Characterization of Recombinant REGα, REGβ, and REGγ Proteasome Activators*

Full-length cDNAs for three human proteasome activator subunits, called REGα, REGβ, and REGγ, have been expressed in Escherichia coli, and the purified recombinant proteins have been characterized. Recombinant α or γ subunits form heptameric species; recombinant β subunits are found largely as monomers or small multimers. Each recombinant REG stimulates cleavage of fluorogenic peptides by human red cell proteasomes. The pattern of activated peptide hydrolysis is virtually identical for REGα and REGβ. These two subunits, alone or in combination, stimulate cleavage after basic, acidic, and most hydrophobic residues in many peptides. Recombinant α and β subunits bind each other with high affinity, and the REGα/β heteromeric complex activates hydrolysis of LLVY-methylcoumaryl-7-amide (LLVY-MCA) and LLE-β-nitroanilide (LLE-βNA) more than REGα or REGβ alone. Using filter binding and gel filtration assays, recombinant REGγ subunits were shown to bind themselves but not α or β subunits. REGγ differs from REGα and REGβ in that it markedly stimulates hydrolysis of peptides with basic residues in the P1 position but only modestly activates cleavage of LLVY-MCA or LLE-βNA by the proteasome. REGγ binds the proteasome with higher affinity than REGα or REGβ yet with lower affinity than complexes containing both REGα and REGβ. In summary, each of the three REG homologs is a proteasome activator with unique biochemical properties.

The proteasome is a large multisubunit protease found in eukaryotes, prokaryotes, and archaebacteria (1,2). The enzyme from the archaebacterium Thermoplasma acidophilum has served as an important structural model for proteasomes from higher organisms. It is composed of two kinds of subunits called ␣ and ␤ (3). These two proteins associate to form four rings stacked upon one another to produce a cylinder roughly 10 nm in diameter and 15 nm in length (4 -6). Seven ␣ subunits form each of the end rings with two rings of seven ␤ subunits sandwiched between them (7,8). The recently determined x-ray structures of the archaebacterial and yeast enzymes (9,10) have shown that the two internal ␤-rings generate a buried chamber containing 14 active sites approximately 30 Å apart from each other.
In higher eukaryotes, there are 7 genes encoding ␣ subunits and 10 genes encoding ␤ subunits (11). The ␤ subunits confer upon the proteasome at least five distinct activities that cleave peptide bonds following basic, acidic, hydrophobic, branched chain, and small neutral amino acids (12,13). In higher eukaryotes, the two ␣-rings interact with regulatory complexes to produce either a large 26 S energy-dependent protease or an activated peptidase. A 700-kDa regulatory complex composed of approximately 15 subunits binds eukaryotic proteasomes to form the 26 S protease that degrades ubiquitylated or unmodified proteins in an ATP-dependent fashion (14,15). In the absence of ATP, an 11 S protein complex can bind to either one or both ends of the proteasome and stimulate its peptidase activities up to 60-fold. This oligomeric complex, known as the PA28 or 11 S REG (16,17), is composed of two highly conserved and homologous subunits that we called REG␣ and REG␤ (18,19).
The proteasome, either as part of the 26 S protease or as an activated peptidase, has been implicated in a variety of cellular processes. These include cell cycle progression (20,21), selective proteolysis (15,22,23), and antigen presentation on class I major histocompatibility (MHC-I) 1 molecules (for reviews, see Refs. 2, 24, and 25). With regard to antigen presentation, the proteasome has been shown to cleave peptide precursors in vitro, generating products with structural properties similar to MHC-I epitopes (26 -28). This proteolytic capacity is enhanced by REG␣ (18); LMP2 and LMP7 (29,30), purified human REG (16); or a combination of LMPs and REG (31,32). These in vitro findings suggest that the proteasome and its regulators play an important role in the processing of antigenic peptides. In support of this idea, in vivo evidence indicates that LMP2 and LMP7, two proteasome ␤ subunits, are up-regulated by ␥interferon (33). A third proteasome ␤ subunit, MECL1 (34), and the ␣, ␤, and ␥ subunits of the 11 S REG (18,19) as well as other components of the antigen presentation pathway such as TAP1, TAP2, and MHC-I molecules are also induced by IFN␥ (35). Furthermore, mice lacking genes for LMP2 or LMP7 exhibit reduced capacity to present antigens on MHC-I molecules (36,37). And recently, it has been reported that overexpression of REG␣ subunits in mouse cells leads to increased antigen presentation (38). Taken together, these findings constitute strong circumstantial evidence that the 11 S REG and the proteasome are involved in production of peptides presented on MHC-I molecules.
We proposed earlier that the K i antigen, a protein homologous to the REG␣ and REG␤ subunits, would also be a proteasome activator (39). In this study, we have characterized the activation properties of recombinant REG␣, REG␤, and K i antigen, which we call REG␥. We have found that although REG␣ and REG␤ subunits can independently activate the peptidase activity of the proteasome, together these two proteins form a superactivating ␣␤ complex. The ␥ subunit did not interact with either ␣ or ␤ subunits. REG␥ is, nonetheless, able to bind the proteasome and activate hydrolysis of certain fluorogenic peptides.  Isolation and Characterization of cDNA Clones-The cloning and in vivo expression of the ␣ subunit of human regulator was described earlier (18). During the screening for REG␣, a 235-bp partial cDNA for REG␤ was isolated. Using this partial clone, [␣-32 P]CTP-labeled random-primed probes were generated (Random Primed DNA Labeling Kit, Boehringer Mannheim) and used to screen approximately 10 5 bacteriophages from a ZAP II HeLa cell cDNA library as described for the cloning of REG␣ (18). A full-length clone for REG␤ activator was isolated, sequenced, and subcloned into the pAED4 expression vector. The published sequence of KI-antigen (40) was used to design a nondegenerate oligonucleotide primer specific for the 3Ј-untranslated region of the REG␥ gene (5Ј-AGACCGACATTGCC-3Ј). The primer served to generate a gene-specific cDNA upon reverse transcription of CsCl 2 -purified total HeLa cell RNA with reverse transcriptase avian myeloblastosis virus. The resulting cDNA and two primers specific for the N and C terminus of the REG␥ gene (5Ј-ATGGCCTCGTTGC-3Ј and 5Ј-AG-GAGTCATGTCTCAGAG-3Ј, respectively), were combined in PCRs. The EditSeq, MegAlign, and Protean algorithms (DNASTAR Inc., Madison, WI) and data base were used to analyze the nucleotide and deduced amino acid sequences of cloned PCR products. Mutation-free clones were subcloned into expression vectors as described below.
