The Proteasome, a Novel Protease Regulated by Multiple Mechanisms*

Investigators have long recognized the potential importance of protein degradation as a mechanism for regulating levels of cellular proteins (1‐3). Nevertheless, only recent studies of the ubiquitin-proteasome pathway have shown how extensively proteolysis is involved in the regulation of crucial cellular processes such as: progression of the cell cycle, oncogenesis, transcription, development, growth and atrophy of developed tissues, flux of substrates through metabolic pathways, selective elimination of abnormal proteins, and antigen processing. The role of a single protease, the proteasome, in such a large and diverse set of events suggests that this enzyme is highly regulated. In fact, regulation of proteasome function occurs on multiple levels and is extraordinarily complex. We first will review briefly the structure of the proteasome because of its inherent role in the regulation of proteasome function. We then will describe emerging themes regarding the multiple mechanisms by which proteasome function is regulated. Details of these themes demonstrate that the proteasome operates within a novel and surprisingly complex proteolytic system that defines a new paradigm for intracellular protein degradation.

Investigators have long recognized the potential importance of protein degradation as a mechanism for regulating levels of cellular proteins (1)(2)(3). Nevertheless, only recent studies of the ubiquitin-proteasome pathway have shown how extensively proteolysis is involved in the regulation of crucial cellular processes such as: progression of the cell cycle, oncogenesis, transcription, development, growth and atrophy of developed tissues, flux of substrates through metabolic pathways, selective elimination of abnormal proteins, and antigen processing. The role of a single protease, the proteasome, in such a large and diverse set of events suggests that this enzyme is highly regulated. In fact, regulation of proteasome function occurs on multiple levels and is extraordinarily complex. We first will review briefly the structure of the proteasome because of its inherent role in the regulation of proteasome function. We then will describe emerging themes regarding the multiple mechanisms by which proteasome function is regulated. Details of these themes demonstrate that the proteasome operates within a novel and surprisingly complex proteolytic system that defines a new paradigm for intracellular protein degradation.

The 20 S Proteasome: an Intracellular Protease with Novel Structural and Functional Features
The proteasome is a 700,000-dalton, cylinder-shaped protease arranged as four axially stacked heptameric rings (4) (Fig. 1). It has multiple catalytic centers located within a hollow cavity of the cylinder. This topology sequesters the catalytic sites from potential protein substrates and has several unusual and important consequences for proteasome function and regulation. The proteasome from the archeon Thermoplasma acidophilum is a prototype for the quaternary structure and topology of the enzyme. Its 28 subunits represent two homologous gene products (␣ and ␤), each present in 14 copies per particle (4,5). The two outer rings are composed exclusively of ␣ subunits, whereas the two inner rings are composed exclusively of ␤ subunits (6) (Fig. 1). In eukaryotes, the subunits represent 14 gene products, seven of which are homologous to the archeal ␣ subunit and seven of which are homologous to the archeal ␤ subunit (7-9). The relative positions of the eukaryotic ␣-type and ␤-type subunits, respectively, are analogous to those in the archeal proteasome, and each ring contains a complete complement of the seven respective gene products (10). Thus, the proteasome is a multimeric dimer with a C2 axis of symmetry about the two inner ␤ rings. The interior of the cylinder contains a cavity consisting of three contiguous chambers joined by narrow constrictions (6,11). The central chamber is formed by abutting ␤ rings; it is connected to two identical "antechambers" formed by abutting ␣ and ␤ rings. The structure of the archeal proteasome initially suggested that the cavity might be continuous to the exterior via narrow 13-Å openings formed by the centers of the ␣ rings (6). However, the structure of the yeast proteasome subsequently showed that the openings are sealed by the N-terminal portions of the ␣ subunits (11). Nevertheless, it is likely that substrates enter the central cavity through this narrow constriction after the seal has been broken by conformational changes in the proteasome. This general structure suggests that the entrance of substrates into the proteasome is regulated much like a gated channel. Probable regulators of this channel include specific proteins that are discussed below.
The eukaryotic proteasome is a multicatalytic protease characterized by three activities with distinct specificities against short synthetic peptides: a "chymotryptic-like" activity with preference for tyrosine or phenylalanine at the P1 position; a "tryptic-like" activity with preference for arginine or lysine at the P1 position; and a "postglutamyl" hydrolyzing activity with a preference for glutamate or other acidic residues at the P1 position (12,13). The identity and nature of these catalytic sites had long remained unclear because no proteasome subunit can be classified within one of four major protease families based on either primary sequence or inhibitor profile. However, the crystal structure of the yeast proteasome (11) and related biochemical data with mutant proteasomes (14,18,19) demonstrated that each catalytic activity is linked with a specific ␤ subunit (4). These subunits have N-terminal threonine residues as catalytic nucleophiles, confirming their suspected novel nature (11,14). Thus, three of the seven different ␤ subunits in the eukaryotic proteasome are presumptive catalysts (present in 2 copies per proteasome molecule), whereas the function of the other four ␤ subunits is unknown. Detailed mechanistic features of the relative functional interactions among the catalytic subunits with respect to the hydrolysis of longer polypeptide chains are still unclear, but they seem to act in a coordinated fashion, which may alter the specificity of a given site as compared with that for short substrates (15). In vitro, the proteasome cannot degrade large proteins unless they are highly denatured and the proteasome has been "activated" by exposure to low concentrations of SDS, heat, or certain ionic conditions (13,16). The exact molecular basis for these diverse effects remains obscure but probably involves a conformational change in the proteasome that opens the central cavity to the exterior and/or allosteric alterations of the catalytic sites. Such in vitro effects are of unlikely direct physiologic significance but probably mimic proteasome activation by regulatory proteins, as described below. It is still uncertain whether the direct activation of the 20 S proteasome by non-protein factors occurs in intact cells.

