ATP Binding by Proteasomal ATPases Regulates Cellular Assembly and Substrate-induced Functions of the 26 S Proteasome*

Background: ATPase subunits mediate 26 S proteasome assembly and function. Results: Defective ATP binding by some ATPase subunits inhibits proteasome assembly, and proteasomes harboring an ATP binding-defective subunit are impaired for proteolysis, substrate-stimulated gating, and ATPase activity. Conclusion: The proteasome requires normal ATP binding for assembly and function. Significance: Protein substrates promote their own degradation by inducing ATP-dependent proteasome functions. We examined the role of ATP binding by six different ATPase subunits (Rpt1–6) in the cellular assembly and molecular functions of mammalian 26 S proteasome. Four Rpt subunits (Rpt1–4) with ATP binding mutations were incompetent for cellular assembly into 26 S proteasome. In contrast, analogous mutants of Rpt5 and Rpt6 were incorporated normally into 26 S proteasomes in both intact cells and an in vitro assembly assay. Surprisingly, purified 26 S proteasomes containing either mutant Rpt5 or Rpt6 had normal basal ATPase activity and substrate gate opening for hydrolysis of short peptides. However, these mutant 26 S proteasomes were severely defective for ATP-dependent in vitro degradation of ubiquitylated and non-ubiquitylated proteins and did not display substrate-stimulated ATPase and peptidase activities characteristic of normal proteasomes. These results reveal differential roles of ATP binding by various Rpt subunits in proteasome assembly and function. They also indicate that substrate-stimulated ATPase activity and gating depend on the concerted action of a full complement of Rpt subunits competent for ATP binding and that this regulation is essential for normal proteolysis. Thus, protein substrates appear to promote their own degradation by stimulating proteasome functions involved in proteolysis.

Intracellular protein degradation requires metabolic energy. Although the cleavage of peptide bonds per se is exergonic, protease machines found in all domains of life use ATP binding and hydrolysis to regulate their function (1,2). In eukaryotes, most ATP-dependent intracellular protein degradation is catalyzed by the 26 S proteasome, a 2.5-MDa complex that degrades polyubiquitylated proteins and certain non-ubiquitylated proteins (3,4). ATP is likely used to promote structural features of the proteasome required for proteolysis and to support and integrate multiple events of substrate processing prior to peptide bond cleavage. However, a comprehensive understanding of how ATP fuels the 26 S proteasome for proteolysis is lacking.
The 26 S proteasome is composed of a cylinder-shaped 20 S proteolytic complex bound at one or both ends to a PA700 (19 S) ATPase regulatory complex (5). The interface of these complexes consists of axially abutting rings of 20 S and PA700 subunits (6,7). The center of the heteroheptameric 20 S ring forms a narrow pore for entry of substrates to the interior of the fourring 20 S cylinder where protease active sites reside (8,9). The pore is reversibly gated by flexible N termini of the 20 S subunits (10). The hexameric PA700 ring is composed of homologous ATPases (termed Rpt1-6) of the AAA protein family (11). AAA proteins feature conserved Walker A and Walker B domains that bind and hydrolyze ATP, respectively (12,13). C-terminal residues of several Rpt subunits induce an open gate conformation upon binding to pockets between 20 S ␣ subunits (14 -16). Gate opening can be induced directly by binding of isolated of C-terminal peptides of gating-competent Rpt subunits in the absence of ATP (5,17) but requires ATP binding in intact proteasomes (18 -20). Yeast 26 S proteasomes containing an ATP binding-defective mutant Rpt2 subunit display defective gating (21). Analogous mutants of other Rpt subunits have disparate effects on gating and on other features of proteasome function, supporting the general conclusion that the six homologous Rpt proteins have largely non-redundant roles (22).
Rpt subunits have other critical structural and functional features required for normal 26 S proteasome action. For example, the Rpt subunit ring interacts with several PA700 subunits oriented proximally to the 20 S complex and forms an interface with another set of PA700 subunits oriented distally to the 20 S proteasome (6,23,24). These various non-ATPase PA700 subunits provide the 26 S proteasome with polyubiquitin chainbinding sites (25)(26)(27)(28), deubiquitylating activities (29 -32), and docking sites for reversibly associated regulatory proteins (26,33,34). Thus, the Rpt subunit ring physically bridges the 20 S proteasome and elements of PA700 that prepare substrates for degradation. This topological feature enables the Rpt ring to use its ATP binding and hydrolyzing functions to coordinate and integrate multiple and diverse processes required for pro-teasomal proteolysis. For example, ATP binding and hydrolysis regulate polyubiquitin chain binding and subsequent substrate engagement (35). Subsequent rounds of ATP binding and hydrolysis probably promote alternating conformations of the Rpt ring that transmit mechanical force to substrates for substrate unfolding and translocation to the 20 S proteasome (36,37). Substrate processing also requires disassembly of the polyubiquitin chain, which cannot traverse the narrow 20 S substrate entry pore. As with peptide bond hydrolysis, deubiquitylation per se is not ATP-dependent but becomes ATP-dependent when integrated with overall substrate degradation (30,32). This feature may reflect an enforced mechanistic linkage of deubiquitylation to other steps in substrate processing to prevent loss of binding affinity prior to committed degradation (35).
