Open-gate mutants of the mammalian proteasome show enhanced ubiquitin-conjugate degradation

When in the closed form, the substrate translocation channel of the proteasome core particle (CP) is blocked by the convergent N termini of α-subunits. To probe the role of channel gating in mammalian proteasomes, we deleted the N-terminal tail of α3; the resulting α3ΔN proteasomes are intact but hyperactive in the hydrolysis of fluorogenic peptide substrates and the degradation of polyubiquitinated proteins. Cells expressing the hyperactive proteasomes show markedly elevated degradation of many established proteasome substrates and resistance to oxidative stress. Multiplexed quantitative proteomics revealed ∼200 proteins with reduced levels in the mutant cells. Potentially toxic proteins such as tau exhibit reduced accumulation and aggregate formation. These data demonstrate that the CP gate is a key negative regulator of proteasome function in mammals, and that opening the CP gate may be an effective strategy to increase proteasome activity and reduce levels of toxic proteins in cells.

T he 26S proteasome, a B2.5-MDa holoenzyme complex, is the sole adenosine triphosphate (ATP)-dependent protease in the eukaryotic cytosol and nucleus, and mediates the irreversible degradation of target substrates conjugated to ubiquitin. It controls intracellular protein levels on a global scale and in particular plays a key role in protein quality control 1,2 . The proteasome holoenzyme (or 26S proteasome) comprises of the 28-subunit core particle (CP, also known as the 20S) and the 19-subunit regulatory particle (RP, also known as the 19S or PA700) 3 . At the interface between the RP and CP, two ring assemblies are axially aligned: the heterohexameric ATPase ring of the RP (known as the RPT ring, and composed of RPT1-RPT6) and the heteroheptameric a-ring of the CP (composed of a1-a7). A number of reversibly associated proteins have been identified, some of which influence the activity of proteasomes [4][5][6] . The overall architecture of the proteasome was recently established through cryo-electron microscopy studies 7,8 .
The CP is composed of four heteroheptameric rings, thus forming an a 7 b 7 b 7 a 7 structure. The outer rings of a-subunits form the substrate translocation channel while the b-subunit-forming inner rings contain six proteolytic active sites (two trypsin-like, two chymotrypsin-like and two caspase-like, in specificity) in their interiors. ATP-dependent protease complexes typically have proteolytic sites sequestered within CP-like cylinders 9 . Broad-spectrum proteasome inhibitors, such as bortezomib, target these sites, and are effective anti-cancer agents 10 . The RP interacts with the polyubiquitin chains of the substrate and translocates the substrates into the CP, with substrate deubiquitination occurring either prior to or contemporaneously with translocation 7 . Deubiquitination on the RP may promote or delay proteasomal degradation, possibly depending on the coordination between the rates of ubiquitin chain trimming and substrate translocation [11][12][13][14][15] . Due to the exceptional complexity of the system, many of the regulatory mechanisms of proteasome activity and homoeostasis remain to be elucidated.
In the free CP (CP that is not engaged with the RP), the N-terminal tails of the a-subunits fill the centre of the ring. They are tightly interlaced to form the gate, blocking substrate access into the proteolytic chamber 16,17 . On binding of the RP, the N-terminal tails are displaced, removing the block to substrate translocation. Gate opening is driven by docking of the C-terminal tails of a subset of RPT proteins into the seven intersubunit pockets of the a-subunits 18 . In addition to the RP, other endogenous activators of the CP gate include proteasome activator 28ab (PA28ab, also known as the 11S), PA28g, PA200/ Blm10 (ref. 1). The RPT ring creates the RP substrate translocation channel that is then attached to the CP channel 7 . A tight co-alignment of the RP and CP channels is generated by conformational change when the proteasome is engaged with polyubiquitinated substrates or ATPgS 19,20 . ATP-driven conformational dynamics of the RPT ring induce substrate translocation and unfolding probably through either concerted or sequential programs of ATP hydrolysis around the ring 21,22 .
Previous studies using the yeast proteasome indicated that, among the key components of the gate, such as a2, a3 and a4, deletion of the N-terminal tail of the a3 subunit resulted in conformational destabilization of other N-terminal residues and consequently opening of the CP channel into the proteolytically active interior chamber 16,23 . Substrate translocation channels and the regulated gates into the proteolytic sites might be a general theme for ATP-dependent proteases. However, the gating of mammalian proteasomes and the consequences of gate opening in mammalian cells are essentially uncharacterized.
