, with each having a different effect on activation. Knowing where AT3 is ubiquitinated and understanding the function of the UIMs in presenting Ub chains to its catalytic site should help our understanding of overall AT3 activity and its enhancement by ubiquitination.

Summary

The functional diversity of DUBs must be achieved through a combination of structural elements, modulatory protein domains, specific protein–protein interactions, and post-translational modifications. DUBs contain various protein motifs and interact with distinct partners that assign DUBs to specific pathways, confer substrate selectivity, and regulate enzymatic activity. We suggest that direct regulation of enzymatic activity by ubiquitination further refines the cellular functions of certain DUBs. In the case of AT3, enhancement of activity through ubiquitination may constitute a feed-forward regulatory process that helps to restore Ub-dependent homoeostasis. Ubiquitination could also serve as a direct enhancer of activity in other classes of enzymes.

Constructs

Recombinant proteins are expressed in pGEX-4T1, pGEX-6P1, and pET28a vectors. FLAG-AT3 is expressed in pVETL (Todi et al, 2007a; Winborn et al, 2008). To inactivate UIMs, conserved alanine and serine residues of each UIM were mutated into glycine and aspartic acid residues, respectively (QuickChange Mutagenesis; Stratagene). Constructs for UbcH5c(WT) and UbcH5c(S22R) are Addgene plasmids 12643 and 12644.

Antibodies

The following antibodies were used: rabbit polyclonal anti-Ub (1:500; Dako); mouse monoclonal anti-Ub (1:500; P4D1; Santa Cruz Biotech); mouse monoclonal anti-AT3 (1:1000; 1H9; provided by Dr Yvan Trottier); rabbit polyclonal anti-MJD (1:40 000; Paulson et al, 1997); rabbit polyclonal anti-HSP70 (1:1000; StressGen); rabbit polyclonal anti-HA (1:500; Y11; Santa Cruz Biotech); goat polyclonal anti-GST (1:10 000; GE Healthcare); peroxidase-conjugated, goat anti-rabbit, goat anti-mouse, and rabbit anti-goat secondary antibodies (1:15 000; Jackson Immunoresearch).

Cellular treatments, transfections, and protein extraction

For proteasome inhibition, lactacystin (Boston Biochem) was used at a final concentration of 10–20 μM in regular cell media. For UPR induction, DTT (Sigma-Aldrich) was used at a final concentration of 5 μM in regular media. Heat shocking was conducted at 39 or 41°C. Cells were transfected using Lipofectamine-PLUS (Invitrogen). For western blotting, cells were lysed in 95°C Laemmli buffer with 100 mM DTT, boiled for 5 min, sonicated, centrifuged, and loaded on 4–20, or 10% SDS–PAGE gels.

Immunopurification and protein preparation

Stringent denature/renature AT3 immunopurification from cells was conducted as described previously (Todi et al, 2007a). Briefly, one 10 cm dish per experimental group was lysed in RIPA buffer (20 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5% deoxycholic acid, 1% Nonidet P-40, pH 7.4) with protease inhibitors (PI; Sigma-Aldrich), denatured for 30 min in 1% SDS, then renatured in 4.5% Triton X-100 for 30 min. Lysates were then incubated with antibody for 2 h at 4°C, rinsed 4 × with RIPA+PI and eluted with Laemmli buffer and boiled for a short time. Anti-AT3 (MJD) antibody used for IPs was cross-linked to protein A beads (Sigma-Aldrich) using 0.2 M triethanolamine.

