A proteasome activator subunit binds calcium.

We recently cloned a cDNA encoding the 29-kDa subunit of human red blood cell regulator (REG), a potent activator of the multicatalytic protease (Realini, C., Dubiel, W., Pratt, G., Ferrell, K., and Rechsteiner, M. (1994) J. Biol. Chem. 269, 20727-20732). The sequence of this subunit contains 28 "alternating" lysine and glutamic acid residues (a KEKE motif). Similar regions are present in a number of Ca(2+)-binding proteins, and using standard filter assays, the recombinant protein is shown to bind 45Ca2+ and ruthenium red. 45Ca2+ is also bound to a ubiquitin extension protein containing the 28-residue KEKE region from the 29-kDa REG subunit. Thus, the 29-kDa REG subunit is a Ca(2+)-binding protein, and its KEKE region is able to bind divalent cations. Ca2+ reversibly inhibits the enhanced peptidase activity of complexes between the multicatalytic protease and recombinant REG. This raises the possibility that multicatalytic protease activity is regulated by calcium in vivo.

The multicatalytic protease (MCP) 1 or 20 S proteasome is a large (ϳ700 kDa) multimeric enzyme found in eukaryotes, prokaryotes, and archaebacteria (for reviews see Refs. [1][2][3]. MCP subunits range in molecular mass from 20 to 30 kDa and can be placed in two families based on their homology to unique ␣and ␤-subunits present in the archaebacterium, Thermoplasma (4). The assembled enzyme consists of four stacked rings that form a hollow cylinder (5). The outer rings consist of 7 subunits of the ␣ family, and the two inner rings each contain 7 ␤-type subunits (6,7). It has been suggested that the enzyme's active sites line a central aqueous channel (3), and the recent x-ray structure of MCP from Thermoplasma confirms this suspicion (8). Site-directed mutagenesis (9) and covalent labeling with a novel protease inhibitor (10) identify the Nterminal threonine residues of ␤-subunits as active site nucleophiles. Thus, MCP is the first example of a threonine protease.
In eukaryotes, MCP serves as the proteolytic core for two larger complexes. In an ATP-dependent reaction, MCP can associate with a regulatory complex that contains at least 15 subunits with apparent molecular masses between 25 and 110 kDa (11)(12)(13). This generates the ATP-dependent 26 S protease responsible for the degradation of ubiquitinated proteins (14) and unmodified ornithine decarboxylase (15). Alternatively, in the absence of nucleotides MCP can associate with an 11 S protein complex that we call the regulator (REG), thereby producing a markedly activated peptidase (16,17). REG binds to each end of MCP (18) and stimulates hydrolysis of selected fluorogenic peptides as much as 50-fold. SDS-polyacrylamide gel electrophoresis of regulator from human red blood cells revealed 2 subunits with apparent molecular masses of 31 and 29 kDa (17). We have recently cloned and expressed the gene for the 29-kDa subunit of human regulator (rREG 29K ), and we have shown that the recombinant species activates MCP in a manner very similar to the molecule purified from human red blood cells (19).
A stretch of 28 alternating lysines and glutamate residues is a striking feature of the amino acid sequence of the 29-kDa subunit. We call such regions KEKE motifs and have proposed that these highly charged regions represent association domains (20); Perutz has also suggested that alternating positive and negative amino acids or "polar zippers" play a role in protein-protein association (21). KEKE motifs are found in a variety of proteins including subunits of the 26 S protease and MCP, microtubule-associated protein 1B, myosin phosphatase, triadin, and chaperonins such as hsp70 and hsp90 (20). They are also present in the calcium-binding proteins calreticulin, calnexin, endoplasmin, and Ca 2ϩ -dependent adenosine triphosphatase (Ca 2ϩ -ATPase). Calreticulin contains two distinct Ca 2ϩ -binding regions, one of which is a KEKE motif that binds calcium with high capacity and low affinity (22). Because KEKE motifs are present in Ca 2ϩ -binding proteins and because Ca 2ϩ is an important regulator of cellular processes (23,24), we asked whether the 29-kDa REG subunit is capable of binding Ca 2ϩ . Here we report that rREG 29K binds 45 Ca 2ϩ , ruthenium red, and the cationic dye, carbocyanine. Using recombinant technology, we show that, when appended to ubiquitin, the KEKE motif from rREG 29K confers Ca 2ϩ binding to the chimeric protein. Furthermore, concentrations of Ca 2ϩ in the midmicromolar range reversibly inhibit the peptidase activity of rREG 29K -MCP complexes.