Expression and Purification of Recombinant REG␣, REG␤, and REG␥ Proteins-Recombinant REG␣ was prepared as described previously (18). Full-length cDNAs for REG␤ and REG␥ were ligated into the NdeI/ClaI or NdeI/BamHI sites, respectively, of the T7 polymerase-dependent expression vector pAED4. Recombinant E. coli BL21(DE3), treated with IPTG according to the manufacturer's instructions, were lysed in 10 mM Tris, pH 8.8, 25 mM KCl, 10 mM NaCl, 1.1 mM MgCl 2 , 0.1 mM EDTA, and 1.0 mM dithiothreitol. The cell lysate was centrifuged at 39,000 ϫ g for 30 min at 4°C. The soluble recombinant proteins were purified by ion exchange chromatography on a DEAE Sephadex A50 column (Pharmacia Biotech Inc.) using a 0 -1 M KCl gradient in TSD, pH 8.8, followed by sizing chromatography on a Superdex 200 (26/60) column equilibrated in 10 mM Tris, pH 8.8, 200 mM KCl, 10 mM NaCl, 1.1 mM MgCl 2 , 0.1 mM EDTA, and 1.0 mM dithiothreitol. The purity of the recombinant proteins was determined by Coomassie Blue and silver staining of the proteins separated on a 15% SDS-polyacrylamide gel. The identity and integrity of the recombinant proteins were assessed by Edman degradation and electrospray mass spectrometry. The purified recombinant proteins were tested for the presence of contaminating proteases and peptidases using ubiquitylated or unmodified 125 I-lysozyme and a panel of 12 fluorogenic peptides. In particular, the preparations were tested for E. coli ATP-dependent HslVU protease activity using GGL-MCA, AAF-MCA, and LLVY-MCA as substrates (41).
By a combination of PCR and cloning approaches, we have isolated cDNAs for a number of REG variants. Four of these REG mutants were expressed in E. coli and purified. The four mutant REGs are REG␣T, which contains a 17-amino acid extension at the C terminus and forms an inactive heptamer; REGm␣, which is an inactive monomer; REGm␤, an inactive monomer; and REGm␥, an inactive heptamer. These four proteins served to control for the presence of E. coli factors that might copurify with the recombinant REGs and fortuitously stimulate the proteasome or cleave fluorogenic peptides directly. In all cases, the purified mutant REGs were completely inactive as proteasome activators, and the final purified fractions did not cleave any fluorogenic peptides. We therefore conclude that none of the activities reported in this paper can be attributed to E. coli proteins.
Metabolic Labeling of REG-␣, -␤, and -␥ and the 26 S Proteasome Subunit S5b-Recombinant BL21(DE3) E. coli carrying the gene for each recombinant protein were exposed to IPTG and L-[ 35 S]methionine (10 Ci/ml) in the presence of an amino acid mixture lacking methionine as described (42). BL21(DE3) host cells containing the empty vector pAED4 plasmid were used to prepare radiolabeled cell lysate. Radiolabeled proteins were analyzed by electrophoresis on 15% SDSpolyacrylamide gels, and the gel was exposed to X-OMAT AR film (Eastman Kodak Co.) or to a PhosphorImager screen. Radiolabeled proteins were diluted in PBS (2.6 mM KCl, 136.8 mM NaCl, 8.1 mM NaH 2 PO 4 , 1.47 mM KH 2 PO 4 ) containing 5% nonfat milk and were used in a protein blot assay.
Production of Homolog-specific Antibodies and Dot Blot Analysis of Recombinant REG␣ and REG␥-Recombinant proteins purified from BL21(DE3) E. coli were applied to a nitrocellulose filter using a dot blot apparatus. The filter was stained with Ponceau S, blocked for 60 min in 5% nonfat milk in TBS-T (25 mM Tris, pH 7.5, 0.9% NaCl, 0.04% Tween 20), and then incubated with polyclonal rabbit serum (1:2000 dilution) directed against either REG␣ or REG␤, washed in TBS, and then incubated for 4 h in the presence of peroxidase-conjugated goat antirabbit IgG. The peroxidase reaction was performed at room temperature following the manufacturer's instructions for the Renaissance Western blot chemiluminescence reagent and exposed to X-OMAT AR films. Polyclonal antibodies were raised in New Zealand White rabbits injected subcutaneously with purified recombinant REG␣ or recombinant REG␥. Polyclonal antiserum against REG␤ was obtained by injecting a ubiquitin fusion protein containing a C-terminal extension specific for the REG␤ insert (Ub-DPPPKDDEMETDKQEKKEC). The fusion protein was produced as described earlier (43) .