Regulation of Proteasome Activity by Composition of Catalytic Subunits
In higher eukaryotes, each of the three catalytic subunits (termed ␤1 or Y or ␦; ␤2 or Z; and ␤5 or X) has a close homolog (LMP2, MECL-1, and LMP7, respectively) (cf. Ref. 4 for nomenclature). The latter proteins are selectively induced under certain conditions, including treatment of cells with ␥-interferon (17). Upon induction, these subunits are assembled into newly synthesized proteasomes in place of their respective homologs, thereby forming proteasomes with distinct subunit compositions and altered catalytic characteristics (Fig. 2). For example, proteasomes containing inducible subunits hydrolyze peptides with hydrophobic and basic amino acids at the P1 position more rapidly, but peptides with acidic amino acids at the P1 position more slowly, than do proteasomes containing the constitutive subunits. Data consistent with these features have been obtained in cells harboring mutations in one or more of these subunits and in experiments where cells have been transfected with genes for the inducible subunits (20). Nevertheless, proteasomes with inducible or constitutive subunits do not differ appreciably in the overall rates at which they degrade large ubiquitinated proteins, the presumptive physiologi-* This minireview will be reprinted in the 1999 Minireview Compendium, which will be available in December, 1999. This is the fourth article of four in the "Proteases in Cellular Regulation Minireview Series." § To whom correspondence should be addressed: . Therefore, regulation of subunit composition may serve to control the qualitative nature of proteolytic products by influencing catalytic specificity and/or the relative rates at which different catalytic sites hydrolyze various peptide bonds. Such regulation may be involved in and important for antigen processing and presentation on class I major histocompatibility complexes, which are biased toward peptides with hydrophobic residues at the C terminus. Consistent with such a specific role, two of the inducible subunits are encoded within the major histocompatibility complex class II locus (22,23), and all three are expressed in response to ␥-interferon, a known regulator of the immune response (24). Despite this attractive model, mechanistic details of how the identity, specificity, and hydrolytic rate of given catalytic sites for short peptides relate to the action of these sites on longer polypeptides remain to be determined.

Proteasome Function Is Mediated by Regulatory Proteins
Structural features of the proteasome, such as sequestration of catalytic sites, impose severe physical restrictions on its interaction with potential substrates and raise critical questions regarding how the protease degrades proteins under physiological conditions. Although the 20 S proteasome can be activated in vitro, it may never function as an isolated enzyme in cells but rather only when bound to regulatory proteins. These proteins appear to mediate proteasome function, in part, by overcoming the restrictions of proteasome structure on substrate access and therefore can be considered as proteasome activators (Fig. 3). Because the proteasome forms complexes with different proteins, it appears to be the center of a modular system in which different regulatory proteins confer unique catalytic and regulatory properties on the proteasome.