To understand the role of the CP gate in mammalian proteasomes, we generated human cell lines that stably express a3DN subunits. We observed enhanced activity of purified mutant proteasomes measured by hydrolysis of fluorogenic peptides and degradation of polyubiquitinated protein substrates. The hyperactivity of a3DN proteasomes was observed for both free CP and holoenzyme complexes. We also found that the increased cellular proteasome activity of a3DN proteasomes stimulated substrate degradation and significantly delayed tau aggregate formation in cultured cells. Finally, multiplexed quantitative proteomics using isobaric tandem mass tags (TMTs) revealed that levels of B200 proteins were significantly reduced in the a3DN cells. These findings indicate the importance of the regulated CP channel in mammals, which functions as a rate-limiting step in proteasome-mediated proteolysis, and suggest that a3DN proteasomes could potentially help cells to cope with the proteotoxic stresses implicated in various neurodegenerative diseases.

Results
Generating open-gated mutant proteasomes. Of the seven a-tails, that of a3 projects most deeply into the centre of the translocation channel, at the same time contacting and potentially stabilizing the N-terminal tails of many other a-subunits (Fig. 1a). In addition, this region is evolutionarily conserved across the eukaryotes (for example, 92.9% identity between humans and yeast a3 N-termini) to a high degree, in contrast to the body of a3, which is less than 50% identical between humans and yeast (Fig. 1b). The virtually complete conservation of a3 N-termini suggests a common gating mode for the CP channel from yeast to mammals. To study gating of the substrate translocation channel in mammals, we stably overexpressed a flag-tagged form of a3 with a 9-residue deletion encompassing the tail element. Overexpression was carried out in the HEK293-b4-biotin cell line that allows for rapid purification of human proteasomes, either 20S and 26S forms, via the b4 subunit of the CP 24 . Two clones of stable cell lines that expressed different amounts of exogenous a3DN-flag were obtained, with the a3DN #2 clone (hereafter referred to as the a3DN cell line) showing the more prominent expression of the mutant subunit (Fig. 1c).
Active human 26S proteasomes were affinity-purified from the parental (wild type) and a3DN cells (Fig. 1d,e). The overall integrity and abundance of a3DN proteasome holoenzymes were virtually identical to those of wild type. In addition, the stoichiometry of a3 within the proteasome appeared to be proper in the mutant complex. Fortuitously, endogenous a3 mRNA expression was dramatically downregulated on a3DN-flag mRNA expression (Fig. 1f). Quantitative RT-PCR using primers specific for either endogenous and exogenous a3s indicated that the mutant a3DN mRNA levels were B18 times higher than the endogenous a3 mRNA (Fig. 1g), suggesting that the stable a3DN cell line had predominantly open-gated proteasomes. At this ratio of mutant to wild type, CP from a3DN cells would be expected to exhibit wild-type gating in less than 1 of 300 complexes. Importantly, the total cellular level of a3 was comparable between the a3DN #2 clone and the parental cell line.
Enzymatic properties of a3DN mammalian proteasomes. We isolated the CP from a3DN cells (Fig. 2a) and found significantly elevated activity compared with wild-type CP, as measured by suc-LLVY-AMC hydrolysis, which is specific for the chymotrypsinlike b5 activity (Fig. 2b). The trypsin-like b2 and caspase-like b1 activities were similarly elevated, measured by Boc-LRR-AMC and Z-LLE-AMC hydrolysis, respectively (Fig. 2c). The parallel effects on all proteolytic sites indicates that the hyperactivity of mutant proteasomes reflects CP gate opening rather than allosteric modulation of active sites in the catalytic chamber.
The hyperactivity of the open-gated CP was also observed when the CP forms the 26S holoenzyme with the RP, especially in the presence of ATPgS, a slowly hydrolyzable analogue of ATP. ATP binding, not ATP hydrolysis, is thought to be sufficient to promote the 26S proteasome assembly from RP and CP, and substrate translocation 25,26 . Under the ATPgS-enriched conditions, the RP undergoes significant intersubunit rearrangement from a preengaged conformation to an engaged conformation, which exhibits coaxial alignment between translocation channels of the RPT ring and the a-ring 19,20,27 . This conformation is similar to that of 26S proteasomes when they are in translocation-competent state when associated with polyubiquitinated substrates 20,27 . Consistent with previous findings 25,26,28 , the peptide hydrolysis activity of wildtype 26S proteasomes was significantly stimulated in the presence of ATPgS (Fig. 2d). This activity stimulation by ATPgS was more dramatic on the a3DN 26S proteasome, which showed B1.6 times higher peptidase activity than wild type (Fig. 2d).