Fluorometric Proteasome Assays-Spectrofluorometric assays were performed in the presence of 100 M fluorogenic peptides and various amounts of proteasome and REG␣, REG␤, or REG␥ in a final volume of 100 l of 10 mM Tris, pH 7.45. Proteasome and recombinant REG proteins were incubated for 10 min at 25°C prior to the addition of fluorogenic peptide substrates. Reactions were performed at 25°C or at temperatures indicated in the figure legends and were terminated by adding 200 l of cold 100% ethanol. Fluorescence was measured with a Perkin-Elmer fluorometer using excitation at 380 nm and emission at 440 nm for peptides containing MCA; hydrolysis of substrates containing ␤NA was monitored at excitation and emission wavelengths of 335 and 410 nm, respectively. Proteasomes were purified from human red blood cells as described previously (44).
125 I-Lysozyme-Ubiquitin Conjugate Proteolysis-Activated Xenopus egg extracts were prepared as described, and 125 I-lysozyme-Ub conjugates were prepared as described (Refs. 45 and 46, respectively). Xenopus egg extracts (22.5 g) were mixed with 25 l of 125 I-lysozyme-Ub conjugates (approximately 420 cpm/l) or unmodified 125 I-lysozyme (approximately 10 6 cpm/l) and various amounts of REG␣, REG␤, and REG␥. Alternatively, purified human proteasomes, or proteasome-REG complexes were tested for the presence of ATP-dependent or ATPindependent proteolytic activity using conjugated or unmodified 125 Ilysozyme. The incubations were at 23°C, and the fraction of hydrolysis of 125 I-lysozyme was determined by acid precipitation (45).
Sizing Chromatography-Individual REG␣, REG␤, or REG␥ or mixtures of REG homologs were separated on a Superdex 200 (26/60) column at 1 ml/min of TSD (10 mM Tris, pH 7.0, 25 mM KCl, 10 mM NaCl, 1.1 mM MgCl 2 , 0.1 mM EDTA, and 1.0 mM dithiothreitol). The elution of REG␣, REG␤, REG␥, REG␣/␤, and REGm␣ was compared with the elution positions of calibration proteins II (Combitek, Boehringer Mannheim). Selected fractions (20 l) were analyzed for proteasome-stimulating activity using LLVY-MCA or LRR-MCA as substrates. Proteins in active fractions were separated by RP-HPLC as indicated in the figure legends. Alternatively, selected fractions were transferred onto a nitrocellulose filter after separation on a denaturing SDS-polyacrylamide gel or directly applied to the filter using a dot blot apparatus. The fractions were then analyzed for the presence of ␣, ␤, or ␥ subunits using homolog-specific antibodies.
Assay for REG-Proteasome Association-The direct binding assay relies on a monoclonal antibody, MCP20 (a kind gift from K. Hendil (University of Copenhagen)), which binds the proteasome but does not prevent binding of REG homologs. Microtiter wells in an enzyme-linked immunosorbent assay plate were coated with 200 l of goat anti-mouse IgG at 20 g/ml in 0.05 M carbonate, pH 9.6. The wells were then rinsed three times in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) and blocked with 200 l of 1.5% nonfat milk in TBS-T for 2 h. The wells were then filled with 200 l of a 1:2500 dilution of ascites fluid containing MCP20 and incubated overnight at 4°C. After three washes with TBS-T, each well was filled with 200 l of human red cell proteasomes at 30 g/ml in TBS-T and incubated overnight. The wells were washed three times with TBS-T and incubated with REG homologs in 10 mM Tris, pH 7.5, for 20 min at 37°C followed by an additional 150 min at 4°C. Each well was then quickly rinsed twice with an excess of cold 10 mM Tris, pH 7.5, 0.1% Tween 20 and once with 10 mM Tris, pH 7.5. The bound REG proteins were eluted with 200 l of 0.5 M NaCl in 20 mM Tris, pH 7.5, and the high salt eluate was dot-blotted onto nitrocellulose for subsequent detection with REG homolog-specific antibodies.

Cloning, Expression, and Purification of Recombinant
Human REG␣, REG␤, and REG␥-cDNAs encoding the ␣ and ␤ subunits of the human 11 S REG were isolated from a ZAPII cDNA library containing HeLa sequences. A cDNA for REG␥ was amplified from purified HeLa cell RNA using PCR primers designed on the basis of the published sequence (40). The deduced amino acid sequences for ␣ and ␥ subunits match those published earlier (18,40). The nucleotide and deduced amino acid sequences of the HeLa REG␤ subunit agree with the published sequence except for Ser 2 and Asn 229 , which were identified by Ahn et al. (19) as Ala 2 and Thr 229 , respectively. Interestingly, position 229 is also Asn in the REG␤ subunit from rats (19). Fig. 1A presents the sequence of each subunit aligned to emphasize the homology between the proteins except for highlighted regions that we call subunit-specific "inserts." The degree of sequence similarity between the three REG homologs is 35-50% using the J. Hein or Clustal methods (Fig. 1B).
