Multiple (a-NH-ubiquitin)Protein Endoproteases in Cells*

Ubiquitin is encoded as a variable, spacerless repeat of the gene terminating with an additional amino acid or as a gene coding for a single ubiquitin with a car- boxyl-terminal extension of 52 to 80 amino acids. We report the identification and partial purification of enzymes that specifically hydrolyze the peptide bond between ubiquitin-ubiquitin conjugate (Ub-Ubase) or ubiquitin fusion proteins (Ub-Xase). The Ub-Ubase was separated from the Ub-Xase by dye-ligand-seph-arose chromatography. The Ub-Xase was purified fur- ther by affinity chromatography on ubiquitin-sepha-rose. The fidelity of the endoprotease reaction was assessed by measuring the ability of the released ubiquitin to be activated by ubiquitin-activating en- zyme ( E l ) which requires intact ubiquitin and by sequence analysis of the released carboxyl extension pro- tein with 52 amino acids after endoproteolysis of human ubiquitin with 52-amino acid carboxyl extension. The failure of both Ub-Ubase and Ub-Xase to cleave a mutant ubiquitin-Gly-76-+Ala-metallothionein showed that the endoproteases distinguish Gly-X precipitated with 20% trichloroacetic acid and washed once with 5% trichloroacetic acid. The 3H contained in the pellet, dissolved in 100 p1 of 0.1 N NaOH, was measured by liquid scintillation spectrometry. Stimulation of Ubiquitin-dependent 32PPi-ATP Exchange Reaction with E,-The conditions employed for this assay have been described (6). The [32P]ATP formed during the reaction is measured after adsorption to acid-washed charcoal.

ubiquitin with a carboxyl-terminal extension of 52 or 80 amino acids (HUBCEP-52 or -80),' to ubiquitin and CEP-52 or -80, respectively (17). Studies with polyubiquitin mutants have shown that the gene is induced under conditions of stress such as heat shock or starvation (1,18). Monoubiquitin is used to target excess/abnormal proteins produced during the stress for proteolysis (18).
Ubiquitin carboxyl extension proteins provide additional sources of ubiquitin to regulate ubiquitin conjugates in cells (19)(20)(21). The gene sequence for the carboxyl extension proteins (CEPs) is well conserved and encodes a basic protein containing -31% lysine and arginine with features suggesting that CEPs direct natural fusions to the nucleus (22). CEPs contain metal binding motifs or "zinc finger sequences" which are found in factors that regulate transcription, receptors, and regulatory proteins that bind to nucleic acids (23- 25).
We report the partial purification and characterization of enzymes that hydrolyze precursor polyubiquitin and ubiquitin-fusion proteins, both of which liberate monoubiquitin. One enzyme cleaves polyubiquitin to monoubiquitin (Ub-Ubase) and others hydrolyze (a-NH-ubiquitin)protein conjugates to monoubiquitin (Ub-Xases).