The Role of the Proteasome in the Ubiquitin Pathway Is
Mediated by the Regulatory Protein, PA700 The vast majority of known protein substrates of the proteasome must be modified by the covalent attachment of a polyubiquitin chain, which serves as a substrate-targeting and recognition signal for the proteasome. Polyubiquitination is accomplished by the sequential action of three enzymes: an ATP-dependent ubiquitinactivating enzyme (E1), 1 a ubiquitin-conjugating enzyme (E2), and a ubiquitin-protein ligase (E3). This cascade covalently links the C terminus of ubiquitin to a free amino group on the target protein, usually the ⑀-amino of a lysine residue. Conjugation of a single ubiquitin to a protein is a weak signal for degradation. However, the ubiquitination reaction is processive, and additional ubiquitin molecules are conjugated to lysine 48 of the preceding ubiquitin. Susceptibility of a protein for polyubiquitination is determined by structural features recognized by E2 and E3 proteins, sometimes in conjunction with a third protein, an example of which is the E6 protein product of papillomavirus for the ubiquitination of p53 (25). Because the E2 and E3 proteins exist as multimembered families, their combinatorial pairing with or without ancillary proteins generates the diversity and specificity required for the ubiquitination of many different proteins with many different ubiquitination signals. Moreover, such signals may be temporally exposed to the ubiquitination machinery in response to other modifications of the protein such as phosphorylation (26). Thus, entry of substrates into the ubiquitin-proteasome proteolytic pathway is regulated independently of selectivity by the proteasome.
Polyubiquitinated proteins cannot be degraded directly by the 20 S proteasome. Rather, proteolysis requires another protein, known alternatively as PA700 (27), ball (28), 19 S cap (37), or -particle (29). PA700 is a 700,000-dalton, 20-subunit complex that binds to one or both of the terminal rings of the proteasome in a cooperative manner (30,31). Formation of the 2,100,000-dalton proteasome-PA700 complex, which also can be purified directly from cell extracts and is commonly referred to as the 26 S proteasome, requires ATP hydrolysis, an inherent function of PA700 (13,(32)(33)(34). Binding of PA700 to the proteasome greatly enhances the ability of the proteasome to degrade both ubiquitinated proteins and non-ubiqutinated peptides. The degradation of the former, but not of the latter, requires continuous ATP hydrolysis beyond that required for complex formation. These results indicate that PA700 plays multiple roles in mediating proteasome function and suggest possible roles for ATP in the proteolytic process. For example, the topology of PA700 binding to the FIG. 2. Regulation of the proteasome activity by alterations in catalytic subunit composition. Proteasome activity can be regulated by altering the composition of its three catalytic ␤ subunits. The three constitutive subunits, ␤1, ␤2, and ␤5, can be replaced by LMP2, MECL-1, and LMP7, respectively, in response to certain physiological stimuli. This type of regulation alters the relative rates at which the subunits with different specificities cleave given peptide bonds and can result in qualitative differences in the peptide products. The 20 S proteasome is essentially inactive because its structure sequesters catalytic sites, depicted by black circles (q), from substrates. Binding of the proteasome to regulatory proteins such as PA700 and PA28 activates degradation of ubiquitinated proteins and peptides, respectively. Part of the activation probably involves opening of channels at the terminal rings to increase substrate access to the catalytic sites. For clarity, the figure depicts proteasomes capped by a single PA700 or PA28, although each activator can doubly cap one proteasome particle (31,47). PA700, PA28, and the proteasome may form tertiary complexes characterized by unique properties (57). Protein inhibitors can inactivate 20 S proteasome that may become activated in the cell and may also inhibit formation of proteasome-activator complexes.
terminal rings of the proteasome suggests that PA700 controls access of substrates to the active sites of the proteasome in the central cavity by inducing conformational opening of the sealed gates at the terminal rings. Because folded proteins are too large to pass through the narrow entry port, PA700 may utilize ATP hydrolysis to unfold them. Furthermore, because substrate binding is distant from the site of peptide bond cleavage and because proteolysis appears to be processive, ATP hydrolysis might be coupled to translocation of the unfolded polypeptide chain through the channel to the active sites. Neither of these ATP-dependent functions would be necessary for accelerated hydrolysis of short peptides, an explanation consistent with the different ATP requirements for degradation of proteins versus peptides by the 26 S proteasome. Although PA700-mediated acceleration of rates of hydrolysis of short peptides may be accounted for by increased access of substrate to the active sites, it is possible that PA700 also enhances proteolysis by an allosteric activation of catalytic centers (19).
ATP hydrolysis by PA700 is accomplished by one or more of six different subunits that share a homologous 200-amino acid domain with ATP binding motifs (35,36). Recent data indicate that these subunits form a ring whose face directly contacts the terminal rings of the proteasome (37). Thus, a subcomplex of PA700 containing all six "ATPases" and several other subunits binds directly to the proteasome and stimulates hydrolysis of short peptides but not of ubiquitinated proteins. Although the relative functions of the AT-Pases are unknown, mutants of individual subunits in yeast result in distinct phenotypes. These results suggest that the subunits may play unique functions with respect to the various ATP-dependent processes promoted by PA700.
PA700 confers specificity to the proteasome for ubiquitinated proteins because it serves as the recognition component for the polyubiquitin chain. This feature provides a mechanism for the targeting of diverse substrates to the 26 S proteasome via a common structural modification. Thus, the proteasome utilizes a novel mechanism for selection of and interaction with protein substrates. Unlike traditional proteases that recognize and bind to substrates based on the affinity of active sites for specific amino acids at or very near the cleaved peptide bond, the proteasome relegates substrate recognition and binding to PA700. PA700, in turn, binds to the substrate via a covalent modification (the polyubiquitin chain) rather than an inherent feature of its primary structure. One identified PA700 subunit binds polyubiquitin chains with high affinity (38). The structure of the polyubiquitin chain suggests a molecular basis for the selectivity of polyubiquitin chains over monoubiquitin for binding (39). Genetic studies in yeast indicate that this subunit cannot be the sole source of polyubiquitin chain recognition because mutants that lack it are viable and degrade most ubiquitinated proteins normally (40).
Finally, PA700 contains an isopeptidase activity that disassembles polyubiquitin chains (41). This activity may be required for accommodation of the polypeptide chain through the narrow proteasome channel or for editing functions of the proteasome. For example, inhibition of this isopeptidase activity with ubiquitin aldehyde increases the relative degradation of proteins with short ubiquitin chains. Thus, PA700 may be involved in kinetic partitioning of proteins between degradation and refolding.
Primary structures of most of the remaining subunits of PA700 in yeast and humans have been determined, but little is known about their relative roles in proteolysis (4). Many of the proteins had been identified initially without realization that they were subunits of PA700. The remarkably varied phenotypes of cells bearing mutations in these respective proteins indicate, not surprisingly, that proteasome-dependent proteolysis is involved in a large number of cellular processes. A major challenge of future work is to define the exact roles of each of the 20 PA700 gene products. The identification of subcomplexes of PA700 with partial functions of the whole complex promises rapid progress toward this goal (37).
Although PA700 utilizes polyubiquitin to select most proteins for proteasomal degradation, it also promotes ATP-dependent proteolysis of some non-ubiquitinated proteins. The best example of the latter category is ornithine decarboxylase (ODC). ODC degradation, however, depends upon antizyme, an inducible protein inhibitor that binds to ODC (42,43). Antizyme appears to target ODC to PA700 and, therefore, may behave as the functional equivalent of a polyubiquitin chain. This mechanism probably operates in vivo, but it is unclear how many other proteins are degraded by the proteasome-PA700 complex in a ubiquitin-independent fashion or to what extent proteins other than ubiquitin target substrates to PA700.