EXPERIMENTAL PROCEDURES
Materials-Calmodulin was obtained from Boehringer Mannheim and ubiquitin from Sigma. The fluorogenic peptide succinyl-Leu-Leu-Val-Tyr-MCA (sLLVY-MCA) was obtained from Peninsula Laboratories. Ruthenium red and Stains All were from Spectrum Chemicals (New Brunswick, NJ); 45 Ca 2ϩ (specific activity, 10 mCi/mg) was purchased from DuPont NEN.
Preparation of REG, Ubiquitin-KEKE, Ub-KEKE 1 ⁄2, and Multicatalytic Protease-Recombinant REG was purified to homogeneity from Escherichia coli cells induced with isopropyl-1-thio-␤-D-galactopyranoside (19). Briefly, recombinant cells were lysed using a French press, and the cell lysate passed over an anion exchange column. Fractions from the peak of activity were pooled, and the regulator was further purified by sizing chromatography. The recombinant regulator was greater than 95% pure as determined by electrophoresis on a SDSpolyacrylamide gel followed by silver staining. Ub-KEKE 1 ⁄2 (Ub-DPV-KEKEKEERKKQQEK) and Ub-KEKE (Ub-DPVKEKEKEERKKQQ-EKEDKDEKKKGEDEDK) were expressed and purified according to published procedures (25). Multicatalytic protease was purified from outdated human red blood cells as described (26).
Fluorometric Protease Assays-Spectrofluorometric assays consisted of 100 M sLLVY-MCA incubated in the presence of MCP and various amounts of recombinant REG in a final volume of 50 -100 l of 10 mM phosphate, pH 7.4. Reactions were initiated by the addition of fluoro-genic peptide and terminated by adding 200 l of cold 100% ethanol. Fluorescence was measured on a Perkin-Elmer fluorometer using an excitation wavelength of 380 nm and an emission wavelength of 440 nm. Iminodiacetic acid was used to eliminate contaminating cations from protein solutions and buffers.
Binding of REG, Ub-KEKE, Ub-KEKE 1 ⁄2, Ubiquitin, and Calmodulin to Ruthenium Red, Carbocyanine, and 45 Ca 2ϩ -Purified rREG 29K , ubiquitin, ubiquitin-KEKE fusion proteins, and calmodulin were applied to a nitrocellulose filter (0.2 M, Schleicher and Schuell) using a slot blot apparatus and tested as described in Ref. 27 for their ability to bind 2 M 45 Ca 2ϩ (1 Ci/ml) and ruthenium red (28) in the presence of 0, 1, or 5 mM MgCl 2 . Interaction of soluble proteins with the cationic carbocyanine dye, Stains All, was measured as described (29).