Expression of each REG subunit in E. coli BL21(DE3) was initiated by the addition of IPTG to growing cells, and the recombinant proteins were purified by ion exchange chromatography, followed by gel filtration (see "Experimental Procedures"). Each subunit constituted more than 90% of the protein pooled after gel filtration (see Fig. 1C). The molecular masses determined by electrospray mass spectrometry (Fig. 1B) (18), REG␤, and REG␥ (40). Amino acid sequences were deduced from the nucleotide sequences using the Edit-Seq algorithm and aligned using the MegAlign algorithm (Clustal method). The insert of each REG is shown black on white. A potential protein kinase C recognition site shared by REG␣ and REG␤ is boxed. Differences from the published sequence of the REG␤ are circled; Ser 2 and Asn 229 were previously identified as Ala 2 and Thr 229 , respectively (19). B, physicochemical properties of REG␣, REG␤, and REG␥. cMr, calculated molecular masses (kDa) for the recombinant REGs lacking Met1; eMr, experimental masses for the recombinant REGs determined by electrospray mass spectrometry; pI, calculated isoelectric points; nMr, the estimated native molecular masses (kDa) were determined by sizing chromatography (SC) as described under "Experimental Procedures" and dynamic light scattering (DLS) using a DynaPro-801 TC instrument (ProteinSolutions, Charlottesville, VA). The theoretical data and sequence similarity values were obtained using the Protean algorithm. The experimental values were obtained as described under "Experimental Procedures." C, Coomassie Blue R250 staining of recombinant human REGs separated by SDS-polyacrylamide gel electrophoresis. S, molecular mass standards (kDa). good agreement with masses expected for each protein lacking its N-terminal methionine. Size exclusion chromatography demonstrated a marked difference in the apparent size of REG␤ as compared with REG␣ and REG␥. The latter two proteins chromatographed with apparent molecular weights of about 200,000. By contrast, REG␤ eluted from Superdex 200 at a size expected for a monomer or dimer (see Fig. 6). Extensive structural analysis, including equilibrium sedimentation and x-ray crystallography, demonstrate that recombinant REG␣ is a heptamer. 2 Proteasome Activation by REG␣, REG␤, and REG␥-Each recombinant REG subunit was tested for its ability to stimulate peptide hydrolysis by proteasomes. Human red blood cell proteasomes were incubated with increasing amounts of each recombinant subunit, and hydrolysis of LLVY-MCA was measured. It is clear from the activation profiles shown in Fig. 2A that the three subunits differ in their ability to stimulate cleavage of LLVY-MCA. Although high concentrations of REG␣ and REG␤ produce equivalent extents of activation, the two homologs differ significantly at lower concentration. Activation by REG␤ is biphasic, with modest stimulation of LLVY-MCA cleavage at concentrations below 30 ng/l and stimulation equivalent to REG␣ at higher concentrations (Fig. 2B). The complex activation curve for REG␤ was observed in three separate titrations. Mixtures of REG␣ and REG␤ produce 2-fold greater activation than either protein alone. Stimulation of LLVY-MCA cleavage by REG␥ was only 2-4-fold, far less than the 30 -50-fold stimulation seen with REG␣ or REG␤. However, marked stimulation of proteasome activity by REG␥ was observed using LRR-MCA as substrate (Fig. 2C).
Because the individual REG subunits diverge significantly at the insert region (see Fig. 1), we suspected that they might differentially activate cleavage of specific fluorogenic peptides. For this reason, we assayed the ability of each homolog to stimulate hydrolysis of 10 fluorogenic peptides. The results presented in Table I show that REG␣ and REG␤ are very similar in their ability to activate hydrolysis of specific peptides by the proteasome. They markedly activate hydrolysis of fluorogenic peptides containing basic, acidic, or hydrophobic amino acids in the P1 position. By contrast, REG␥ stimulates cleavage of fluorogenic peptides with positive residues adjacent to the fluorescent leaving group. It is particularly striking that REG␥ barely activates cleavage of LLVY-MCA or LLE-␤NA, two substrates that are highly responsive to REG␣ and REG␤.
Effects of REG Homologs on K m and V max -Further comparison among the three activators is provided by kinetic studies performed with three peptides frequently used as model substrates: LRR-MCA, LLE-␤NA, and LLVY-MCA. Increasing concentrations of each substrate were incubated with proteasomes or proteasome-activator complexes. The graphs in Fig. 3 show that REG␣␤ complexes exhibit higher activity than equimolar amounts of either REG␣ or REG␤. Again, it was found that REG␥ preferentially stimulates the cleavage of  Fig. 2A were replotted to emphasize the difference between REG␤ and the other homologs in proteasome stimulation at low concentration. C, increasing amounts of recombinant REG␥ were incubated for 10 min at 24°C with 340 ng of proteasome and 100 M LRR-MCA. Peptide hydrolysis was measured as described under "Experimental Procedures." and REG␥ Enzymatic reactions consisted of 100 l of 10 mM Tris, pH 7.45, containing 340 ng of human red blood cell proteasomes and 6 g of REG␣, REG␤, and REG␥. Proteasomes and REGs were incubated for 10 min at 25°C prior to addition of fluorogenic peptides to a final concentration of 100 M. The reactions were stopped after 15-40 min by addition of 200 l of ethanol. The markedly reduced activation of LLVY-MCA and LLE-␤NA hydrolysis by REG␥ cannot be attributed to cleavage of alternate bonds within these two fluorogenic peptides, since addition of aminopeptidase did not increase fluorescence. In fact, aminopeptidase addition did not produce increased fluorescence for any combination of substrates and enzymatic components presented in the table. This indicates that the amino acid-MCA and amino acid ␤NA bonds are the exclusive sites of hydrolysis. The values (fold stimulation, S) were calculated as follows: S ϭ F prot ϩ REG/F prot , where F prot and F protϩREG are the fluorescence values resulting from the cleavage of fluorogenic substrates with proteasomes and proteasome-REG complexes, respectively. Each value represents an average of four to six measurements.  (18,47,48), and it has been shown that levels of mRNAs for REG␣, -␤, and -␥ increase upon treatment with IFN␥ (19). These observations suggest that REG proteins may be up-regulated during viral infections. Because viral infection is often accompanied by fever, we examined the ability of each recombinant REG to stimulate peptide hydrolysis at various temperatures near mammalian body temperature. This revealed a significant difference between REG␣ or REG␥ on one hand and REG␤ on the other. The data in Fig. 4 show that stimulation of the proteasome by REG␣ and -␥ is unabated at temperatures up to 45°C. By contrast, the ability of REG␤ to enhance peptide hydrolysis markedly decreases above 40°C. Although these findings may or may not be physiologically relevant, they do reveal another biochemical difference between REG␤ and the other two REG homologs.