Using suc-LLVY-AMC, we measured the enzyme kinetics of translocation-competent 26S proteasomes. The k cat value of a3DN 26S (2,376 min À 1 ) was significantly higher than that of wild-type 26S (1,565 min À 1 ) while K M values were comparable (93.92 mM for a3DN versus 93.47 mM for wild type) (Fig. 2e). These kinetic data indicate that the deletion of the N-terminal tail of a3 mainly affects substrate entry rather than the proteolytic sites of the CP. When they were in the non-engaged conformations or in the presence of ATP, a3DN holoenzymes showed only modestly enhanced proteolytic activity (Fig. 2d). In addition, the peptidase activity of both the wild type (B10-fold when CP:RP molar ratio was 1:2) and a3DN CP (B5-fold) was significantly stimulated when complexed with purified RP (Fig. 2f). By reconstituting purified CP and RP with different molar ratios, we identified maximum stimulation when the molar ratio of CP and RP was 1:2. Similar to the results obtained using purified 26S proteasomes, the reconstituted CP-RP complex showed only modestly increased proteasome activity with a3DN CP in comparison with wild-type CP. However, as shown above, RP stimulation is not sufficient for the proteasomes to achieve their fully activated status required for efficient substrate degradation ( Fig. 2d; Supplementary Fig. 1). Our data imply that gate opening by the RP may be incomplete, and that the a3 tail is critical for the residual occlusive effect of the gate in the holoenzyme state.
We next examined whether the open-gated mutant proteasome has enhanced proteolytic activity using a more physiologically relevant protein substrate, polyubiquitinated Sic1 PY (Ub-Sic1 PY ), a CDK inhibitor from Saccharomyces cerevisiae, instead of (g) Same as f, except quantitative RT-PCR (qRT-PCR) was used to compare a3 mRNA levels in the a3DN #2 cell line, which determined that B18 times more a3DN subunits were expressed compared with endogenous a3. *Po0.001 (n ¼ 3, two tailed Student's t-test).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10963 ARTICLE fluorogenic peptide substrates. A modified form of Sic1, in which the PY element signals polyubiquitination with mixed Ub-linkage types, was employed in these in vitro degradation assays 29 . The purified a3DN 26S proteasomes showed more rapid degradation of Ub-Sic1 PY than wild-type proteasomes (Fig. 2g,h). Thus, opening the central gate of the CP in the mammalian proteasome promotes degradation of protein substrates when the RP is bound to the CP. Furthermore, the facilitated degradation of Ub-Sic1 PY substrates by a3DN holoenzymes may reflect a more fully open state of the CP channel as revealed by the peptide hydrolysis data. The enhancement of a3DN CP activity and the maintenance of higher activity as the 26S proteasome with engaged conformations suggests that substrate proteolysis in the catalytic core is not only CP gate-dependent, but also closely linked with other regulatory mechanisms on the proteasome. To investigate whether there are additional layers of activation required for proteasomal degradation, we tested whether blocking ATP hydrolysis influenced the proteasomal degradation of Ub-Sic1 PY . Addition of excess ATPgS in the in vitro degradation assay significantly delayed the degradation of Ub-Sic1 PY by both wild type and a3DN 26S proteasomes ( Supplementary Fig. 2), probably due to the loss of substrate translocation functionality. However, a3DN 26S proteasomes showed still facilitated Ub-Sic1 PY degradation and inhibited polyubiquitin chain trimming on the proteasome ( Supplementary  Fig. 2). The constitutive opening of the CP gate probably does not affect the substrate translocation function of the RPT ring. Thus the enhanced proteolytic capacity of mutant 26S proteasome might originate from the facilitated substrate entry rate (and possible product release as well) through the opened gate of proteasomes with engaged conformations. Taken together, our data suggest that the open-gate mutation enhances the activity of not only free CP but of proteasome holoenzyme as well. Gate opening appeared to be a regulated process even in the assembled holoenzyme, being subject to control by nucleotide and most likely substrate occupancy, and aspects of this control remain in place in the a3DN mutant. These results predict that the a3DN mutation should accelerate the degradation of ubiquitinated substrates of the proteasome in living cells, which was borne out as described below.
Open-gated proteasomes facilitate substrate degradation in cells.