PA28, an ATP-and Ubiquitin-independent
Proteasome Activator Mammalian cells (but not yeast) contain a proteasome activator called PA28 (also known as 11 S regulator) (44,45). PA28 is composed of two 28,000-dalton subunits (␣ and ␤) that are 50% identical in primary structure (46). The subunits form a ringshaped molecule of about 180,000 daltons, whose exact quaternary structure is unclear but is likely to be either heterohexameric or heteroheptameric (47)(48)(49). PA28, like PA700, binds to the terminal rings of the proteasome and activates the hydrolysis of short peptides (44,45,47). However, in contrast to PA700, binding of PA28 to the proteasome does not require ATP or any other known cofactor. Furthermore, PA28 does not activate the degradation of large proteins, regardless of their state of ubiquitination. Recombinant PA28␣ subunit forms a homoheptamer whose crystal structure has been solved (50). However, neither PA28␣ nor PA28␤ appears to exist alone in intact cells. Nevertheless, recombinant PA28␣ is sufficient for proteasome activation but is much less efficient than native heteromeric PA28 (51,52). Recombinant PA28␤ is a monomer. Effects of isolated PA28␤ on proteasome activation have produced conflicting data, whose basis may relate to the very high concentrations of the subunit required to achieve proteasome activation (51,53). In any case, reconstitution of the ␣ and ␤ subunits restores full and efficient PA28 activity. This result has been explained alternatively by additive activities of each subunit or by modulatory effects of the ␤ subunit on the ␣ subunit (51,53). At least two domains of PA28 are important for its function. The identity of C-terminal residues (tyrosine in the wild-type proteins) is a critical determinant for binding to the proteasome (48,53). Binding of PA28 to the proteasome is necessary but not sufficient for activation; another domain near the center of the primary structure is required for activation after PA28 binding (53). A third member of the PA28 family, PA28␥, forms a homomultimer and also stimulates the proteasome's hydrolysis of short peptides (52). PA28, like PA700, probably activates the proteasome by opening the channel at the terminal rings, thereby increasing access of substrates of the catalytic sites and/or by allosteric activation of the catalytic sites (Fig. 3). The inability of PA28 to activate hydrolysis of large proteins probably reflects its inability to unfold and/or translocate these substrates in an ATP-dependent manner.
The physiological role of PA28 is unclear. The absence of PA28 in yeast indicates that it is not essential for general features of ubiquitindependent proteolysis but may serve a specialized role either supplemental to or independent of the ubiquitin-dependent pathway. Several lines of evidence suggest that one such function is to regulate the proteasome's production of antigenic peptides for presentation by class I molecules, a process also known to involve the ubiquitinproteasome pathway. First, PA28 (both ␣ and ␤ subunits) is upregulated by ␥-interferon, an important regulator of the immune response and the inducer of the three alternate catalytic proteasome subunits described above (46,54). Second, PA28 alters the cleavage specificity of peptides by the proteasome in a manner that may favor production of antigenic peptides (55). Finally, overexpression of PA28␣ by transfection resulted in more efficient antigen presentation in an indirect cytotoxicity assay (56). PA28 could influence antigen processing in a two-step mechanism whereby peptides produced initially via degradation of ubiquitinated proteins by proteasome-PA700 complexes are processed further by proteasome-PA28 complexes or via a "hybrid" proteasome complex containing both PA700 and PA28 (Fig. 3). Such recently identified hybrids could have distinct catalytic properties for the production of unique peptide products (57). Nevertheless, very little is known about the relative qualitative or quantitative activities of these various proteasome-activator complexes with respect to the production of specific peptides. Additional work will be required to determine the role and extent to which PA28 regulates this and possibly other aspects of proteasome activity in intact cells.