Interactions among REG␣, REG␤, and REG␥-The data in Figs. 2 and 3 demonstrate that each REG subunit can, by itself, activate the proteasome. Also, REG␣ forms a multimer known from x-ray crystallography and other physical analyses to be a heptameric ring. 2 For the sake of discussion, we will call REG␥ a heptamer, although at present we cannot rule out hexameric or octameric arrangements for this subunit. Recombinant REG␤ does not form a hexamer or heptamer stable to gel filtration (see Fig. 6, B and F). Nonetheless, in REG molecules isolated directly from human red blood cells, the ␤ subunit sediments at 11 S (17). Because the active REG fractions from red cells contain both ␣ and ␤ subunits, it is reasonable to assume that ␣ and ␤ subunits form stable heteromeric rings in vivo. Indeed, this assumption has received support from several recent studies (48 -50).
We performed two sets of experiments that demonstrate strong binding between ␣ and ␤ subunits. In the first experiment, a filter assay was used to examine interactions between the REG homologs. Each recombinant REG subunit was dotblotted onto nitrocellulose, and the filter was incubated in a solution containing metabolically labeled REG subunits. The autoradiograms presented in Fig. 5 reveal highly specific subunit interactions. 35 S-REG␥ bound only to REG␥; 35 S-REG␣ bound strongly to REG␤ and weakly to itself; 35 S-REG␤ bound only to REG␣. The ␣-␣ and ␥-␥ associations evident in the filter assay are consistent with the gel filtration properties of recombinant REG␣ and REG␥ subunits (Fig. 6, panels A and E, respectively); i.e. these two subunits form distinct homomeric species, known for REG␣ to be a heptamer. Similarly, the absence of strong ␤-␤ interactions can explain why REG␤ oligomers are not present upon gel filtration of this recombinant subunit (Fig. 6, B and F). The strong ␣-␤ interaction seen in the filter binding assay supports the idea that individual REG oligomers from human red blood cells contain both ␣ and ␤ subunits.
The chromatographic behavior of mixtures of recombinant REG subunits provided further evidence for strong associations between ␣ and ␤ subunits. Purified REG␤ subunits were mixed with equal amounts of either REG␣ or REG␥ and chromatographed on Superdex 200 fractogel. Comparison of the elution profiles in Fig. 6B with that in Fig. 6C show that ␤ subunits elute as heptamers or hexamers after being mixed with REG␣. Western blot analysis, RP-HPLC separation, and Edman degradation revealed that the hexa-/heptameric species eluting at 160 ml in Fig. 6C contained equal amounts of both ␣ and ␤ subunits. In agreement with the data shown in Figs. 2 and 3, the newly formed ␣␤ heteromer exhibited approximately 2-fold higher proteasome-stimulating activity than the same amount of the REG␣ homomer (Fig. 6D). By itself REG␤ did not stimulate the proteasome, since the concentrations present in the tested fractions were below the threshold required for activation (see Fig. 2B). In contrast, REG␥ oligomers did not bind REG␤ (Fig. 6, E-G). Electrophoretic analysis of selected fractions from the profiles shown in Fig. 6, E-G, failed to reveal major changes in the elution of either REG␤ or REG␥ (not shown). Furthermore, the REG␥ activity profile was virtually unchanged after mixing with REG␤ (Fig. 6H).
The possibilities that individual ␤ subunits bind to the surface of ␣ heptamers and are simply carried along during gel filtration or that REG␣ heptamers promote assembly of REG␤ homo-oligomers were excluded using a mutant REG␣ subunit. A number of inactive REG␣ proteins have been generated by PCR mutagenesis. Many do not form heptamers by themselves, but they readily become active multimers when mixed with REG␤ subunits. 3 An example of the conversion of an inactive monomeric REG␣ mutant (REGm␣) and wild type REG␤ monomers to a higher molecular weight, active multimer is shown in the gel filtration profiles in Fig. 7, A-C. Both REGm␣ and REG␤ migrate as monomers (Fig. 7, A and B, respectively), but the REGm␣/REG␤ mixture eluted as an oligomer (Fig. 7C). Neither monomeric species by itself activated the proteasome, but the mixed heteromer was active (Fig. 7D). Selected fractions from the profiles shown in Fig. 7, A-C, were analyzed by denaturing SDS electrophoresis (Fig. 7, E-G). The stained gels show that, when fractionated individually, REGm␣ and REG␤ chromatograph as monomers (solid arrowheads in panels E and F, respectively), but when mixed they elute at a position corresponding to heptamers or hexamers. Although REGm␣ and REG␤ subunits did not resolve in the gel shown in Fig. 7G, both subunits were present at an apparent 1:1 molar ratio upon RP-HPLC (Fig. 7H). We consider the formation of active oligomers from two inactive monomeric proteins, REG␤ and REGm␣, as very strong evidence that the two proteins form a heteromeric complex.
Association of REG␣, -␤, and -␥ with the Proteasome-The

FIG. 5. Filter binding assay for REG homolog interactions.