In addition to regulating the function of intact 26 S proteasome, Rpt subunits play important roles in 26 S proteasome assembly. Interactions between C-terminal residues of Rpt subunits and 20 S proteasome are critical determinants of 26 S proteasome formation (38 -41). Certain Rpt subunits lacking C-terminal residues cannot assemble into 26 S proteasome in intact cells. Moreover, removal of C-terminal residues from any of several Rpt subunits in purified PA700 prevents its assembly into 26 S proteasome in an in vitro assembly reaction with 20 S proteasome (5). 26 S proteasome assembly in this reaction depends on ATP binding but not hydrolysis (5). This effect may reflect ATP-induced conformational changes in Rpt subunits that optimize the availability of C-terminal residues for 20 S proteasome binding. Such a mechanism appears to be conserved among other Rpt-like ATPases and their interactions with proteasomes. For example, PAN, an archaebacterial AAA protein ortholog of eukaryotic Rpt subunits, also requires ATP binding to associate with and promote gating of 20 S proteasome in vitro (18). Nevertheless, several aspects of the physiologic significance of ATP-dependent proteasome assembly remain unclear. First, the exact cellular pathway of 26 S proteasome assembly remains undefined (42). Despite some supporting evidence, it is not unambiguously established that 26 S proteasome is assembled in cells by binding of intact PA700 to 20 S proteasome. Second, conflicting results regarding the role of ATP binding by Rpt subunits in the cellular assembly of 26 S proteasome have been reported. For example, ATP binding mutations in individual yeast Rpt subunits produced a range of effects that vary according to the subunit (22). As with other features of proteasome function (43), these findings suggest non-redundant roles for Rpt subunits in ATP-dependent proteasome assembly. In contrast, a recent report indicated that ATP binding mutations of any Rpt subunit blocked 26 S proteasome assembly in mammalian cells (44).
This work had two related initial goals: first, to determine the general and individual roles of ATP binding by Rpt subunits in 26 S proteasome assembly in intact mammalian cells and second, to evaluate the physiologic significance of ATP-dependent in vitro assembly of 26 S proteasome from purified 20 S proteasome and PA700. Our results show that defective ATP binding differentially affected cellular proteasomal assembly of Rpt subunits and unexpectedly allowed us to study the role of ATP binding in functions of purified 26 S proteasomes harboring either of two different single ATP binding-defective Rpt subunits. Results of these studies reveal critical roles of ATP binding in multiple aspects of proteasome function and provide insights to molecular mechanisms of ATP-regulated proteasome activity.

EXPERIMENTAL PROCEDURES
DNA Constructs-cDNAs encoding each full-length wildtype human Rpt subunit (Rpt1-6; PSMC2, PSMC1, PSMC4, PSMC6, PSMC3, and PSMC5, respectively) were prepared as described previously (39). For each cDNA, a conserved lysine residue in the Walker A motif (GPPGTGKT) known to be essential for ATP binding was mutated to alanine (45,46). The mutant cDNA was subcloned into pIRESpuro3 expression vector (Clontech) featuring an N-terminal FLAG epitope (supplemental Fig. 1). Each construct was confirmed by DNA sequencing. As reported previously, wild-type Rpt2 was modified, apparently by proteolysis, upon stable expression in HEK293 cells (39). In contrast, Rpt2 lacking the last three (C-terminal) residues was expressed without modification and was assembled normally into 26 S proteasome. Therefore, the C-terminally truncated construct was used as the "wild-type" control for Rpt2 bearing an ATP binding mutation.

Preparation of HEK293 Cell Lines with Stable Expression of FLAG-tagged Rpt Subunits Containing a Walker A ATP Binding
Mutant-HEK293 cells stably expressing wild-type or ATP binding mutant FLAG-Rpt subunits were established and cultured as described previously (39). Cells expressing these proteins were morphologically indistinguishable from and grew at rates similar to those of normal HEK293 cells.
Preparation of Cell Extracts-HEK293 cell lines were grown to ϳ90% confluence, harvested, and washed with phosphatebuffered saline. Cells were disrupted in ice-cold buffer consisting of 50 mM Tris-HCl, pH 7.5 at 4°C, 0.05% Nonidet P-40, 1 mM ATP, 5 mM MgCl 2 , and 1 mM ␤-mercaptoethanol by 15 passages through a 27-gauge needle. The lysates were centrifuged to remove debris to obtain a crude soluble fraction. Expression of Rpt proteins was determined by Western blot analysis using anti-FLAG M2 antibody (Sigma) and corresponding anti-Rpt antibodies.
Affinity Purification of FLAG-Rpt Protein Complexes-FLAG-Rpt protein complexes were affinity-purified on anti-FLAG beads as described previously (39).