The results above indicated that the a3DN proteasomes, both free CP and holoenzyme complexes, have significantly enhanced proteolytic activity. The effects of gate-opening in living cells were then investigated using a3DN cells. The steady-state levels of various transiently overexpressed proteasome substrates, including GFP U (a Ub-dependent substrate), GFP-ODC (Ub independent), Arg-GFP and RGS4-GFP (two Ub-dependent N-end rule substrates), were significantly lower in the a3DN cell line, while levels of cotransfected lacZ were comparable (Fig. 3a). The GFP mRNA levels were also not changed in the mutant cell line in the presence of all substrates (Fig. 3b), indicating that the reduction in model substrate levels is post-translational. Chase experiments were performed after release from short-term MG132 treatment because of the rapid turn-over rates of these substrates. The result confirmed facilitated degradation of GFP U and GFP-ODC in the a3DN cells (Fig. 3c). The GFP u and GFP-ODC protein levels in two cell lines were comparable after 93.92 2,376 1,565 MG132 treatment, further indicating their accelerated degradation by the hyperactive mutant proteasome ( Supplementary  Fig. 3). Moreover, among the substrates, GFP-ODC, a Ub-independent proteasome substrate, was more responsive to the gate-opening mutation ( Supplementary Fig. 3). This dramatic effect in cultured cells may reflect the fact that the proteasome exists as free CP, RP-CP (singly capped) and RP 2 -CP (doubly capped proteasomes) forms in the cell and that ODC proteins are degraded by both free CP and holoenzyme complexes 30 . Enhancing proteasome activity in the cell also resulted in reduced levels of the cell cycle checkpoint protein p53 and the selective autophagy receptor p62, which were accompanied by increased free (unconjugated) Ub and decreased polyubiquitin levels (Fig. 3d). The conjugated forms of Ub are expected to be more sensitive to proteasome activity than free forms 31,32 . We also observed increased LC3-II levels in the a3DN cells compared with wild-type cells, but this effect was lost when bafilomycin A1, an inhibitor of the late stage of autophagy, was used (Fig. 3d). These findings suggested that the autophagic flux was inhibited at the autophagosome-lysosome fusion step when cellular proteasome activity was enhanced. Consistent with this, a significantly increased number of GFP-LC3 puncta were observed in the hyperactive a3DN cells (Supplementary Fig. 4). Therefore, the dynamic activity regulation between the ubiquitinproteasome system (UPS) and the autophagy-lysosome system appears to be linked through the proteasome activity.
We then examined the degradation of various proteotoxic proteins including tau and a-synuclein (a-Syn), which are implicated in Alzheimer's and Parkinson's diseases, respectively, when accumulated and aggregated 33,34 . Both of these proteins are substrates of the proteasome and impaired proteasomal activity may be related to the progression of these diseases 35 . In the a3DN cells, levels of both overexpressed tau and a-Syn were dramatically decreased compared with those in control cells ( Fig. 3e-g;  Supplementary Fig. 5). This outcome is also likely contributed by both the open-gated CP and holoenzyme complexes because the CP is known to degrade intrinsically unstructured proteins, including tau and a-Syn 36 . However, adding back wild-type a3 to the mutant cells effectively abrogated the CP gate-opening effect by a3DN on tau and a-Syn degradation (Fig. 3f,g). Moreover, the hyperactivity of proteasomes in the a3DN cells significantly delayed the formation of a-Syn aggregates (Fig. 3h), which might be preceded by the accelerated degradation of soluble a-Syn 37 . No effects on a-Syn or tau mRNA level were observed either as a consequence of a3DN mutation or by rescuing wild-type a3 ( Supplementary Fig. 6), further indicating that facilitated proteasomal degradation results in the decreased levels of these proteins in mammalian cells.

Enhanced tau degradation by open-gated proteasomes.
Enhanced proteasome activity may be beneficial to cells by delaying proteotoxic protein accumulation and aggregation. Tau   is thought to undergo degradation via the UPS, especially during the early stages of tauopathy and Alzheimer's disease progression 35 . We used a HEK293-derived cell line that expresses the longest isoform of human tau (htau40) on doxycycline (Dox) induction (an inducible tau cell line) 38 . These cells expressed htau40 in a tightly dose-dependent manner 24 and produced SDSresistant tau aggregates, a pathological hallmark of AD, after B2 days with a high dose of Dox in culture (Fig. 4b,e). When a3DN was transfected to cells treated with 300 pg ml À 1 Dox, the levels of induced tau proteins were mildly decreased compared with that of a3 transfection (Fig. 4a), although this effect was weaker than that of stable open-gated a3DN expression (Fig. 3f). When tau was induced with 700 pg ml À 1 Dox, significantly reduced amounts of tau oligomers were observed (Fig. 4b). We observed weak effects of a3DN overexpression on monomeric tau degradation when this Dox concentration was used. Therefore, it appears that the overall levels of induced tau limit the effect of hyperactive proteasomes in mammalian cells and consequently its propensity to aggregate. Phosphorylated tau forms intraneuronal filamentous oligomers called paired helical filaments, which are the principle constituent of neurofibrillary tangles in Alzheimer's diseases. Levels of tau proteins phosphorylated at Ser 396 or Ser 199/202 were also significantly reduced in 300 pg ml À 1 Dox-treated a3DN cells, and mildly reduced in 700 pg ml À 1 Dox conditions (Fig. 4c). Under those conditions, tau mRNA levels between wild-type and a3DN cells were virtually identical (Fig. 4d), indicating that accelerated tau degradation occurs at the post-translational stage by hyperactive proteasomes. Considering that neurodegeneration and cognitive dysfunction are critically linked to the accumulated tau level in neurons, these results indicate that enhancing proteasome activity using the open-gated proteasome could be an effective therapeutic strategies for Alzheimer's and other related neurodegenerative diseases.
