Relative Functions of the α and β Subunits of the Proteasome Activator, PA28*

PA28 is a 180,000-dalton protein that activates hydrolysis of small nonubiquitinated peptides by the 20 S proteasome. PA28 is composed of two homologous subunits, α and β, arranged in alternating positions in a ring-shaped oligomer with a likely stoichiometry of (αβ)3. Our previous work demonstrated that the carboxyl terminus of the α subunit was necessary for PA28 to bind to and activate the proteasome. The goals of this work were to define the exact structural basis for this effect and to determine the relative roles of the α and β subunits in proteasome activation. Each subunit and various mutants of the α subunit were expressed in Escherichia coli and purified. PA28α stimulated the proteasome, but had a much greaterK act than native heteromeric PA28. In contrast, PA28β was unable to stimulate the proteasome. Mutants of the α subunit in which the carboxyl-terminal tyrosine residue was deleted or substituted with charged amino acids could neither bind to nor activate the proteasome. However, substitution of the carboxyl-terminal tyrosine with other amino acids resulted in proteins which could stimulate the proteasome to various extents. Tryptophan mutants stimulated the proteasome as well as did native PA28, whereas serine or phenylalanine mutants stimulated the proteasome much poorer than did wild type PA28α. Deletion of the “KEKE” motif, a 28-amino acid domain near the amino terminus of PA28α, had no effect on proteasome stimulatory activity. Hetero-oligomeric PA28 proteins were reconstituted from isolated wild type and mutant subunits. PA28 reconstituted from wild type subunits had structural and functional properties that were indistinguishable from those of the native hetero-oligomeric protein. PA28 molecules reconstituted from inactive α subunits and wild type β subunits remained inactive. However, PA28 molecules reconstituted from suboptimally active α mutants and wild type β subunits had the same activity as native heteromeric PA28. These results indicate that the β subunit modulates PA28 activity, perhaps by influencing the affinity of PA28 for the proteasome.

The proteasome is a 700,000-dalton protease composed of 28 similarly sized subunits (1)(2)(3)(4)(5). These subunits, which in eukaryotes represent the products of 14 distinct genes, are arranged as a stack of four heptameric rings that form a cylin-drical particle (6 -8). Although all 14 gene products have similar amino acid sequences, they can be divided into two subgroups of seven, termed ␣ and ␤, whose respective members' sequences are more closely related to one another (9). Each of the two outer or terminal rings of the proteasome are composed of all seven ␣-type subunits, whereas each of the two inner rings are composed of all seven ␤-type subunits (7). Certain archaebacteria contain a simpler form of the proteasome, which consists of only two gene products, one representing the ancestor of the eukaryotic ␣-type subunits and the other representing the ancestor of the eukaryotic ␤-type subunits (9). Nevertheless, the archaebacterial proteasome has the same overall morphology as eukaryotic proteasomes, and the ␣ and ␤ subunits occupy the same relative positions within the particle as do those of eukaryotic proteasomes (10,11). Thus, regardless of source, the proteasome is a multisubunit dimer that exhibits a C2 symmetry about an axis through the two inner rings (7,11).
Recently, crystal structures for both archaebacterial and yeast proteasomes were solved (12,13). The structures have clarified several issues left unresolved by previous biochemical and molecular studies. For example, it is now clear that the proteasome's catalytic activities are located on ␤ subunits and that at least three of the seven eukaryotic ␤ subunits are catalysts. Furthermore, the catalytic nucleophile for each of these distinct subunits is an amino-terminal threonine oriented toward a chamber present in the interior of the cylindrical particle (12)(13)(14). Access to this chamber by protein substrates appears to be highly restricted. In archaebacteria, substrates must traverse a narrow 13-Å portal formed by the center of the terminal ring subunits (12). In yeast proteasomes, this portal is blocked by the amino-terminal portions of the ␣ subunits, and there is no obvious path by which substrates can reach the active sites (13). These findings imply that there must be mechanisms that regulate the influx of substrates through the terminal rings for delivery to the sequestered active sites.