Purification of Proteasome Complexes from Bovine Red Blood Cells-26 S proteasome, latent 20 S proteasome, and PA700 were purified from bovine red blood cells as described previously (20,47,48).
Assay of 20 S Proteasome Activation by PA700-Proteasome activating activity by PA700 binding was assayed as described previously (20). In brief, purified PA700 was incubated with 1 nM purified latent 20 S proteasome from bovine red blood cells. After incubation for 30 min in 45 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 200 M ATP, and 5 mM MgCl 2 in a final volume of 50 l, proteasome activity was assayed by addition of 200 l of Suc-Leu-Leu-Val-Tyr-AMC to a final concentration of 100 M. Control assays included exclusion of either PA700, 20 S proteasome, or both.
Preparation of Polyubiquitylated Sic and Unanchored Polyubiquitin Chains-Polyubiquitylated His 6 -Sic was prepared as described previously (49). In brief, His 6 -Sic containing a PY (Pro-Tyr) motif fused to the C terminus was expressed in Escherichia coli and purified by affinity chromatography on nickel beads. The purified His 6 -Sic was incubated with purified recombinant E1, E2 (Ubc4), E3 (Rsp5), and ubiquitin. A high molecular weight fraction of polyubiquitylated Sic (M r Ͼ 150,000) was isolated by glycerol density gradient centrifugation and dialyzed extensively against a buffer consisting of 20 mM Tris-HCl, pH 7.6 and 100 mM NaCl. Unanchored polyubiquitin chains were prepared by incubation of E1, GST-E2 25K , and ubiquitin. Unanchored chains were separated from GST-E2 25K by chromatography on glutathione beads. A high molecular weight fraction of polyubiquitin chains (M r Ͼ 100,000) was prepared by glycerol density gradient centrifugation.
Assay of ATPase Activity-ATPase activity was determined by measuring the rate of hydrolysis of ␥-32 PO 4 from [␥-32 P]ATP (50). Assays contained 45 mM Tris-HCl, pH 7.8, 11 mM dithiothreitol, 10 mM MgCl 2 , 0.2 mM ATP, and proteins specified in individual experiments in a final volume of 25 l. After incubation at 37°C, free ␥-32 PO 4 was quantified. All assays were conducted under conditions in which the rate of phosphate hydrolysis was linear with respect to time and enzyme concentration. To compare the specific ATPase activities directly among various FLAG-protein complexes, we isolated affinity-purified complexes by glycerol density gradient centrifugation and verified their identities by Western blotting for FLAG-Rpt subunits and other constituent complex subunits. The content of protein complexes in assays was normalized to FLAG content and at least one other constituent subunit, typically Rpt2. Absolute ATPase activities for given complexes varied modestly among different experiments possibly because of variations in calculation of specific radioactivity of the ATP substrate and/or in estimates of protein content among various preparations. All direct comparisons of ATPase activity among complexes were are made within the same experiment.
Polyacrylamide Gel Electrophoresis-Native and SDS-polyacrylamide gel electrophoresis were conducted as described previously (20). Native gels were either stained with silver, blotted with anti-FLAG antibodies, or overlaid with Suc-Leu-Leu-Val-Tyr-AMC substrate for in-gel detection of proteasome activity. SDS gels were stained for protein with silver or blotted with antibodies against proteins indicated in specific experiments.
Comparison of Specific Enzymatic Activities among Proteasome Complexes-To determine and compare specific activities of enzymatic functions such as ATPase activity and various protease activities among various wild-type and mutant FLAGcontaining proteasome complexes, affinity-purified samples were subjected to glycerol density gradient centrifugation. Gradient fractions were analyzed by Western blotting for the distribution pattern of FLAG protein and for other selected proteasome subunits such as Rpt2, Rpt 5, and Rpn12. Gradient fractions also were assayed for ATPase activity and peptide hydrolyzing activity using the substrate Suc-Leu-Leu-Val-Tyr-AMC. The combined protein and activity distribution patterns permitted identification of gradient fractions containing 26 S proteasome and/or PA700 based on the known sedimentation positions of these complexes. This procedure separated 26 S proteasome and PA700 from one another and from any FLAGcontaining complexes not incorporated into either of them. Proteasome complexes containing equal content of representative subunits based on Western blotting and independent protein determination were assayed for ATPase activity, peptidase activity, and protease activity with protein substrates. Parallel analysis and comparison were conducted with highly purified 26 S proteasome or PA700 from bovine red blood cells.