Tau aggregates was further examined by separating the Triton X-100 insoluble fraction from the tau cell line induced with 700 pg ml À 1 Dox. Consistently, we found significantly reduced levels of pelleted insoluble tau monomers and tau aggregates in the 16,000g (P2) and the 200g (P1) centrifugation runs, respectively (Fig. 4e). To visualize and quantify tau oligomerization in living cells, we utilized a htau40-expressing cell line with the biomolecular fluorescence complementation system (a tau-BiFC cell line) 39 , where fluorescence becomes strongly 'turned-on' on tau oligomerization. Consistent with inducible tau cells, tau-BiFC cells overexpressing open-gated a3DN proteasomes showed significantly less tau aggregation compared with cells expressing a3 (Fig. 4f). Next, we directly delivered the purified open-gated proteasomes into inducible tau cells using silica-based mesoporous nanoparticles. The nanoparticles had pore sizes between 25 and 30 nm and nickel (Ni 2 þ ) moieties, which enabled them to harbour a proteasome holoenzyme molecule through noncovalent interactions with the poly-histidine tag of proteasomes 37 . The levels of induced tau decreased more significantly after direct delivery of hyperactive mutant proteasomes than wild-type proteasomes ( Supplementary Fig. 7), indicating that exogenous a3DN proteasomes delivered using nanoparticles can delay the aggregation process of tau proteins in proteotoxic conditions. Again, the magnitude of tau depression on enhancement of proteasome activity appeared to be partially dependent on the total tau levels in cells (Fig. 4a-c). These results suggest that hyperactive proteasomes may more efficiently degrade protein substrates that impose an unusual load on the UPS, such as overexpressed tau.
Next, we examined the effect of CP gate-opening on degradation of oxidized proteins, which are an important subset of misfolded substrates of proteasomes and accumulated with age. After reactive oxygen species (ROS) was induced by menadione, oxidized proteins were labelled with 2,4-dinitrophenylhydrazine, and visualized through their carbonyl group modification. The a3DN cells showed strikingly reduced levels of oxidized proteins compared with wild type after the treatment of menadione (Fig. 4g), suggesting that hyperactive proteasomes may have accelerated oxidized proteins clearance in cells. In addition, a3DN cells showed significant resistance to cytotoxicity from menadione-mediated oxidative stress (Fig. 4h). Consequences of protein aggregates in neurons include excessive generation of free radical and oxidatively damaged proteins, which are also closely linked to neuronal dysfunction and death 40 . Our results indicate that enhancing proteasome activity through opening of the CP gate might be beneficial in protecting cells under oxidative stress conditions during neurodegeneration.
TMT-MS-based identification of a3DN proteasome targets. The global effects of enhanced proteasome activity in mammalian cells were characterized by multiplexed quantitative proteomics based on tandem mass tags-mass spectrometry (TMT-MS) (Fig. 5a). To date, many proteomic strategies aimed at identifying proteasome substrates and ubiquitination profiles using proteasome inhibitors 41,42 , but a quantitative study of the UPS proteome in response to activation of the proteasome has been unavailable. Protein samples were obtained from three independent cultures of wild-type and hyperactive a3DN cells, which showed excellent reproducibility evaluated by the intra-group component analysis and hierarchical clustering. The six samples were independently labelled with 6-plex isobaric TMT reagents, pooled for parallel comparison, fractionated using basic RP-HPLC, and analysed using MS 3 methods to quantify a total of 7,031 proteins ( Fig. 5a; Supplementary Table 1). The initial threshold for data evaluation was a more than two-fold increase or decrease with a P value o0.05. By these criteria, 332 proteins showed significant changes ( Fig. 5b; Supplementary  Table 2). Among these responding proteins, 201 were depleted in a3DN cells, many of which presumably due to accelerated protein degradation via the proteasome, given the model substrate data using cultured cells above. However, 131 proteins were enriched in the mutant cells, raising the possibility that some changes are mediated by a non-proteolytic manner or through secondary effects, for example, possibly as a part of UPS-autophagy communication (see below).
Our global proteomic analysis was initially validated by comparison with the immunoblotting data on endogenous proteins (Fig. 3d). Consistent with our previous data, levels of proteasome substrates p53 and p62 were both significantly reduced in hyperactive a3DN cells, as measured by TMT-MS (Fig. 5c). Degradation of p62, which is subject to both autophagic and proteasomal regulation, appeared to be more directly affected by the hyperactive proteasome. To the contrary, LC3 protein levels were significantly increased, consistent with the increased levels of autophagic selective substrate LC3-II, indicating proteasome activation may negatively regulate autophagy (Figs 3d and 5c, and Supplementary Fig. 2). Accumulating evidence has suggested that the overall activity of UPS affects the autophagy flux in cells: for example, suppression of UPS activity of UPS resulted in induced autophagy 42,43 . However, these systems are not communicated by a simple compensatory mechanism in cellular catabolism, because impaired autophagy leads to a decrease of UPS flux, rather than upregulation of UPS activity 43 . The underlying molecular mechanism of this crosstalk is to be determined.