In eukaryotes, mechanisms that regulate entry of substrates to active sites may be mediated by specific regulatory proteins that bind to the terminal rings of the proteasome. In fact, two proteins, PA700 and PA28, stimulate proteasome activity after binding with such topologies (1,3,(15)(16)(17). PA700 (also known as 19 S cap or ATPase regulator) is a 700,000-dalton multisubunit complex that mediates the proteasome's ability to degrade ubiquitinated proteins in an ATP-dependent fashion (18 -21). This complex contains at least one subunit that binds polyubuiqutinated protein substrates (22)(23)(24) and six homologous subunits that contain ATP binding domains (16,(25)(26)(27)(28). Assembly of the proteasome-PA700 complex as well as subsequent degradation of ubiquitinated proteins requires ATP hydrolysis (18 -20, 29). Although the exact role of ATP in either process is unknown, it is reasonable to speculate that ATP may be used to translocate the polypeptide chain from binding sites on PA700 to the catalytic sites within the interior of the protein (17).
PA28 is a 180,000-dalton activator of the 20 S proteasome. It is a ring-shaped molecule composed of two gene products, termed ␣ and ␤ (30 -32). These subunits have primary structures that are approximately 50% identical (33)(34)(35). Previously, work by us and others indicated that the ␣ and ␤ subunits occupy alternating positions in the ring, suggesting that PA28 has a hexameric structure of (␣␤) 3 (36,37). PA28 activates the proteasome by binding to one or both terminal rings, but unlike PA700 does not require ATP for either binding or activation (32). PA28 regulates the proteasome's hydrolysis of small nonubiquitinated peptides by increasing V max and decreasing apparent K m , and therefore functions as a positive allosteric effector (30,31). Previous work in our laboratory has demonstrated an important role for the carboxyl terminus of the ␣ subunit in proteasome activation. Specifically, treatment of PA28 with carboxypeptidases resulted in complete loss of activity due to the loss of only one or two amino acids from the carboxyl terminus of the ␣ subunit (36,38). Thus, a limited structural domain of one of the two subunits appeared to be required for binding of PA28 to the proteasome and resultant proteasome activation. Further support for a key role of the ␣ subunit in PA28 function was obtained in studies showing that the isolated ␣ subunit, prepared as a recombinant protein or electrophoretically separated from denatured native PA28, could activate the proteasome although not as efficiently as the native protein containing both ␣ and ␤ subunits (34,36,39,40). Thus, we hypothesized that the ␤ subunit might act to mediate function of the ␣ subunit, perhaps by increasing the affinity of PA28 for the proteasome.
The purpose of the present work was to further define the structural requirements of the carboxyl terminus of the ␣ subunit in proteasome activation and to examine the functional role of the ␤ subunit in PA28 function. Therefore, we performed mutational analysis of the ␣ subunit to study structure/function relationships for this protein. We have also established the role of the ␤ subunit by reconstituting heteromeric PA28 from it and wild type or mutant ␣ subunits.

MATERIALS AND METHODS
Plasmid Construction of the PA28␣ Mutants-Mutations of rat PA28␣ were generated by PCR 1 using the previously described PA28␣ plasmid as a template (35,36). Mutations were introduced in the 3Ј-PCR primer, changing the last codon for tyrosine to codons for tryptophan, phenylalanine, serine, lysine, glutamic acid, or to a stop codon. The 5Ј-PCR primer has a NcoI restriction site at its end and the 3Ј-PCR primer has a BamHI site. The PCR products as well as the pET16b vector (Novagen) were digested with both NcoI and BamHI (Life Technologies, Inc.). Digested PCR product and vector were ligated with T4 DNA ligase. The ligation mixture was transformed into Escherichia coli strain DH5␣. For construction of the mutant lacking the KEKE domain, PCR was used to generate two fragments of PA28␣ cDNA, one encoding amino acid residues 1-69 and a second encoding amino acid residues 96 -249. The former has a NcoI site at its 5Ј end and a SalI site at its 3Ј end. The latter has a SalI site at its 5Ј end and a BamHI site at its 3Ј end. The restriction enzyme-digested PCR products and the pET16b vector DNA digested with NcoI and BamHI were ligated with T4 DNA ligase. The ligation mixture was transformed into E. coli strain DH5␣. The PA28␣-KEKE/-Y249 mutant was generated using the same strategy as that for the other carboxyl-terminal mutants, but used the PA28␣-KEKE DNA as the template. The 5Ј-PCR primer has a NcoI restriction site at its end, and the 3Ј-primer has a BamHI site. The PCR products as well as the pET16b vector were digested with both NcoI and BamHI. The digested PCR products and vector were ligated with T4 DNA ligase. The identities of all mutant constructs were verified by sequencing using an ABI PRISM TM 377 automated DNA sequencer (Perkin-Elmer Corp).