Protein Inhibitors of the Proteasome
Several protein inhibitors of the proteasome have been identified but in general they are much less well characterized than the activators (58 -61). Protein inhibitors were discovered by assaying fractionated tissue extracts for the ability to inhibit the 20 S proteasome, but in most cases they have not been tested for their ability to affect proteasome activity in the presence of activators. Thus, the physiological role of these proteins may be to inhibit 20 S proteasome that becomes activated in vivo or to regulate proteasome-activator complexes. For example, inhibitors could function by inhibiting such complexes directly or by blocking their formation. Additional work will be required to advance the understanding of the molecular mechanisms by which these proteins interact with the proteasome and the physiological significance of such interactions. Recently, potent and specific pharmacological proteasome inhibitors have been identified. Utilization of these inhibitors in cellular studies has identified many cellular roles of the proteasome and further highlights the physiological potential of regulated proteasome inhibition (13).

Physiological Regulation of the Proteasome System
Proteasome function in cells can be regulated by altering levels of the proteasome, proteasome regulatory proteins, or proteins of the ubiquitin conjugation system. Thus, cells may adjust their capacity for ubiquitin/proteasome-dependent proteolysis by changing levels of proteins that participate in substrate selection and/or in substrate degradation. Up-regulation of cellular proteasome levels suggests that proteolytic capacity of the proteasome per se is rate-limiting under certain conditions. This conclusion is surprising because cells normally contain a high proteasome concentration. Nevertheless, changes in expression of the proteasome and other regulatory proteins have been documented in a large number of physiological and pathological conditions, many of which are associated with atrophy of skeletal muscle as a consequence of increased rates of global protein degradation (62,63). Therefore, in addition to the selective degradation of individual proteins, the ubiquitin-proteasome pathway appears to be responsible for alterations in tissue size. The complexity of this system involving over 50 gene products has hindered a systematic examination of altered expression of all components under a given condition and the exact functional consequences of such changes. Thus, many of these initial studies are at a necessarily descriptive stage. Moreover, little is known about the signaling pathways responsible for altered expression and regulation of the pathway's proteins. Early results, however, suggest that many components of the system are likely to be regulated coordinately and highlight the need for continued work in this important area.