Rpt Subunits Defective for ATP Binding Are Differentially
Assembled into 26 S Proteasome-To evaluate the role of ATP binding by Rpt subunits in the cellular assembly of 26 S proteasome, we expressed FLAG-tagged versions of individual subunits as either a wild-type or an ATP binding-defective mutant protein in HEK293 cells (supplemental Fig. 1). We compared the fate of each mutant Rpt subunit with that of its wild-type counterpart and with the fates of other mutant Rpt subunits with respect to 26 S proteasome assembly. With the exception of Rpt4, we engineered and selected cells that stably expressed given FLAG-Rpt subunit pairs at levels approximately equal to one another and to the corresponding endogenous protein (supplemental Fig. 2). We were unsuccessful in generating HEK293 cells that stably expressed mutant Rpt4 at levels sufficient for further analysis because the mutant protein appeared to be unstable and was rapidly degraded after expression (data not shown). Nevertheless, mutant FLAG-Rpt4 was produced at suitable levels during transient expression, and this method was used for analysis of this protein.
As an initial assessment of 26 S proteasome assembly, we subjected cell extracts to glycerol density gradient centrifugation. The 26 S proteasome and its separate 20 S proteasome and PA700 component subcomplexes sediment at characteristic and distinguishable positions by this method (31). Thus, the distribution profile of a FLAG-tagged protein in gradient fractions reflects the steady-state presence of the protein in these protein complexes. As reported previously, FLAG-tagged wildtype Rpt subunits were assembled into intact 26 S proteasome albeit with different efficiencies (39). For example, nearly all wild-type FLAG-Rpt3 and FLAG-Rpt6 were found in 26 S proteasome, whereas other FLAG-Rpt proteins were found both in 26 S proteasome and to variable extents in slower sedimenting complexes including both intact PA700 and smaller complexes that are likely intermediates of PA700 assembly (Fig. 1). ATP binding mutations had differing effects on the incorporation of the six Rpt subunits into the 26 S proteasome. Mutants of Rpt1, Rpt2, Rpt3, and Rpt4 were largely excluded from gradient fractions characteristic of 26 S proteasome and in each case accumulated in slower sedimenting gradient fractions characteristic of intact PA700 and/or of smaller complexes (Fig. 1A). In contrast to the assembly defects displayed by these mutant subunits, mutant Rpt5 was assembled into 26 S proteasome and featured a distribution profile similar to that of wild-type Rpt5. Mutant Rpt6 also assembled into 26 S proteasome although not as efficiently as its wild-type counterpart. Thus, although an appreciable portion of the mutant Rpt6 had a distribution profile characteristic of 26 S proteasome, a significant portion was shifted to gradient fractions characteristic of PA700 (Fig. 1B). In sum, these results demonstrate that Rpt subunits have different and characteristic requirements for ATP binding for their steady-state assembly into 26 S proteasome. Notably, mutant Rpt subunits defective in 26 S proteasome assembly appeared to be competent for assembly into intact PA700 (see below).
Characterization of Purified Protein Complexes Containing ATP Binding-defective Rpt Subunits-To characterize the structural and functional features of complexes containing FLAG-Rpt subunits, we purified wild-type and mutant FLAG-Rpt proteins using anti-FLAG beads and then compared the properties of the resulting protein complexes. Affinity purification of each wild-type FLAG-Rpt subunit yielded structurally intact and functionally active 26 S proteasomes as judged by multiple criteria. For example, these complexes migrated on native PAGE as a mixture of singly and doubly capped 26 S proteasomes ( Figs Similar analysis of affinity-purified mutant FLAG-Rpt5 and mutant FLAG-Rpt6 demonstrated that each of these proteins also was incorporated into structurally intact 26 S proteasomes (Figs. 2 and 3). These results are consistent with those obtained with crude cell extracts and confirm that ATP binding mutants of either Rpt5 or Rpt6 are competent for 26 S proteasome assembly. Smaller portions of these mutant subunits were found in other complexes including intact PA700. These complexes were obvious by distribution patterns in glycerol gradient fractions (Figs. 2C and 3C) but were not well resolved by native PAGE. Further characterization of 26 S proteasomes and PA700 containing mutant Rpt5 and mutant Rpt6 is presented below.
Affinity purification of ATP binding mutants of FLAG-Rpt1-4 isolated no or very low levels of intact 26 S proteasome as expected from results of glycerol density gradient centrifugation of corresponding crude cell extracts (Figs. 4 and 5 and supplemental Figs. 3 and 4). Instead, these mutant FLAG-Rpt subunits were associated with a heterogeneous group of protein complexes including PA700. Comparison of the distribution profiles of individual mutant subunits on density gradient centrifugation before and after affinity chromatography suggested that some complexes present in the initial extract were unstable and dissociated during affinity purification. For example, most mutant FLAG-Rpt1 sedimented significantly more slowly after affinity purification than before purification (Figs. 1A and 5C). Likewise, the recovery of mutant FLAG-Rpt2 as intact PA700 was poor after affinity purification (supplemental Fig. 4 and data not shown). We speculate that salt conditions used for elution of unbound proteins during affinity chromatography promote dissociation of complexes with lower stabilities due to the ATP binding mutations. In any case, these results highlight an important role for ATP binding by Rpt1-4 in the formation and stability of 26 S proteasome and distinguish these subunits from Rpt5 and Rpt6 for which the ATP binding defects had little effect on 26 S proteasome assembly and stability.