Expression and Purification of PA28␣ Mutants from E. coli-PA28␣ mutants were expressed in E. coli strain BL21 (DE3) and purified as described previously for the wild type PA28␣ (36). Purified proteins were subjected to mass spectrometry to confirm structural alterations, as described previously (36).
Construction of PA28␤ Expression Plasmid-cDNA encoding rat PA28␤ was amplified by PCR to yield a fragment lacking the initial ATG (35). This fragment was cloned into pET16b digested with NcoI and NdeI and blunt ended with Klenow fragment; the resulting plasmid was designated pJVK39. The identity of the PA28␤-encoding sequence was verified by sequencing.
Expression of PA28␤-For expression of PA28␤, pJVK39 was transformed into E. coli BL21 (DE3). Expression was induced with isopropyl-1-thio-␤-D-galactopyranoside (1 mM final concentration). After 3.5 h of induction, the cells were harvested, washed with buffer H (20 mM Tris-HCl, pH 7.6, at 4°C, 20 mM NaCl, 1 mM EDTA, 5 mM ␤-mercaptoethanol), collected by centrifugation, and frozen at Ϫ70°C. The cells were thawed on ice and resuspended in 15 ml of buffer H containing 0.3 mg/ml lysozyme and incubated on ice for 20 min. After sonication, insoluble material was removed by centrifugation (40,000 ϫ g for 15 min). PA28␤ was present in both the soluble and insoluble fraction as determined by Western blot analysis with anti-PA28␤ antibodies. PA28␤ was purified from the soluble fraction by column chromatography. The insoluble pellet was resuspended in 8 M urea and dialyzed extensively against buffer H. The small amount of precipitate that formed during dialysis was removed by centrifugation. The supernatant was purified by ion exchange chromatography. The sample was applied to a 5 ϫ 2.5-cm column of DEAE-Fractogel equilibrated with buffer H containing 150 mM NaCl and eluted with a 100-ml linear gradient of 150 -400 mM NaCl in buffer H. The data presented in this study use PA28␤ purified from the pellet fraction, although similar results were obtained for the originally soluble protein.
Reconstitution of PA28 from Recombinant Subunits-Heterodimeric PA28 was reconstituted from purified recombinant proteins including native PA28␣ or various PA28␣ mutants, and PA28␤. Equal volumes of proteins at 0.2 mg/ml (i.e. final concentration of 0.1 mg/ml for each protein) were preincubated in buffer H containing 5% glycerol for 16 h at 4°C.
Immunoprecipitation of PA28 -Immunoprecipitation of PA28 reconstituted from recombinant subunits was conducted as described previously for native PA28 with minor modifications, using antisera specific for either PA28␣ or PA28␤ (36). Seven l of serum (except for the precipitation of protein containing the PA28␣-Y249 and PA28␣-KEKE/-Y249 constructs, where 14 l of serum were used) was preincubated with 300 l of a 1/6 suspension of protein A-Sepharose CL-4B in TTBS (TBS containing 0.05% Tween 20) for 1.5 h at 4°C. 15 l of reconstituted PA28, prepared as described above (0.2 mg/ml in buffer H containing 5% glycerol) and 250 l of bovine serum albumin (0.2 mg/ml in TTBS) were added to the washed beads. The suspension was shaken gently for 1.5 h at 4°C, after which the pellet was collected by centrifugation. The equivalent of 8 l of supernatant was subjected to SDS-PAGE. The pellet was washed three times with TTBS and heated to 100°C with 100 l of 2 ϫ SDS sample buffer lacking ␤-mercaptoethanol. Ten l of this sample were subjected to SDS-PAGE.