Purified REG-␣, -␤, and -␥ (2 g) were applied to a nitrocellulose filter (0.1 M). The filter was blocked in a solution containing 5% nonfat milk in PBS, rinsed briefly with PBS, and then incubated at 4°C for 3 h in PBS containing 5% dry nonfat milk and [ 35 S]methionine-labeled REG-␣, -␤, or -␥, recombinant subunit 5b (rS5b) of the 26 S proteasome (63), or radiolabeled bacterial lysate (Cell Lys) (see "Experimental Procedures"). After incubation, the filters were briefly rinsed in PBS, dried, and exposed to an x-ray film. Nitrocellulose spots containing bound radioactive material were boiled in 62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5% ␤-mercaptoethanol, and the eluted radioactivity was separated on 15% SDS-polyacrylamide gels and analyzed using a PhosphorImager. In all cases, the bound 35 S-proteins migrated at approximately 30 kDa, indicating that they were REG homologs (not shown). fact that REG␣, -␤, or -␥ stimulates peptide hydrolysis demonstrates that each is capable of binding the proteasome. Two experiments indicate that the individual REG subunits differ in their affinity for the enzyme. The first experiment exploits the different activation properties of REG␣, REG␤, and REG␥. As shown in Fig. 2, REG␥ stimulates the proteasome to hydrolyze LRR-MCA to a greater extent than LLVY-MCA. REG␣ or REG␤, on the other hand, stimulates cleavage of the two peptides almost equally. These properties allowed us to perform competition experiments. Proteasomes were mixed with saturating amounts of REG␣ or REG␤, and increasing amounts of REG␥ were added. As shown in Fig. 8, high levels of REG␥ suppress hydrolysis of LLVY-MCA activated by REG␣ (closed squares) or REG␤ (open diamonds). We interpret the observed inhibition as a result of REG␣ or REG␤ being displaced from the proteasome by the added REG␥ oligomers. Inhibition was not observed when proteasomes activated by a mixture of REG␣ and REG␤ were challenged with increasing amounts of REG␥ (Fig. 8, closed circles). Thus, the competition experi-ments indicate that REG species bind the proteasome with decreasing affinities in the order ␣/␤ Ͼ ␥ Ͼ ␣ or ␤.
Further support for differential proteasome binding by REG homologs was obtained using a direct binding assay. In this assay, REG homologs competed for binding to antibody-tethered proteasomes, and the identities of the proteasome-bound REGs were determined using homolog-specific antibodies. Results from the competition experiments are presented in Fig. 9, where the competing species are indicated for each dot blot. It is evident that REG␥ displaced REG␣ from the proteasome but did not compete away REG␣/␤ complexes (Fig. 9, upper panel), and as expected from the enzymatic assay in Fig. 8, only REG␣/␤ heteromeric complexes markedly reduced REG␥ binding to the proteasome (Fig. 9, lower panel). Thus, both enzymatic and direct binding assays demonstrate that REG␣/␤ binds the proteasome tighter than REG␥, which in turn binds tighter than REG␣ or REG␤ .
REG-␣, -␤, and -␥ Bind Ca 2ϩ -We have previously shown that REG␣ and a Ub fusion protein containing the REG␣ KEKE insert (Ub-KEKE) are able to bind Ca 2ϩ (51). Using a 45 Ca 2ϩ overlay assay, we compared the calcium binding properties of the three REG homologs, calmodulin, ubiquitin, and Ub-KEKE. REG␣, -␤, and -␥, the Ub-KEKE fusion protein, and calmodulin bind 45 Ca 2ϩ in the presence of 5 mM MgCl 2 , whereas no signal could be detected for ubiquitin (Fig. 10). We also showed previously that the addition of calcium inhibited peptide hydrolysis by REG␣-proteasome complexes (51). When we repeated these measurements, concentrations of Ca 2ϩ that were 3-4-fold higher than previously reported were needed to inhibit peptide cleavage by proteasome-REG␣ complexes (not shown). The reason for this discrepancy with our previous findings is not known. DNA sequencing indicates that no mutations occurred in our expression plasmid, so we suspect that the difference reflects some unexplained variability in our proteasome and/or recombinant REG preparations. However, the results in Fig. 10 and those in Ref. 51 do show that each of the recombinant REGs can bind calcium, although with a lower affinity than other well characterized Ca 2ϩ -binding proteins such as calmodulin.