ATP Binding Mutants Have Different Effects on in Vitro Assembly of 26 S Proteasome and Mimic Their Cellular Roles-
To further evaluate the significance of these distinctions among Rpt subunits for ATP binding-dependent 26 S proteasome assembly, we purified intact PA700 from cells expressing Rpt subunits of each category and tested PA700 function in an in   FEBRUARY 1, 2013 • VOLUME 288 • NUMBER 5 vitro 26 S assembly assay with purified 20 S proteasome. As expected, PA700 from cells expressing wild-type Rpt subunits activated proteasome hydrolysis of peptide substrates, a functional indication of 26 S proteasome assembly, and did so comparably with activation by purified bovine PA700 (Fig. 6). PA700 isolated from cells expressing ATP binding mutants of Rpt5 or Rpt6, subunits that were incorporated normally into 26 S proteasome in intact cells, activated the 20 S proteasome to an extent similar to PA700 containing their wild-type counterparts. In contrast, PA700 containing ATP binding mutants of Rpt1 or Rpt3 had no detectable activating effect on proteasome activity. Thus, ATP binding mutant Rpt subunits that were either competent or incompetent for assembly into 26 S proteasome in intact cells mimicked their respective assembly competencies in this in vitro assembly assay. These results suggest that ATP binding by different individual Rpt subunits plays different roles in 26 S proteasome assembly and are consistent with a model in which cellular 26 S proteasome is normally assembled by the binding of intact PA700 to 20 S proteasome (see "Discussion").

S Proteasome and PA700 Containing a Single ATP Binding Mutant Rpt Subunit Have Near Normal ATPase Activities-To
determine the effect of ATP binding-defective subunits on functions of proteasome complexes, we purified complexes containing wild-type and mutant Rpt subunits and compared various activities. Mutant FLAG-Rpt5 and mutant FLAG-Rpt6 were each assembled into both 26 S proteasome and PA700. Proteasome complexes containing either of these mutant subunits had ATPase specific activities indistinguishable from those of complexes containing their respective FLAG wild-type subunits and from those of 26 S proteasome and PA700 purified from bovine red blood cells (Table 1). Thus, the presence of a single ATP binding-defective Rpt subunit did not significantly affect the ATPase function contributed by the five remaining wild-type subunits of the complexes.
ATP binding mutants of other FLAG-Rpt subunits did not assemble into 26 S proteasome but did assemble into PA700 to variable extents. To determine whether mutant Rpt subunits defective for 26 S proteasome assembly conferred different ATPase properties on intact PA700 compared with their assembly-competent mutant counterparts, we measured ATPase activity in two purified assembly-defective PA700 complexes. PA700 containing either mutant FLAG-Rpt1 or FLAG-Rpt3 had ATPase activities similar to their wild-type counterparts or to PA700 from bovine red blood cells. Collectively, these results demonstrate that the presence of a single ATP binding-defective subunit in these complexes has little effect of the ability of the five normal ATPase subunits to hydrolyze ATP and suggest that basal ATPase activity of proteasome complexes does not require a mechanism in which Rpt subunits are obligately coupled (see below).

S Proteasomes Containing a Single ATP Binding Mutant Rpt Subunit Are Defective in the Degradation of Ubiquitylated and Non-ubiquitylated Substrates but Not of Short Peptides-
To further analyze effects of Rpt subunits with ATP binding defects on functions of proteasome complexes, we compared protease activities of 26 S proteasomes with wild-type or ATP binding mutant Rpt subunits. As suggested by the initial characterization, 26 S proteasomes containing either mutant Rpt5 or mutant Rpt6 featured peptidase activities comparable with those of normal 26 S proteasomes (Fig. 7A). These results indi- PA700s containing either WT FLAG-Rpt subunits or corresponding ATP binding mutant (Lys to Ala) FLAG-Rpt subunits were purified by glycerol density gradient centrifugation of anti-FLAG affinity-purified samples. Equivalent amounts of PA700 based on normalization by FLAG and Rpt2 content were assayed for ATP-dependent activation of latent 20 S proteasome activity. The peptide hydrolyzing activity of 20 S proteasome in the absence of PA700 (20S) was assigned a value of 100, and activities in the presence of PA700 (ϩPA700) with the indicated FLAG-Rpt protein are expressed as a percentage of that value. 20 S proteasome activation by an equivalent amount of purified bovine PA700 ("Bovine") was determined for comparison. Data represent mean values triplicate assays ϮS.D. (error bars). Similar results were obtained with two independent preparations of proteins.  26 S proteasomes or PA700 containing either wild-type or ATP binding mutants of the indicated FLAG-Rpt subunits were isolated by glycerol density gradient centrifugation after affinity purification on anti-FLAG beads as described under "Experimental Procedures." Bovine 26 S proteasome and PA700 were purified from bovine red blood cells. Results for the indicated individual experiments were obtained with complexes assayed concurrently. Data represent mean values of triplicate assays ϮS.D. Similar results were obtained with at least two independent preparations of each protein.