Protein isolated from Cos7 cells for DUB reactions was collected by lysing cells in RIPA+PI. For unmodified AT3, cells were transfected with FLAG–AT3. Lysates from four 10 cm dishes per experiment were incubated in bead-bound anti-FLAG antibody for 2 h at 4°C, rinsed 5 × with RIPA+PI, 3 × with DUB reaction buffer (50 mM HEPES, 0.5 mM EDTA, 1 mM DTT, 0.1 mg/ml ovalbumin, pH 7.5), retained on beads, and used for reactions. We have not observed FLAG–AT3 ubiquitination in the absence of coexpressed Ub. To prepare AT3-Ub, cells were transfected with FLAG–AT3 and HA–Ub. FLAG–AT3 was isolated as above. AT3 was eluted using 3 × FLAG peptide (Sigma-Aldrich), incubated with bead-bound anti-HA antibody (2 h; Dako) to isolate AT3-Ub, rinsed 5 × with RIPA+PI and 3 × with DUB reaction buffer, and retained on beads for reactions. For quantification, a subset of the protein was eluted from beads using Laemmli buffer and quantified using serial dilutions and silver staining. Recombinant protein was prepared as described previously (Todi et al, 2007a; Winborn et al, 2008). Recombinant protein was eluted using reduced glutathione or PreScission Protease (GE Healthcare), and quantified using serial dilutions, Coomassie staining, and UV spectrophotometer (NanoDrop; Thermo Scientific).

Brain lysate preparation

Flash-frozen brains from YAC transgenic mice expressing AT3(Q15) or AT3(Q84) (Cemal et al, 2002) were homogenized in RIPA+PI, sonicated, and ultracentrifuged. For western blotting, supernatants were supplemented with 1% final SDS and 100 mM DTT, boiled, and loaded on 10% SDS–PAGE gels. Stringent immunopurification was conducted as outlined above.

Ubiquitination reactions

GST-bound AT3 was ubiquitinated in vitro using 1 μM CHIP (E3), 8 μM UbcH5c (E2), 0.16 μM E1 (Boston Biochem), 50 μM Ub, 4.5 μM MgCl₂, and 4.5 μM ATP in kinase reaction buffer (50 mM TRIS, 50 mM KCl, 0.2 mM DTT, pH 7.5) for 2 h at 37°C. AT3 that was not ubiquitinated underwent the same treatment, without Ub or ATP/MgCl₂. AT3 and AT3-Ub were purified using glutathione sepharose beads (GE Healthcare), and retained on beads, or eluted in DUB reaction buffer (PreScission Protease; GE Healthcare). Protein was quantified using Coomassie staining and UV spectrophotometer.

Deubiquitination reactions

All DUB reactions were continuous reactions. Protein was quantified before use with UV spectrophotometer and/or serial dilutions and silver or Coomassie staining. Ub chains (250 nM; Boston Biochem) were incubated with AT3 species (50–100 nM) in DUB reaction buffer at 37°C. Fractions were collected at the indicated times in 2% Laemmli, 100 mM DTT, and boiled for 1 min. Samples were loaded in 4–20, 10, or 15% SDS–PAGE gels.

Western blotting and quantification

Western blotting was conducted as described previously (Todi et al, 2007a; Winborn et al, 2008). Blots were imaged using autoradiography film (Kodak) or VersaDoc 5000 MP (Bio-Rad). For semi-quantification, images were collected exclusively using VersaDoc 5000 MP below-saturation levels, and quantified with Quantity One (Bio-Rad). Background was subtracted equally among lanes. Reaction products were added and divided by the sum of all bands (e.g., ∑(Ub5, Ub4, Ub3, Ub2, Ub1)/∑(Ub6, Ub5, Ub4, Ub3, Ub2, Ub1) for Ub6 chains). Two-tailed Student's t-test was used for statistical analyses.

Ub chain binding assay

Binding assays were conducted as described previously (Winborn et al, 2008). Briefly, GST–AT3 species (250 nM) were incubated with 250 nM Ub3-7 chains (Boston Biochem) for 30 min at 4°C to minimize proteolysis. Unbound supernatant fractions were removed and added to loading buffer. Beads were washed four times with buffer A (20 mM HEPES, 120 mM NaCl, 10% glycerol, 1% Triton X-100, pH 7.4) supplemented with 0.1% Nonidet P-40. Proteins were eluted with loading buffer.

Size exclusion chromatography

A volume of 100 μl of sample containing 2 μg of purified, recombinant AT3 or AT3-Ub were separated at 25°C on a Superdex 200 HR 10/30 column (Amersham) in U buffer (50 mM Tris (pH 7.5), 50 mM KCl, 0.2 mM DTT) with a flow rate of 1.0 ml/min and a fraction size of 1 ml. Gel filtration standards (Bio-Rad) were used to predict apparent molecular weight.