The 26 S Proteasome Is Unaffected by High Levels of REG Homologs--It has been shown that REGs do not promote FIG. 7. Formation of an active REG oligomer from REGm␣ and REG␤ monomers. Partially purified REGm␣ (750 g) or REG␤ (750 g) was fractionated individually (m␣ and ␤) or as a mixture (m␣/␤; 750 g of total protein). Eluting proteins were localized by absorption at 280 nm (A 280 ; panels A-C). Aliquots (20 l) of selected fractions from each elution were tested in a fluorometric assay (see "Experimental Procedures") for proteasome-stimulating activity. Panel D shows the activity measured after elution of REGm␣ (m␣), REG␤ (␤), or a REGm␣/REG␤ mixture (m␣␤). Selected fractions were also analyzed on SDS-polyacrylamide gels stained with Coomassie Blue R250 (panels E-G). Each gel is shown next to the corresponding elution profile. Fractions from the active peak in panel G (5 g of total protein) were analyzed by RP-HPLC on a 4.6 ϫ 250-mm C18 column using a 0 -45% acetonitrile gradient in 0.1% trifluoroacetic acid (panel H). The eluting proteins were detected by absorption at 214 nm. hydrolysis of any fluorogenic peptide that is not already a substrate of the proteasome (32), and they do not convert the proteasome to an enzyme capable of degrading intact, folded proteins (16,17). It is known that the red blood cell 11 S REG is displaced from proteasomes by the regulatory complex of the 26 S proteasome (52). We have reexamined these points using each recombinant REG. Purified human proteasomes were unable to degrade 125 I-lysozyme or 125 I-Ub-lysozyme conjugates in the presence of REG␣, REG␤, REG␥, or REG␣/␤ complexes (data not shown). Furthermore, amounts of recombinant REG homologs (molar concentrations estimated to be 2000-fold higher than the 26 S proteasome) were unable to prevent hydrolysis of 125 I-Ub-lysozyme conjugates in Xenopus egg extracts (not shown). These results confirm that the 11 S REGs or individual REG subunits do not convert the proteasome to an enzyme capable of degrading 125 I-lysozyme or 125 I-Ub-lysozyme conjugates. Furthermore, they do not impair the function of the 26 S proteasome in Xenopus laevis egg extracts. DISCUSSION The experiments presented in Figs. 3 and 4 demonstrate that each of the three REG homologs, ␣, ␤, and ␥, activates peptide hydrolysis by the proteasome. For REG␣ and REG␥, this observation is not surprising, since the two proteins are closely related (Fig. 1), each forms a defined oligomer of about 200 kDa (Fig. 6), and recombinant REG␣ was shown previously to activate the proteasome (18). The finding that REG␤ is a proteasome activator might be considered unexpected for two reasons. First, there are two reports that REG␤ does not activate the proteasome (2,53). Second, we have produced 31 monomeric mutants of REG␣ that cannot stimulate peptide hydrolysis by the proteasome. 3 Because REG␤ chromatographs on Superdex 200 as a monomer (see Fig. 6), one might expect it to be inactive as well. However, relatively high concentrations of REG␤ are required to detect its activity. In fact, REG␤ appears to be inactive in Figs. 7 and 8 because the fractions were assayed prior to concentration. High concentration might promote REG␤ heptamers or hexamers that are either required for or are more efficient at proteasome activation. Alternatively, monomers of REG␤ may assemble into a heptameric ring only on the proteasome surface. It should be noted that dynamic monomer-oligomer equilibrium for ␤ subunits may explain the biphasic activation curve obtained for this REG homolog (see Fig. 2B).
The activity of REG␤ is difficult to detect for another reason: it is temperature-sensitive. Although the data in Fig. 4 show REG␤ active at 42°C, other REG␤ preparations were markedly inactivated at 35°C. Accordingly, enzyme reactions incubated at 37°C may not detect REG␤ activity, and this is the reason that we assayed the recombinant REG homologs at room temperature rather than at 37°C as in previous studies (18,53). In conclusion, we are confident that REG␤ activates the proteasome, but its activity is more difficult to detect than that of REG␣ or REG␥.
Although each REG homolog activates the proteasome, there are differences in the extent to which they stimulate hydrolysis of specific fluorogenic peptides. REG␣ and REG␤ activated the cleavage of 9 of the 10 peptides listed in Table I. Indeed, the two proteins produce virtually identical patterns of stimulated peptide hydrolysis. By contrast, REG␥ barely increased the cleavage of LLE-␤NA and LLVY-MCA, yet it activated the proteasome to hydrolyze all of the peptides with Arg or Lys next to the fluorescent leaving group. As noted in the Introduction, REGs have been implicated in the generation of peptides presented to the immune system on MHC-I molecules. Presented peptides are 8 -11 residues long, and their C termini almost always consist of hydrophobic or basic amino acids (54). In this context, it appears that of the three REG homologs, ␥ would produce peptides best suited for class I presentation, since it does not markedly increase hydrolysis after the negatively charged residue in LLE-␤NA. Of course, this inference rests on the controversial assumption that the substrate specificities obtained using small fluorogenic peptides reflect cleavage preferences in FIG. 9. Direct binding assay for relative association of REG homologs with proteasomes. Analysis of proteasome-bound REGs. Specific antibodies directed against REG␣ (upper panels) or REG␥ (lower panels) were used to detect REG binding to proteasomes in a competition assay, REG homologs were incubated with antibody-tethered proteins and rinsed, and the bound species were identified using homolog-specific antibodies (see details under "Experimental Procedures"). ; REGm␣ (m␣); Ub; a ubiquitin fusion protein extended at the C terminus with the KEKE insert of REG␣ (DPVKEKEKEERKKQQEKEDKDEKKKGEDEDK (Ub-KEKE)); and calmodulin (CaM) were applied to a nitrocellulose filter using a slot blot apparatus. The filter was exposed to 45 Ca 2ϩ (2 Ci/ml) in 10 mM imidazole, pH 6.8, 60 mM KCl, 5 mM MgCl 2 at room temperature for 15 min (64), briefly rinsed with distilled water, and exposed to x-ray film. longer, naturally occurring peptides and proteins (see Refs. 32 and 55 for discussions on this point).