Proteasome complex
ATPase activity cate that defective ATP binding by these specific subunits has no significant effect on PA700-induced gating of the proteasome. In contrast, each mutant 26 S proteasome was severely defective for the degradation of proteins such as polyubiquitylated Sic, a model polyubiquitylated 26 S proteasome substrate, and casein, a structurally disordered protein known to be degraded by the 26 S proteasome in a ubiquitin-independent fashion (Fig. 7, B and C). Thus, despite their near normal rates of ATP hydrolysis and an open substrate entry gate, 26 S proteasomes harboring a single ATP binding-defective subunit were crippled for degradation of protein substrates. In the presence of proteasome inhibitors such as MG132, wild-type 26 S proteasomes deubiquitylated substrates, which accumulated as unmodified proteins (Fig. 7B). Deubiquitylation is catalyzed by multiple deubiquitylase subunits of PA700 (Fig. 7B) but is coupled to normal 26 S proteasome protease activity (20, 30 -32). 26 S proteasomes with mutant Rpt5 or Rpt6 also deubiquitylated substrates but did so at a modestly but reproducibly slower rate (Fig. 7B). Collectively, these results suggest that 26 S proteasome-catalyzed protein hydrolysis is linked to ATPase activity by a mechanism that requires coordinated function among all Rpt subunits.

S Proteasomes Containing a Single ATP Binding Mutant Rpt Subunit Are Defective in the Substrate-regulated ATPase
Activity and Gating-We previously showed that polyubiquitin stimulated peptidase and ATPase activities of 26 S proteasome, suggesting a regulatory link between substrate binding and these functions (49). Here, we have extended those findings by showing that either ubiquitylated Sic, unanchored polyubiquitin chains, or a disordered non-ubiquitylated substrate protein such as casein also stimulated ATPase activity of normal 26 S proteasome by 3-6-fold (Fig. 8). Substrate-anchored or unanchored polyubiquitin chains also stimulated proteasome gating as indicated by 3-5-fold increased peptidase activity (Fig. 8B). In contrast, tightly folded substrates refractory to degradation such as GFP and titin I27 had no effect on either activity (Fig.  8D). Substrate-induced stimulation of ATPase did not appear to depend on or be obligately linked to proteolysis because stimulation occurred similarly with proteasomes inhibited by MG132 or epoxomicin ( Fig. 8E and data not shown). These results suggest that substrate binding or engagement rather than substrate hydrolysis increased ATPase activity and proteasome gating. Interestingly, protein substrates did not stimulate ATPase activity of isolated PA700 even though PA700 likely binds substrates similarly to 26 S proteasome (Fig. 8A). Thus, physical interaction between PA700 and 20 S proteasome may be required for substrate-induced stimulation of ATPase activity. To determine the role of ATP binding by individual Rpt subunits in substrate-induced stimulation of ATPase and peptidase activities, we repeated these experiments with 26 S proteasomes containing ATP binding mutants of either Rpt5 or Rpt6. Neither ATPase activity nor peptidase activity was stimulated by protein substrates or unanchored polyubiquitin chains in these mutant 26 S proteasomes (Fig. 8, B, C, and  E). These results indicate that although 26 S proteasomes with ATP binding-defective subunits have normal ATPase activity and gating properties in the absence of substrates they are unable to regulate these features in the presence of protein substrates. Defective protein degradation by these proteasomes suggests that these regulatory features are important for normal proteasome function and likely require coordinated action among a full complement of functional Rpt subunits.