PA28 Activity Assays-PA28 was assayed for its ability to stimulate the hydrolysis of Suc-Leu-Leu-Val-Tyr-AMC by the purified 20 S proteasome, as described previously (30). One unit of PA28 activity is defined as the increase of 1 unit of proteasome activity, and is expressed as the difference between proteasome activity in the presence and absence of PA28. Additional details are found in appropriate figure legends.
Glycerol Density Gradient Centrifugation-Glycerol density gradient centrifugation was carried out in 1.98-ml gradients (10 -40% glycerol) containing 50 mM Tris-HCl buffer, pH 7.6, and 5 mM ␤-mercaptoethanol. Protein samples (2 g in 10 l) were centrifuged for 4.5 h at 55,000 rpm in a Beckman TL100 ultracentrifuge using a TLS55 rotor at 4°C. Tubes were fractionated into 22 90-l fractions, and the distribution of proteins was determined by Western blotting.

RESULTS
The Carboxyl-terminal Tyrosine of PA28␣ Is Required for Proteasome Activation-Our previous work demonstrated that the carboxyl terminus of the ␣ subunit of PA28 is essential for activation of peptide hydrolysis by the 20 S proteasome (36,38).
Treatment of either native PA28, which contains both ␣ and ␤ subunits, or recombinant PA28␣ with carboxypeptidase Y resulted in the loss of each protein's ability to activate the proteasome (36). To define the precise modification of PA28␣ responsible for loss of proteasome stimulating activity, a mutant of this subunit lacking only the carboxyl-terminal tyrosine residue was expressed in E. coli and purified to homogeneity (Fig. 1). This protein, designated ␣-Y249, had no detectable proteasome stimulatory activity (Fig. 2). Therefore, deletion of tyrosine from the carboxyl terminus of PA28␣ results in complete loss of function for this protein. Furthermore, the lack of stimulatory activity of this mutant results from its inability to bind to the proteasome (data not shown).
To further examine the structural requirements of the carboxyl terminus of PA28␣ for proteasome activation, additional mutants were constructed, expressed in E. coli, and purified as described under "Materials and Methods" (Fig. 1). These altered proteins included those in which the carboxyl-terminal tyrosine was changed to: tryptophan (Y249⌬W), phenylalanine (Y249⌬F), serine (Y249⌬S), lysine (Y249⌬K), or glutamic acid (Y249⌬E). As reported previously, PA28␣ stimulates the proteasome to the same extent as does native PA28 (containing both ␣ and ␤ subunits), but requires about 5-10 times more protein to achieve half-maximal activation (K act ) (36). The Y249⌬W mutant, however, had the same K act as native heteromeric PA28. The Y249⌬F and Y249⌬S mutants were much poorer proteasome stimulators than wild type PA28␣ and had 3 and 10 times higher K act values, respectively. In contrast, the Y249⌬K and Y249⌬E mutants failed to activate the proteasome. These results demonstrate that activation of the proteasome by PA28␣ is dependent on the identity of a single amino acid at its carboxyl terminus.
The KEKE Domain of PA28␣ Is Not Required for Proteasome Activation-PA28␣ contains a segment of 28 amino acids (residues 70 -97) with an unusually high content of lysine and glutamic acid residues (34,35). This feature, known as a KEKE motif, has been postulated to participate in protein-protein interactions (41). KEKE motifs are present in numerous proteins including some subunits of the proteasome. Interestingly, however, a KEKE motif is not present in PA28␤ (35). To investigate the role of the KEKE domain for PA28␣ function, a mutant in which this domain was deleted was expressed in E. coli, purified, and assayed for proteasome stimulatory activity. As shown in Figs. 3 and 4, this mutant (␣-KEKE) bound to the proteasome and stimulated proteasome activity as efficiently as did native PA28␣. Thus, the KEKE motif is not required for proteasome activation by PA28␣. A second mutant lacking both the KEKE motif and the carboxyl-terminal tyrosine was also generated and analyzed. As expected, this protein had no proteasome stimulatory activity.