The 29-kDa REG Subunit Is a Calcium-binding Protein-
There are three generally accepted methods for measuring calcium binding to proteins. Some calcium-binding proteins interact with the cationic carbocyanine dye, Stains All, and produce a characteristic absorption peak at 615 nm (29). Similarly, ruthenium red has been shown to bind many proteins capable of forming complexes with Ca 2ϩ (28). Finally, direct association of 45 Ca 2ϩ to filter-bound proteins is considered diagnostic for Ca 2ϩ -binding proteins (27). We used all three methods to determine whether the rREG 29K subunit binds calcium. Because we suspected that the KEKE region in the subunit confers calcium binding, we constructed two ubiquitin peptide extensions. One fusion protein consists of ubiquitin followed by 31 amino acids (DPVKEKEKEERKKQQEKEDK-DEKKKGEDEDK); the other is ubiquitin to which the first 17 amino acids of the KEKE motif (DPVKEKEKEERKKQQEK) are appended. These chimeric proteins were also assayed for their ability to bind Ca 2ϩ . rREG 29K was purified from E. coli lysates as described under "Experimental Procedures" and mixed with 0.001% Stains All; equivalent solutions containing calmodulin or ubiquitin were prepared for comparison. The absorption spectra obtained from the three proteins are presented in Fig. 1A. A peak at 615 nm, characteristic of Ca 2ϩ -binding proteins, is present in the spectra from calmodulin and rREG 29K but absent in the spectrum from ubiquitin. According to the Stains All assay, rREG 29K qualifies as a Ca 2ϩ -binding protein. This conclusion is supported by both ruthenium red and 45 Ca 2ϩ binding assays. The slot blots in Fig. 1B show that rREG 29K , calmodulin, and Ub-KEKE fusion proteins bind radioactive calcium and ruthenium red. Neither 45 Ca 2ϩ nor ruthenium red was bound by ubiquitin, and there was only minimal binding of 45 Ca 2ϩ to Ub-KEKE 1 ⁄2.
We also measured Ca 2ϩ binding to the test proteins in the presence of increasing concentrations of Mg 2ϩ (Fig. 2). Whereas ubiquitin and bovine serum albumin failed to bind Ca 2ϩ even in Mg 2ϩ -free solution, rREG 29K , Ub-KEKE, and calmodulin bound Ca 2ϩ in the presence of the competing cation. These results indicate that the 29-kDa REG subunit is a Ca 2ϩ -binding protein and that its KEKE motif is likely to confer this ability. Ub-KEKE 1 ⁄2 contains the first 14 residues of the KEKE motif; nonetheless, it failed to bind 45 Ca 2ϩ in the presence of competing Mg 2ϩ (Fig. 2). This suggests that either the entire KEKE motif is required for significant Ca 2ϩ binding or that the second half of the KEKE motif (EDKDDKKKGEDEDK) is responsible for interactions with Ca 2ϩ . Effect of Ca 2ϩ on the Activity of rREG 29K -MCP Complexes-To address the relevance of Ca 2ϩ binding to rREG 29K , we assayed MCP peptidase activity in the presence of rREG 29K , Ca 2ϩ , and EGTA. A standard peptidase assay was initiated and then rREG 29K was added, followed several minutes later by Ca 2ϩ ; EGTA was finally added after another several minutes. A typical spectrofluorometric trace is shown in Fig. 3A. Strong stimulation of peptidase activity was immediately observed upon addition of purified recombinant REG (Phase I 3 Phase II); peptide hydrolysis by the rREG 29K -MCP complex was inhibited by Ca 2ϩ (Phase III), but activity recovered after EGTA was added to chelate the Ca 2ϩ (Phase IV). MCP alone did not respond to the addition of Ca 2ϩ or EGTA (see Fig. 3B). The peptidase activity of rREG 29K -MCP complexes was inhibited by Ca 2ϩ in a dose-dependent manner (Fig. 3C); only 30% of the original activity remained at 300 M Ca 2ϩ . The inhibitory effect of Ca 2ϩ was not observed in the presence of EGTA, and the activity of MCP was only minimally affected by Ca 2ϩ in the absence of rREG 29K . The reversibility of the Ca 2ϩ -dependent inhibition of rREG 29K -MCP complexes is illustrated by the data in Fig. 3D; rREG 29K -MCP complexes preformed in the presence of 300 M Ca 2ϩ recovered up to 90% of their activity after increasing amounts of EGTA were added. The dose dependence and reversibility of Ca 2ϩ inhibition of rREG 29K -MCP complexes were observed using two separate preparations of MCP and three different preparations of recombinant REG. These effects were also observed with the 11 S regulator obtained directly from red blood cells. However, the Ca 2ϩ -activated protease, calpain, is present in the partially purified red blood regulator fraction, 2 which complicates interpretation of calcium effects on MCP peptidase activity.  