DISCUSSION
Despite their structural homology, the six Rpt subunits of the 26 S proteasome appear to play different roles in multiple aspects of proteasome function. For example, the C termini of Rpt subunits have different roles in 26 S proteasome activation and cellular assembly (38 -40, 51). Here, we show that defective ATP binding has different effects on 26 S proteasome assembly for various Rpt subunits. Several previous studies explored the role of ATP binding by Rpt subunits in cellular 26 S proteasome assembly. In yeast, ATP binding mutations of different Rpt subunits produced disparate effects on proteasome assembly and function (22). Although the general conclusions of those stud-ies are similar to ours, the exact pattern of assembly defects for the Rpt subunits differed perhaps because different specific mutations were used or because of differences in assembly mechanisms between yeast and mammalian cells (22). While our work was in progress, Lee et al. (44) reported the effect of ATP binding mutations on assembly of Rpt subunits into 26 S proteasomes in HeLa cells. In contrast to our results, they reported defective 26 S proteasome assembly for every mutant subunit. The basis for this discrepancy is unclear, but differences between the studies include the cell lines, methods of protein expression, and amino acid substitutions for the mutant subunits. Of these, we consider differences in stable FIGURE 8. 26 S proteasome complexes with a single ATP-binding mutant Rpt subunit are defective in substrate-activated ATPase and peptidase activities. 26 S proteasomes containing the indicated WT or ATP binding mutant FLAG-Rpt subunits were affinity-purified on anti-FLAG beads and then isolated by glycerol density gradient centrifugation. Bovine 26 S proteasome and PA700 were purified from red blood cells. A, the indicated proteasome complexes were assayed for ATPase activity in the absence (Control) or in the presence of polyubiquitylated Sic (480 nM) or casein (500 nM). B, the indicated 26 S proteasomes were assayed for ATPase in the absence (Control) or in the presence of polyubiquitylated Sic (480 nM), casein (500 nM), or unanchored polyubiquitylated chains (450 nM). C, the indicated 26 S proteasome were assayed for Suc-Leu-Leu-Val-Tyr-AMC hydrolyzing activity in the absence (Control) or presence of polyubiquitylated Sic (480 nM). D, bovine 26 S proteasome was assayed for ATPase activity in the absence (Con) or presence of the indicated proteins (500 nM). "GFP" is green fluorescent protein; "Titin I27 " is the I27 domain of titin. E, the indicated 26 S proteasomes were assayed for ATPase activity in the absence (Control) or in the presence of polyubiquitylated Sic (480 nM) and/or MG132 (100 M). In all panels, activity of the control was assigned a relative value of 1.0, and other activities are expressed relative to that. Each value represents the mean value ϮS.D. (error bar) of at least two independent experiments (and in most cases four independent experiments) in which each assay was performed in triplicate. Ub, ubiquitin.
versus transient expression to be the most likely cause of discrepant results. In fact, we noted that mutant Rpt4 behaved differently under transient and stable expression conditions (see "Results"). We also observed that some mutant Rpt subunits reduced the postassembly in vitro stability of the 26 S proteasome. Thus, specific details of in vitro manipulation and analysis of the mutant proteasomes may determine whether the mutant subunits are judged to have an assembly defect.
We undertook this work in part to evaluate the physiologic relevance of the in vitro assembly of 26 S proteasome from purified 20 S proteasome and PA700 (20,52). This reconstitution is strictly dependent on ATP binding, but the relative roles of individual Rpt subunits in ATP binding in this process are unknown. Furthermore, the significance of 26 S proteasome formation from 20 S proteasome and intact PA700 subcomplexes is uncertain because data supporting alternative models of cellular PA700 assembly into the 26 S proteasome have been presented (42). One model features formation of PA700 by sequential addition of multiple PA700 subcomplexes to 20 S proteasome, which serves as a required assembly template (40,53). By this mechanism, intact PA700 would not exist as an isolated cellular complex. A second model, compatible with our in vitro assembly reaction, involves formation of intact PA700 prior to its binding to 20 S proteasome. Support for this model includes cellular accumulation of PA700 with Rpt subunits lacking required 20 S binding elements or when 20 S proteasome content is reduced by RNAi (38,39). Moreover, PA700 can be reconstituted in vitro from three subcomplexes in the absence of 20 S proteasome (54). The current data provide additional support for the 20 S-independent model because ATP binding-defective Rpt subunits that are incompetent for 26 S proteasome assembly also accumulated in intact PA700 (Fig. 1). Finally, we purified intact PA700 containing either wild-type or mutant Rpt subunits and demonstrated that their relative abilities to reconstitute 26 S proteasome in vitro mirrored their effects on cellular 26 S proteasome assembly. These results further support direct binding of intact PA700 to 20 S proteasome in cells. Despite these data, an appreciable fraction of the intact PA700 containing certain mutant Rpt subunits was lost during affinity purification perhaps by dissociation of mutant PA700 into smaller complexes. These results suggest that PA700 assembly and stability are sensitive to ATP binding and are consistent with in vitro data in which reconstitution of intact PA700 from three PA700 subcomplexes depended on ATP binding (54). Postassembly dissociation of mutant PA700 could also account for some of the discrepancies regarding the effects of mutant Rpt subunits noted above. Our current data do not address the molecular basis for how ATP binding by various Rpt subunits is either required or unnecessary for 26 S proteasome assembly. Previous work established the role of dedicated assembly chaperones for 26 S proteasome assembly, and the function of these factors may be governed by ATP binding by their client Rpt subunits (42,55).