PA28␤ Does Not Activate the Proteasome-Our previous studies of the relative roles of the ␣ and ␤ subunits of PA28 for proteasome activation were conducted prior to the successful expression of the ␤ subunit in E. coli. We now have expressed PA28␤ in E. coli and have purified the protein to homogeneity (Fig. 5). Purified PA28␤ does not stimulate the proteasome at concentrations up to 10 times greater than those which produced maximal stimulation by PA28␣. These and similar results reported during the preparation of this manuscript demonstrate that the ␤ subunit has no direct or independent role in proteasome stimulation by PA28 (39,40). Despite its inability to stimulate proteasome activity, the ␤ subunit could bind to the proteasome with sufficiently high affinity to be isolated in a complex after glycerol gradient centrifugation (data not shown).
Reconstitution of PA28 from ␣ and ␤ Subunits-The results described above demonstrate that the ␣ subunit is necessary and sufficient for proteasome activation. However, they also suggest that the ␤ subunit plays a modulatory role in PA28 function because the native heteromeric protein has a lower K act than the isolated ␣ subunit. To further investigate the relative structural and functional roles of the ␣ and ␤ subunits of PA28 in proteasome activation, we have developed a procedure, as described under "Materials and Methods," for the reconstitution of PA28 hetero-oligomers from isolated recombinant ␣ and ␤ proteins. Reconstitution was documented by immunoprecipitation of the resulting complexes with antibodies specific for either ␣ or ␤ subunits followed by Western blot analysis of the immunoprecipitates with each of the antibodies (36). The reconstitution of PA28 from wild type ␣ and ␤ subunits is shown in Figs. 6 and 7. Thus, after the reconstitution protocol, these subunits were coimmunoprecipitated with antibodies specific for either protein. These results demonstrate that the recombinant ␣ and ␤ proteins had assembled into a common complex. Additional support for this conclusion was obtained by comparing structural features of the reconstituted PA28 to those of each isolated recombinant protein and to  Fig. 1 and purified native bovine PA28 were tested for their ability to activate purified latent 20 S proteasome using the fluorogenic substrate Suc-Leu-Leu-Val-Tyr-AMC, as described under "Materials and Methods." Each assay contained 0.24 g of proteasome (1.9 units/assay) and the indicated amounts of PA28 proteins. native PA28. During glycerol density gradient centrifugation, native PA28 sedimented at a position corresponding to an approximate size of 200,000 daltons, a value in good agreement with its expected hexameric or heptameric structure (Fig. 7, panel E). Recombinant PA28␣ sedimented at the same position, indicating that this subunit formed a homomeric complex, probably a hexamer or heptamer (Fig. 7, panel C). In contrast, recombinant PA28␤ sedimented very slowly, indicating that it is a monomer in the absence of the ␣ subunit (Fig. 7, panel D).
After the reconstitution protocol, however, the ␣ and ␤ subunits cosedimented and had a distribution profile indistinguishable from that of native PA28 (Fig. 7, panels A and B). These data provide additional strong evidence that the recombinant ␣ and ␤ subunits had reconstituted into a heterohexameric or heteroheptameric complexes.
Reconstitution experiments were also performed using wild type ␤ subunit and each of the mutant ␣ subunits described above. As shown in Fig. 6, each ␣ subunit mutated at the carboxyl terminus formed a heteromeric complex with the ␤ subunit. The mutant lacking the KEKE motif also formed a heteromeric complex, but not as efficiently as did the other mutants, as judged by the incomplete coprecipitation of subunits by subunit-specific antibodies (Fig. 3). These results may indicate that ␣ and ␤ subunits interact via the KEKE domain.