The distinct patterns of enhanced peptide hydrolysis produced by the REG homologs provide clues regarding the mechanism(s) by which they activate the proteasome. The recently published x-ray structure of the yeast proteasome reveals that the enzyme's active sites are virtually inaccessible from the particle's surface (10). Hence, it seems almost certain that activators like the 19 S regulatory complex (44,56,57) or 11 S REGs (16,17) increase substrate access to the central proteolytic chamber within the proteasome. The generation of channels leading into the proteasome could be the only mechanism for activation of peptide hydrolysis. For example, association of the rabbit 19 S regulatory complex with the proteasome results in a uniform 3-fold faster hydrolysis of LLE-␤NA, PFR-MCA, and LLVY-MCA (52). Since it is well established that each of these peptides is hydrolyzed by a specific proteasome ␤ subunit (58 -60), equivalent stimulation for each fluorogenic peptide can be interpreted in favor of increased substrate access or product egress as the mechanism by which the 19 S regulatory complex activates peptide hydrolysis. Alternatively, binding of proteasome activators might also induce conformational changes that increase the catalytic efficiency of one or more proteasome ␤ subunits. This would be consistent with activation by the 11 S REGs, since they increase the hydrolysis of specific peptides to widely varying extents. Furthermore, as shown in Table I, the pattern of activation by REG␥ differs significantly from that produced by REG␣ or REG␤. Taken at face value, the data in Table I indicate that all three REG homologs activate the proteasome trypsin-like active site(s), whereas only REG␣ and REG␤ activate the subunit(s) largely responsible for cleavage of LLE-␤NA and LLVY-MCA. We conclude that REG homologs activate proteasomes both by opening channels to the particle's central chamber and by directly activating specific subunits in the proteasome ␤-rings.
Comparisons presented in Fig. 1B show that the amino acid sequences of REG␤ and REG␥ diverge about equally from the sequence of REG␣. With this in mind, it is interesting that REG␣ and REG␤ activate hydrolysis of specific peptides in an identical manner quite distinct from the pattern of activation produced by REG␥. The insert regions of REG␣ and REG␤ may explain their equivalent activation properties at saturation. Although the REG␤ insert is shorter than that of REG␣, both inserts are highly charged regions characterized by "alternating" lysine and glutamate residues. We have previously speculated that such regions, known as KEKE motifs, promote association of proteasomes with the 19 S regulatory complex and the 11 S regulator (61). Here, we suggest that interaction between the KEKE motifs found at the C termini of proteasome ␣ subunits, C6 and C9, and the KEKE inserts in REG␣ and REG␤ results in enzymatic activation of those proteasome ␤ subunits responsible for cleaving LLE-␤NA and LLVY-MCA. The distinctly different sequence of amino acids in REG␥'s insert would, according to this model, explain its inability to stimulate hydrolysis of these two fluorogenic peptides. Alternatively, the inserts may promote homolog-specific association or bind other cellular factors. The ideas that REG inserts influence peptide cleavage specificity or partner selection are, without doubt, speculative. Fortunately, these ideas can be readily tested by reciprocal transfer of insert regions between REG homologs.
Several experiments have revealed highly specific interactions among the REG homologs. The gel filtration profiles in Fig. 6 demonstrate strong interactions between REG␣ and REG␤ subunits and the absence of interaction between REG␥ and REG␤. The conversion of REG␤ monomers to a distinct oligomeric species when mixed with REG␣ heptamers (Fig. 6, B and C) can be rationalized by assuming that the REG␣ heptamers are in equilibrium with a small pool of ␣ monomers. The addition of ␤ subunits would shift the equilibrium to a mixed oligomer because of the high affinity between ␣ and ␤ subunits. The absence of interaction between REG␤ and REG␥ could result either from a lack of inherent affinity between the two proteins or from the absence of a significant monomer-heptamer equilibrium for REG␥. The filter binding assay in Fig. 5 favors the first possibility, since REG␥ subunits clearly bind themselves but not REG␣ or REG␤. Presumably, the assay in Fig. 5 works because filter-bound monomers or small oligomers of each REG are in conformations that permit formation of specific subunit interfaces with their binding partners. Thus, both solution experiments and filter binding assays indicate that among REG homologs the principal oligomeric species will consist of ␣/␤, ␣/␣, and ␥/␥ subunits.
The absence of oligomers of recombinant REG␤ and its high affinity for ␣ subunits strongly suggest that REG␤ subunits will be present as hetero-oligomers with REG␣ in the 11 S complexes isolated directly from mammalian cells. Recent data also indicate that REG␣ and REG␤ are coordinately regulated by ␥-interferon in a fashion distinct from the regulation of REG␥ (19). These findings taken together suggest that 11 S REGs obtained from mammalian cells will likely consist of distinct REG␥ heptamers or REG␣-REG␤ complexes. Whether these presumed REG␣-REG␤ oligomers will be found to contain six or seven subunits remains, in our opinion, an open question.
The recombinant REG homologs not only display distinct affinities for each other; they bind the proteasome with characteristic affinities. The enzymatic competition assays in Fig. 8 and the direct binding assays in Fig. 9 yield a consistent hierarchy of proteasome binding in which heteromeric REG␣/␤ complexes bind tighter than REG␥ and these two complexes bind tighter than either REG␣ or REG␤ alone. There is no reason to believe that these measurements are in error. Still, they may not be physiologically relevant for two reasons. First, it seems likely that proteasomes are in excess within mammalian cells (62). Consequently, there may not be significant competition in vivo. Second, isoelectric variants of both REG␣ and REG␤ are present in the 11 S REG isolated from human red blood cells (17). Quite possibly, ␣ and ␤ subunits are posttranslationally modified in mammalian cells, and such modifications could affect their affinities for proteasomes and/or for each other.
In summary, we have shown that each of three recombinant REG homologs is a proteasome activator with distinct biochemical properties. Previous comparisons between recombinant REG␣ and the 11 S REG purified from red blood cells revealed little difference in their ability to activate peptide hydrolysis by the proteasome (32). Presumably, the properties reported here will apply to the 11 S REG species isolated directly from mammalian cells. But this is clearly a presumption in view of potential modifications to the REG homologs synthesized in mammalian cells. Characterization of REGs purified from mammalian tissues, especially REG␥, and comparison with the recombinant REG species is under way. This approach should provide further insight into the roles of the proteasome activators in vivo.