To further validate the legitimacy of the target proteins that are sensitive to hyperactive proteasomes, we used immunoblotting to examine several proteins with significant depletion in the a3DN cells from TMT-MS 3 (Fig. 5d,e). Many target substrates of hyperactive proteasomes identified by quantitative TMT-MS 3 were validated by immunoblot analysis (Supplementary Fig. 8). For example, proteins whose functions involve cell motion, such as SGPL1, UNC5B, DCDC2, ITGA4, SCARB1 and ApoB, were significantly depleted in a3DN cells, while levels of proteasome subunits and relative stable proteins, such as a7, ADRM1/RPN13, GAPDH and b-actin, were unchanged ( Fig. 5e; Supplementary  Fig. 8). Moreover, when comparing the 201 hyperactive proteasome-sensitive substrates with different ubiquitome data sets [44][45][46] , B55% (121 out of 201) of these proteins overlapped between the lists (Supplementary Fig. 9; Supplementary Table 3). These data provide strong evidence that many of the protein targets from our TMT-MS analysis are true substrates of hyperactive proteasomes.
From gene ontology analysis, we identified that a substantial fraction of the hyperactive proteasome targets is enriched in several metabolic and biological processes, including the UPS, protein folding, oxidation/reduction, growth regulation and cellular metabolism (Supplementary Fig. 10; Supplementary  Table 4). Further work will be required to determine what distinguishing features of substrates they share to be susceptible to the a3DN proteasome. It will be also important to determine the capacity and selectivity of hyperactive proteasomes, especially for the clearance of various proteotoxic proteins.
Next, we examined the levels of various Ub-linkage types, which are crucial determinants of substrate fates. Moreover, different linkages are expected to be regulated and recognized independently although many of related biochemical questions are still unanswered 47,48 . We found that, in the a3DN cells, the Lys63 (K63)-linked polyubiquitin chains were significantly depleted while most other linkage types were relatively comparable ( Fig. 5f; Supplementary Fig. 11; and Supplementary  Table 5). Recently, the K63 chain was identified as a novel sensor/ regulator of cellular oxidative stress 49 . This result and our previous finding that the hyperactive cells are more resistant to ROS-induced protein oxidation and cytotoxicity (Fig. 4g,h) provide strong evidence that enhanced proteasome activity may relieve oxidative stress from cells. Interestingly, K33-linked polyubiquitin chains, whose biological role has only been studied 50 , were also significantly increased in a3DN cells (Fig. 5f). This atypical Ub-linkage type was reported to take only a small portion of the whole ubiquitome in the cell and to be not significantly accumulated after proteasome inhibitor treatments, unlikely other Ub-linkage types 51 . We speculate that K33-linked polyubiquitin chains may function as a sensor of proteasome activity with non-proteolytic consequences, perhaps responding to massive changes of UPS substrates. Collectively, we found that opening the CP gate of proteasomes resulted in global but tolerable proteomic changes in mammalian cells. It has been suggested that proteasomes function under tonic inhibitory states under normal conditions 5,52 . Therefore, our proteomic data further indicate enhancing proteasome activity may be a potentially beneficial intervention for cells under mild proteotoxic or oxidative stress.

Discussion
Here we report that deletion of the a3 subunit's N-terminal tail resulted in activation of mammalian proteasomes, which showed significant increase in hydrolysis of fluorogenic substrate suc-LLVY-AMC and in degradation of Ub-Sic1 PY proteins in vitro. Opening CP gate enhanced the activity of both free CP and proteasome holoenzymes with translocation-competent conformations, implicating that the gating system may function as a critical regulator of the substrate translocation rates from the RP to the catalytic core. Because the proteasome is a major degradation machinery that regulates the levels of toxic, aggregation-prone proteins and their pathological accumulation 53 , enhancing proteasome activity through gate opening may be beneficial to suppress toxicity and related pathophysiology of proteotoxic diseases, such as Alzheimer's disease 54,55 . We observed that cells expressing a3DN proteasomes had reduced levels of tau proteins and their aggregates. In addition, mammalian cells with open-gated proteasomes effectively promoted cell survival against ROS-mediated oxidative stress. Considering that CP gate opening was tolerable to cells, the present strategy could be an effective approach to study the regulatory mechanisms of mammalian proteasomes, to identify the molecular link between proteasome activity and autophagic flux, and to modulate the levels of aggregation-prone proteins in the cell. The application of hyperactive proteasomes is actually not limited to neurodegenerative diseases, because numerous other diseases are caused by toxic, misfolded, oxidized, aggregation-prone proteins 56,57 . Thus, hyperactive proteasomes with open-gate mutation may have a potentially beneficial effects for cells under various proteotoxic or oxidative stress.