The ␤ Subunit Modulates PA28 Activity-To assess directly the effect of the ␤ subunit on the function of heteromeric PA28, the various reconstituted PA28 molecules described above were analyzed for their ability to stimulate the proteasome. PA28 reconstituted from wild type ␣ and ␤ subunits stimulated proteasome activity in an indistinguishable manner from that of native PA28 (Fig. 8). This result indicates that the ␤ subunit functions to decrease the K act of the ␣ subunit, perhaps by increasing the affinity of PA28 for the proteasome. PA28 reconstituted from ␣ subunits with tryptophan as the carboxylterminal residue stimulated proteasome activity as well as did native PA28. Thus, the ␤ subunit had no effect on this already maximally efficient ␣ subunit. In contrast, PA28 proteins reconstituted from ␣ subunits containing either serine or phenylalanine as the respective carboxyl-terminal residues stimulated the proteasome as well as did native PA28. In these cases, the ␤ subunit restored the ability of ␣ subunit mutants with suboptimal activities to stimulate the proteasome optimally. On the other hand, no proteasome stimulatory activity was detected in heteromeric proteins reconstituted from ␣ subunits lacking the carboxyl-terminal tyrosine, or from those containing either lysine or glutamic acid as the respective carboxyl-terminal residues. Thus, the ␤ subunit could not restore the function of these completely inactive ␣ subunits.
Finally, PA28 reconstituted from ␣ subunits lacking the KEKE domain stimulated the proteasome to a lesser maximal degree than did native PA28. This result was surprising because the isolated ␣ mutant stimulated the proteasome as well as did the wild type ␣ subunit. The decreased level of proteasome activation may be related to the incomplete formation of heteromeric complexes achieved with this mutant subunit ( Fig.  3; see "Discussion"). DISCUSSION We have expressed in E. coli each of the component subunits of the proteasome activator, PA28, and have used native and mutant forms of these recombinant proteins to investigate their structure/function relationships. The results reported here confirm and greatly extend our previous work by demonstrating that the nature of a single amino acid residue at the carboxyl terminus of the ␣ subunit determines whether PA28 can bind to and activate the proteasome. The wild type ␣ subunit has a tyrosine at this position, and previous results showed that its removal by carboxypeptidases was associated FIG. 6. Reconstitution of heteromeric PA28 from isolated ␣ and ␤ subunits. Wild type PA28␣ and the various mutants described in the text were prepared and incubated with purified recombinant PA28␤ as described under "Materials and Methods." The resulting proteins were subjected to immunoprecipitation using antibodies specific for either subunit (36). The precipitated proteins (P) and the unprecipitated proteins (S) were subjected to SDS-PAGE and immunoblotted with each of the antibodies. IP indicates the specificity of the antibody used for the immunoprecipitation. Blot indicates the specificity of the antibody used for the Western blot analysis.
FIG. 7. Glycerol density gradient centrifugation of reconstituted PA28. PA28 reconstituted from wild type ␣ and ␤ subunits was subjected to glycerol density gradient centrifugation as described under "Materials and Methods." Panels A and B, fractions from the same gradient of the reconstituted protein immunoblotted with antibodies specific for either PA28␣ or PA28␤, respectively. Panel C, fractions from a gradient of PA28␣ blotted with an antibody specific for PA28␣. Panel D, fractions from a gradient of PA28␤ blotted with an antibody specific for PA28␤. Panel E, fractions from a gradient of native bovine PA28 blotted with an antibody made against this protein. Proteins of known molecular weight (thyroglobulin, M r ϭ 660,000 and catalase, M r ϭ 240,000) were centrifuged in separate tubes and their respective sedimentation positions are indicated by the arrowheads.
with inactivation of PA28. The latter experiments, however, could not establish whether this modification was sufficient for inactivation or whether the removal of the penultimate amino acid (isoleucine) was also necessary for loss of PA28 function. The current analysis of recombinant PA28␣ mutants lacking only tyrosine has resolved this issue, and demonstrates that the removal of remarkably little structural information results in complete elimination of PA28 function. This information is specific for the chemical properties of the amino acid at the carboxyl terminus because substitution of the tyrosine with charged amino acids such as glutamic acid or lysine resulted in inactive PA28, whereas substitution with tryptophan greatly improved the ability of the isolated ␣ subunit to activate the proteasome.