We were surprised that 26 S proteasomes or isolated PA700s containing a mutant Rpt subunit had basal ATPase specific activities similar to those of normal complexes. A recent study provided evidence for a highly concerted mechanism of proteasomal ATPase activity involving ordered cyclical ATP binding, ATP hydrolysis, and product release by pairs of Rpt subunits oriented opposite to one another in the ATPase ring (56). Accordingly, we expected a single defective ATP-binding subunit to significantly disrupt ATPase activity. However, the near normal ATPase activity of an Rpt ring with an ATP bindingdefective subunit indicates that a highly concerted mechanism is not an absolute requirement for ATP hydrolysis and that ATP binding and hydrolysis can occur by a stochastic mechanism characteristic of certain bacterial Clp AAA protein rings (57). Likewise, the basal level of peptide hydrolysis by 26 S proteasomes containing mutant Rpt subunits did not differ from that of wild-type proteasomes. Peptide hydrolysis is a monitor of 20 S proteasome gating and is promoted upon PA700 binding (18,20). However, gating does not appear to be a simple twostate function but rather is variably modulated by factors including the status of Rpt-bound ATP and binding of polyubiquitin by the assembled 26 S proteasome (35,49,58). Previous work demonstrated a role for ATP binding by Rpt2 in gating of yeast 26 S proteasome, an effect we were unable to evaluate here because of the assembly defect caused by ATP binding mutant Rpt2 (21). Nevertheless, our results show that ATP binding defects of either Rpt5 or Rpt6 have little effect on gating of 26 S proteasomes under basal conditions and are consistent with differential roles of various Rpt subunits in the regulation of proteasome function.
In contrast to their normal basal ATPase and peptidase activities, 26 S proteasomes containing either mutant Rpt5 or Rpt6 were severely crippled for ATP-dependent degradation of both polyubiquitylated proteins and disordered, non-ubiquitylated proteins. These results suggest that processing of protein substrates requires the coordinated action of a full complement of functional Rpt subunits. We propose that defective proteolysis by mutant proteasomes is a consequence of an impaired regulatory response to protein substrates. Proteins susceptible to proteasomal degradation as well as unanchored polyubiquitin chains enhanced both ATPase and peptidase activities of normal 26 S proteasome by up to 6-fold. However, neither of these effects occurred with 26 S proteasomes containing mutant Rpt5 or Rpt6, suggesting that these defects were closely related to the failure of proteolysis. Thus, substrate-stimulated ATPase activity and gating appear to require and may impose tight coordination and cooperation among functional Rpt subunits for protein degradation. Although this type of regulation is prevented by the presence of a single ATP binding-defective subunit, the precise mechanistic relationships among these processes are unclear, and these processes may not be obligately linked. For example, degradation of non-ubiquitylated proteins requires ATP binding but not hydrolysis (18,20), indicating that degradation is enhanced by increased gating without increased ATPase activity. Likewise, substrate-activated ATPase activity does not require concomitant protein hydrolysis because it occurred in MG132-inhibited proteasomes. Finally, both ATPase activity and gating were stimulated by unanchored polyubiquitin chains, showing that each effect can be achieved in the absence of protein hydrolysis (49). It is possible that defects in substrate-regulated gating and ATPase activity as well as in overall proteolysis are consequences of reduced substrate binding by mutant proteasomes (35). Deubiquitylation by MG132-inhibited proteasomes was modestly but reproducibly inhibited in mutant proteasomes, an effect that also could be accounted for by reduced substrate binding. Regardless of the exact relationships among these processes, our results indicate that protein substrates promote their own degradation by enhancing proteasomal functions required for their degradation. These substrate-induced functions require concerted action among a fully functional complement of Rpt subunits.
Quantitative comparison of activities among protein complexes depends on accurate measurements of enzymatic activities and protein content of the complexes. The assays for ATPase, peptidase, and protease activities have been validated by multiple criteria essential for quantification of enzyme activity. A larger source of possible error in these comparisons is the quantification of proteasome complex content. To achieve protein quantification necessary for normalization among samples, we isolated individual proteasome complexes by glycerol gradient centrifugation after affinity purification and then measured and compared levels of both FLAG-Rpt protein and other constituent subunits using standards of highly purified bovine PA700 and 26 S proteasome (supplemental Fig. 5). This approach verified the identity of proteasome complexes and ensured that measures of their levels used for calculations were not confounded by including the content of mixed complexes, unassembled Rpt subunits, or other proteins. The similar values of specific activities obtained for wild-type FLAG-Rpt proteasome complexes and highly purified bovine proteasome complexes provide additional confidence in the quantitative comparison of specific activities among various samples. Finally, comparison of complexes containing wild-type and mutant Rpt subunits must account for possible contributions of the corresponding endogenous wild-type Rpt subunits to measured activities of mutant complexes. This concern is not applicable to isolated PA700 or to 26 S proteasomes with a single PA700 cap (which typically represented about half of the 26 S proteasome content) because they should contain FLAG-tagged subunits exclusively. In contrast, doubly capped 26 S proteasomes could contain one normal and one mutant PA700. Although wild-type subunits may contribute normal activities to such "mutant" samples, the large defects in proteolysis and in substrate-stimulated ATPase and peptidase activities of these same samples indicate that mutant subunits dominate their functional features.
In summary, we have demonstrated that ATP binding by certain Rpt subunits but not each Rpt subunit is an important requirement for cellular assembly of 26 S proteasome in mammalian cells. ATP binding-defective Rpt subunits severely inhibited the ability of 26 S proteasomes to degrade protein substrates possibility as a result of loss of substrate-induced activation of ATPase activity and gating. These results indicate that proteolysis by the 26 S proteasome requires highly concerted functions among its ATPase subunits.