Methods
Plasmids. Plasmids expressing a3, a3DN, a3-flag and a3DN-flag were generated by PCR amplification using specific primers. The PCR products encoding a3 derivatives were digested using restriction endonucleases BamH1 and Xba1. The products were then inserted into the corresponding sites of the pcDNA3.1 plasmid, and digested with the same restriction enzymes to construct the pcDNA3.1-a3 derivatives. The plasmids were then transformed into bacterial strain DH5a to screen for recombinant plasmids. These recombinants were identified by DNA sequencing. Plasmid DNA was prepared and purified using a plasmid midi kit (GeneAll, Korea), according to the manufacturer's instructions, and stored at À 20°C until use. Arg-GFP, RGS4-GFP and LC3-GFP plasmids were previously generated 58  RT-PCR. Total RNA from cultured cells was prepared using TRIzol reagent (Invitrogen), followed by further purification through RNeasy mini-columns (Qiagen, USA) with on-column DNase I treatment. cDNA samples were prepared by reverse transcription using Accupower RT-pre mix (Bioneer, Korea). Endogenous a3 was amplified by PCR using forward (5 0 -ATGTCTCGAAGATAT GACTCCAG-3 0 ) and reverse primers (5 0 -CTATTTATCCTTTTCTTTCTGT TC-3 0 ). Exogenous a3DN-flag was amplified using forward (5 0 -ATGATATTTTC TCCAGAAGGTCGCTTAT-3 0 ) and reverse primers (5 0 -CTACTTGTCGTCAT CGTCTTTGTAGTCTTTA-3 0 , which is on the C-terminal flag tag). Amplified DNA was visualized by using ethidium bromide after agarose gel electrophoresis.
Purification of the 26S human proteasome and a3DN-proteasome. Human proteasomes and a3DN proteasomes were affinity-purified from a stable HEK293 cell line harbouring biotin-tagged human b4, as previously described, with slight modifications 37 . The cells were cultured in 15-cm culture dishes, collected in lysis buffer (50 mM NaH 2 PO 4 (pH 7.5), 100 mM NaCl, 10% glycerol, 5 mM MgCl 2 , 0.5% NP-40, 5 mM ATP and 1 mM DTT) containing protease inhibitors, and homogenized using a Dounce homogenizer. After centrifugation, the supernatants were incubated with streptavidin agarose resin (Millipore, Billerica, MA) for 5 h at 4°C. The beads were washed with lysis buffer and tobacco etch virus buffer (50 mM Tris-HCl (pH 7.5) containing 1 mM ATP and 10% glycerol). The 26S proteasomes were eluted from the resin by incubating with TEV protease (Invitrogen) in TEV buffer containing 1 mM ATP for 1 h at 30°C and were concentrated using an Amicon ultra-spin column (Millipore).
Measurement of proteasome activity with fluorogenic peptide substrates. Hydrolysis of fluorogenic substrates suc-LLVY-AMC Boc-LRR-AMC and Z-LLE-AMC was measured to determine the proteolytic activity of the chymotrypsin-like, trypsin-like and caspase-like sites of proteasomes, respectively. For example, a suc-LLVY-AMC hydrolysis assay was carried out using 0.5 nM purified proteasome and 12.5 mM of suc-LLVY-AMC (Enzo Life Sciences). The reaction mixture contained 50 nM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mg ml À 1 BSA, 1 mM ATP and 1 mM DTT. Proteasome activity, when it is in the engaged conformation, was measured in the presence of 25 nM unmodified or ubiquitinated proteins, and ATPgS was used instead of ATP. Proteasomal activity was monitored by measuring free AMC fluorescence in a black 96-well plate using a TECAN infinite m200 fluorometer.
In vitro ubiquitination of Sic1 and Ub-Sic1 degradation. Polyubiquitinated Sic1 with PY motifs (Ub-Sic1 PY ) was prepared as previously described 29  Oxidized protein assays. Oxidized proteins were detected using the OxyBlot protein oxidation detection kit (Millipore). Briefly, total proteins from cells were isolated after treatment with 25 mM of menadione for 2 h, and 15 mg of protein was used for derivatization with 2-4-dinitrophenyl hydrazine for 25 min. Samples were resolved by SDS-PAGE and anti-DNP antibody was used for subsequent immunoblotting.