1. Calcium binding by recombinant regulator. A, absorption spectra of Stains All in the presence of recombinant regulator (REG), calmodulin (CaM), and ubiquitin (Ub). Samples were 10 g of protein in 1 ml of 10 mM Tris, pH 8.8, 0.001% Stains All, and 0.1% formamide. B, binding of regulator, Ub-KEKE 1 ⁄2, Ub-KEKE, calmodulin, and ubiquitin to 45 Ca 2ϩ and ruthenium red (RR). The proteins (5 g) were applied to a nitrocellulose membrane and probed with either ruthenium red (25 mg/ml) or 2 M 45 Ca 2ϩ (1 Ci/ml) as described under "Experimental Procedures." Filters were stained with Ponceau S to confirm that equal amounts of protein were bound in each slot. . The largest family, by far, consists of proteins with EF-hands such as troponin C, calmodulin, or calpain. Other Ca 2ϩ -binding proteins use "elbows" (␣lactalbumin) or EF-hand-like motifs (annexins). Some Ca 2ϩbinding proteins do not possess either structure; nonameric repeats (LXGGXGNDX) in E. coli homeolysin (30) and acidic stretches in calsequestrin (31) have been proposed to be Ca 2ϩbinding sites. Ruthenium red and 45 Ca 2ϩ binding assays were used to show that recombinant fragments from calreticulin (22) and the ryanodine receptor (32) bind calcium in regions containing KEKE motifs. However, the Ca 2ϩ -binding sites were not precisely localized within the expressed peptides, which were generally 100 -200 amino acids long. The experiments presented above provide strong evidence that calcium binds directly to a KEKE motif since a Ub-KEKE extension protein bound 45 Ca 2ϩ and ruthenium red but ubiquitin did not (Fig. 2). This conclusion is reinforced by circular dichroism spectra from the free 31-residue KEKE peptide of REG. A significant loss of ␣-helix was observed in the presence of Ca 2ϩ . 3 KEKE motifs may generally be involved in binding Ca 2ϩ and/or Mg 2ϩ since they are present in triadin (33), a protein of the sarcoplasmic reticulum thought to bind Ca 2ϩ . Also, calnexin (34), endoplasmin (35), and Ca 2ϩ -dependent adenosine triphosphatase (36) contain such regions.
The trace in Fig. 3 shows that Ca 2ϩ binding can reverse the increased peptidase activity conferred by rREG 29K . This raises the possibility that calcium regulates proteolytic activity by the multicatalytic protease in vivo. It should be noted, however, that 60 M Ca 2ϩ was required to suppress peptidase activity by half; this concentration is higher than the accepted values of 0.1-10 M for intracellular calcium levels (37).
The experiments presented above do not address the mechanism by which Ca 2ϩ inhibits rREG 29K -MCP peptidase activity. The stimulatory effect of the REG depends on its physical interaction with MCP (17). Glycerol gradient and native gel analysis of rREG 29K -MCP mixtures suggest that the complexes dissociate in the presence of Ca 2ϩ , but the experiments are not conclusive because of the low affinity of rREG 29K for the multicatalytic protease. 4 It is also unclear that the Ca 2ϩ effect is mediated only by rREG 29K . Native MCP applied to nitrocellulose binds both 45 Ca 2ϩ and ruthenium red, 4 and ␣-subunits of MCP contain KEKE motifs (20). Thus, the rREG 29K -MCP interaction could result in conformational changes that activate Ca 2ϩ binding by MCP subunits, and this event might inhibit peptide hydrolysis. Discovering how Ca 2ϩ inhibits rREG 29K -MCP peptidase activity will require further experimentation.
In summary, we have demonstrated that the 29-kDa subunit of REG is a calcium-binding protein. We have also provided strong evidence that the KEKE motif present in rREG 29K is a Ca 2ϩ -binding site. Although the experiments show that Ca 2ϩ reversibly inhibits peptide hydrolysis by rREG 29K -MCP complexes, the molecular mechanism has not been discovered. Moreover, the physiological significance of this finding remains an open question. Nonetheless, there is a real possibility that intracellular calcium levels regulate proteolysis by the multicatalytic protease.