The current results directly test speculations about functional roles of KEKE motifs (41). These motifs are found in a number of interacting proteins, including PA28␣ and certain ␣-type proteasome subunits, thereby raising the possibility that binding might occur via direct interactions between the two KEKE motifs or via an interaction between a KEKE motif of one protein and a distinct structure to which it can bind on the second protein. The current data show that deletion of the KEKE motif from PA28␣ does not prevent its binding to or stimulation of the proteasome. Thus, direct KEKE-KEKE interactions between PA28 and the proteasome are not required for the binding of these proteins. The PA28␣ mutant lacking the KEKE domain also formed a multimeric complex, suggesting that the association of ␣ subunits does not depend on this motif. This mutant, however, was poorly reconstituted with the ␤ subunit and it is possible that the KEKE motif plays a role in association between ␣ and ␤ subunits. Additional work will be required to determine the structural and functional significance of KEKE motifs in PA28.
The current results also provide important new information about the role of the ␤ subunit of PA28 in the mechanism of proteasome activation. Although the isolated ␤ subunit cannot directly activate the proteasome, it modulates the stimulatory effect of the ␣ subunit in heteromeric PA28 complexes. Thus, both native PA28 and PA28 reconstituted from recombinant ␣ and ␤ subunits activate the proteasome with a much lower K act than does PA28␣ alone. This effect was even more pronounced with two PA28␣ mutants in which the carboxyl-terminal tyrosine was replaced by either phenylalanine or serine. Each of these mutants was a very poor proteasome activator as an isolated protein, but functioned indistinguishably from native PA28 when reconstituted into a heteromeric protein with the wild type ␤ subunit. The mechanism by which the ␤ subunit influences ␣ subunit function is unclear, but the finding that isolated ␤ subunits can bind to the proteasome indicates that it may contribute significant binding energy to the PA28-proteasome interaction. Additional work will be required to determine the exact roles of the ␣ and ␤ subunits in the interaction between PA28 and the proteasome.
In addition to their distinct functions, isolated ␣ and ␤ subunits have very different structures. Isolated ␣ subunits form multimeric complexes of the same size as that of the native heteromeric PA28 as judged by density gradient centrifugation or gel filtration chromatography, whereas isolated ␤ subunits exist principally as monomers. These results indicate that the affinity of ␣ subunits for one another is appreciably greater than the affinity of ␤ subunits for one another. This may explain why the reconstitution of hetero-oligomeric PA28 from isolated ␣ and ␤ subunits was achieved much more efficiently when the subunits were incubated at low protein concentrations (0.1 mg/ml) as opposed to high protein concentrations (1.0 mg/ml). Low protein concentrations should shift the equilib-rium of ␣ subunit interaction toward monomers, and in the presence of ␤ subunits, the high affinity of ␣ and ␤ subunits for one another would favor formation of heteromeric structures. In contrast, high protein concentrations should shift the equilibrium of ␣ subunit interaction toward multimeric ␣ structures, thereby impeding the formation of heteromeric PA28. Interestingly, reconstitution of heteromeric PA28 could be achieved at high protein concentrations in the presence of urea, which may dissociate the ␣ subunits and promote heterologous subunit interactions. 2 The differences in ␣ and ␤ subunit function and interaction exist even though these proteins are about 50% identical in primary structure. These observations indicate that the dissimilar regions of these proteins may contain structures that define their unique functions. Additional studies are underway to define which regions of the two proteins are responsible for various properties such as subunit interactions and proteasome activation. In this regard, the ␣ subunit mutant lacking the KEKE motif did not efficiently form heteromeric complexes with ␤ subunits. This finding may indicate that the KEKE structure is involved in the interaction with the ␤ subunit. Curiously, protein samples from this reconstitution experiment stimulated the proteasome to a lesser degree than did the isolated mutant protein. This result could be explained if non-reconstituted ␤ subunits (which do not stimulate the proteasome) competed with non-reconstituted ␣ subunits and the heteromeric protein (both of which do stimulate the proteasome) for proteasome binding. This explanation, however, predicts that isolated ␤ subunits should competitively inhibit proteasome stimulation by the ␣ subunit; we have been unable, however, to detect significant inhibition in this type of experiment. 3