Mass spectrometry analysis. Protein samples were prepared from wild-type and a3DN cells in three separate 150 mm dishes. Cells were washed three times with ice-cold PBS, then scraped in PBS, spun down and lysed in 8 M urea lysis buffer (8 M urea, 75 mM NaCl, 50 mM HEPES pH 8.0, with added Complete protease inhibitors (Roche) and PhosSTOP phosphatase inhibitors (Roche)). Cell debris was spun down for 10 min at 13,000 r.p.m. at 4°C, after which protein concentrations were determined using the BCA assay (Thermo Fisher Scientific). Subsequently, 400 mg of lysate was reduced with 5 mM TCEP (tris(2-carboxyethyl)phosphine) for 30 min and alkylated with 14 mM iodoacetamide for 30 min in the dark. Proteins were precipitated using methanol/chloroform precipitation and resuspended in digestion buffer (8 M urea, 50 mM HEPES pH 8.5 and 1 mM CaCl 2 ). The protein extracts were diluted to 4 M urea, after which they were digested for 2 h at 37°C with LysC (Wako) at a 1:250 LysC/protein ratio. They were then further diluted to 2 M urea, and incubated overnight at 37°C with LysC. The next day, urea was further diluted to 1 M and trypsin (Promega) was added at a 1:50 trypsin/protein ratio for 6 h at 37°C. The samples were acidified with formic acid (FA) to a pH of o2, and then desalted using Sep-Pak C18 solid-phase extraction cartridges (Waters). Peptide concentrations were determined using the micro-BCA assay (Thermo Fisher Scientific), after which the samples were labelled with the 6-plex TMT reagents (Thermo Fisher Scientific). TMT labelling and subsequent MS analysis were performed largely as described previously 59 . Briefly, 0.8 mg of TMT reagents was dissolved in 40 ml anhydrous acetonitrile (ACN) and 10 ml was added to 100 mg peptides in 90 ml of 200 mM HEPES, pH 8.5. After 2 h, the reaction was quenched with 8 ml of 5% hydroxylamine (Sigma). Labelled peptides were combined at a ratio of 1:1:1:1:1:1 for the six channels, acidified with FA, diluted to a final concentration of 3% ACN, and then desalted with a Sep-Pak column. The peptides were then subjected to basic-pH reverse-phase HPLC fractionation as described 60 and fractionated into 24 fractions. Half of these fractions were dissolved in 3% FA/3% ACN, desalted via StageTip, dried in a SpeedVac, and then dissolved in 8 ml of 3% FA/3% ACN for LC-MS/MS analysis on an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific) as described previously 61 . Briefly, peptides were separated on an in-house packed column using a gradient of 85 min from 6 to 24% ACN in 0.125% FA at 575 nl per minute. FTMS1 spectra were collected at a resolution of 120k with a maximum injection time of 100 ms and a 200k automated gain control (AGC) target. A top-10 method was used to select the 10 most intense ions for MS/MS. ITMS2 spectra were collected with a maximum injection time of 150 ms with an AGC target of 4k and CID collision energy of 35%. FTMS3 spectra were collected using the multi-notch method described previously 59 to reduce interference and to increase quantitative sensitivity and accuracy. In brief, synchronous-precursor selection was used to include 10 MS2 fragment ions in the FTMS3 scan. To create TMT reporter ions, the higher-energy collisional dissociation collision energy was set at 55%. An AGC target of 50k and maximum injection time of 250 ms were used. Mass spectra were processed using an in-house software pipeline as described previously 60 . In short, mass spectra were searched against the human Uniprot database (February 2014) and a reverse decoy database. Precursor ion tolerance was set at 20 p.p.m. and product ion tolerance at 0.9 Da. Addition of a TMT tag ( þ 229.1629 Da) on lysine residues and peptide N-termini, and cysteine carbamidomethylation ( þ 57.0215 Da) were added as static modifications, and methionine oxidation ( þ 15.9949 Da) was set as a variable modification. A separate search was done for the ubiquitin linkages, in which a differential modification of þ 114.0429 Da for the GG-peptide on lysine residues was added. False discovery rate was set at 1%, and peptide spectral match filtering was performed using linear discriminant analysis as described previously 60 . After exporting the protein quantification values, the data was further analysed in Excel and Perseus 1.5.1.6. For protein quantitation, the signal-to-noise values for each reporter ion channel were summed across all quantified peptides, and then normalized assuming equal peptide loading across all samples. A two-tailed t-test was then performed to identify significantly changed proteins between the wildtype and a3DN cells triplicates, after which the P values were corrected for multiple testing using the Benjamini-Hochberg method 62 . For the ubiquitin linkage searches, GG-sites were localized using a modified version of the Ascore algorithm 63 , using a localization threshold of 13. The relative ubiquitin linkage abundance was determined by normalizing the quantified linkage-specific peptide to the amount of total ubiquitin in each channel. Gene ontology analysis was performed using the DAVID Bioinformatics Resource 6.7 functional annotation tool (http://david.abcc.ncifcrf.gov/) 64,65 . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 66 partner repository with the data set identifier PXD003577.
Statistical analysis. Statistical significance of difference between various groups was determined by one-way analysis of variance followed by the Bonferroni post hoc test in most data. Differences were considered to be significant Po0.05. The Michaelis-Menten kinetic parameters were obtained by fitting the experimental data to a nonlinear regression model, using GraphPad Prism 5 (GraphPad Inc.).