MATERIALS AND METHODS
Ubiquitin-activating enzyme (E1) used in these studies was kindly provided by Dr. Avram Hershko, Haifa, Israel. Ubiquitin-aldehyde (Ub-al) and ubiquitin-methionine (Ub-Met) were gifts of Dr. Irwin A. Rose, Philadelphia, PA. Rabbit reticulocyte-rich blood was obtained from Pel-Freeze, and 10-20 or 17-27% polyacrylamide gradient SDS gels were obtained from The Integrated Separation Systems, Hyde Park, MA. The green dye-ligand-Sepharose was obtained from Pierce. [2,fb3H]ATP was purchased from Du Pont-New England Nuclear.
Preparation of Substrates-Pentaubiquitin expressed in Escherichia coli was purified as described previously (lo), with the last step (preparative electrophoresis of pentaubiquitin) being replaced by Mono S ion exchange chromatography. 0-1 M KC1-eluted fractions contained pentaubiquitin with varying amounts of contaminating tetra-and triubiquitin but not di-and monoubiquitin. These fractions were used as substrate for Ub-Ubase. The normal and mutant forms of ubiquitin or Ub-Mt used in this study were purified according to published procedures (9,30). The construction, expression, isolation, and properties of HUBCEP-52 and HUBCEP-80 form E. coli are published (17).
Partial Purification of (a-NH-ubiquitin1Protein Endoproteases-"Ub-Ub" and the "Ub-X" endoproteases were partially purified from reticulocyte extracts. "Fraction 11" was prepared as described previously (27). Fraction-I1 was applied to an ion exchange column (Fast The abbreviations used are: HUBCEP-52, human ubiquitin with 52 amino acid carboxyl extension; CEP-52, carboxyl extension protein with 52 amino acids; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; Ub, ubiquitin; Ub-al, ubiquitin where the terminal glycine acid group was converted to an aldehyde group; Ub-Met, ubiquitin with an additional methionine; DTT, dithiothretol; Ub-Mt, ubiquitin-metallothionein; Ub-Ubase, Ub-Ub endoprotease; Ub-Xase, Ub-X endoprotease. Flow Q Sepharose) equilibrated with 50 mM Tris-HC1, pH 7.5, containing 2 mM DTT and washed with 5 column volumes of this buffer. The adsorbed proteins were eluted with a gradient of 0-1 M KCl. Aliquots were assayed for (a-NH-ubiquitin)protein endoprotease activity with pentaubiquitin or ubiquitin-metallothionein (Ub-Mt), which were expressed and isolated from E. coli (10,26). Two peaks (Peak-I and Peak-2 of Fig. 1, A and B ) of pentaubiquitin hydrolysis to monoubiquitin were observed. However, when these column fractions were assayed with Ub-Mt as substrate, there was one peak of Ub-Mt hydrolysis, corresponding to Peak-2 of Fig. 1B (data not shown). Ub-Mt is hydrolyzed to ubiquitin and metallothionein. However in SDS-PAGE monoubiquitin and metallothionein migrate as a single band, and the conditions to observe these as separate bands are published elsewhere (30). Peak-1, eluted by 0.1-0.2 M KC1 predominantly hydrolyzed pentaubiquitin to monoubiquitin (Ub-Ubase). Peak-2, eluted by 0.4-0.5 M KC1 predominantly hydrolyzed Ub-Mt to monoubiquitin and metallothionein (Ub-Xases). After dilution, 2-3fold with 50 mM Tris-HC1, pH 7.4, Peak-1 was applied to a column of Green Dye-Ligand-Sepharose. This column was washed (described above) and eluted with a gradient of 0-0.6 M KCl. The Ub-Ubase (detected by the hydrolysis of pentaubiquitin to monoubiquitin) eluted from this column at -0.1 M KC1 and was not contaminated with Ub-Xases (detected by the hydrolysis of ubiquitin-metallothionein to monoubiquitin and metallothionein (30) or HUBCEP-52 to monoubiquitin and CEP-52). The Ub-Xases from Peak-2 were purified further from the Ub-Ubase by covalent affinity chromatography on ubiquitin Sepharose (-4.5 mg of ubiquitin/ml Sepharose). Ubiquitin-Sepharose (linked through the a-or +amino groups of monoubiquitin to the Sepharose (27)) was equilibrated with 50 mM Tris-HC1, pH 7.4. The Ub-Xases were brought to the above buffer concentrations and applied to the column, which was washed with 5 volumes of 50 mM Tris-HC1, pH 7.4, and then with the same volume of buffer containing, in addition, 1 M KC1. The affinity matrix was washed with 5 volumes of 50 mM Tris-HC1, pH 7.4. The Ub-Xase(s) was eluted with 5 volumes of Tris-HC1, pH 9, containing 10 mM DTT. Fractions were collected into tubes containing one-tenth the volume of 1 M Tris-HC1, pH 7. The partially purified Ub-Xase had no contaminating Ub-Ubase activity. The presence of Ub-Ubase and Ub-Xases in Fraction I1 is shown in Fig. 2

Assays for Monoubiquitin with Ubiquitin-activating Enzyme (E,)-
Ubiquitin formed during the ree-tion of Ub-Mt or HUBCEP-52 with (a-NH-ubiquitin)protein endoprotease was assayed by analysis with E,. The endoprotease reaction mixture (20 pl, final concentrations of 50 mM Tris-HC1, pH 7.4, and 2 mM DTT), was incubated for 30 min at 37 "C, transferred to a 70 "C water bath for 10 min and then centrifuged in a microfuge for 10 min. The supernatant was assayed for El reactivity with ubiquitin released duringproteolysis. The ability of E, to activate ubiquitin was assayed by the following methods.
Ubiquitin-dependent End Point Assay with E,-The procedure was similar to one published previously (7). Ubiquitin-activating enzyme was treated with iodoacetamide and incubated with 50 mM Tris-HC1, pH 7.4, 5 mM MgC12, 0.2 mM EDTA, 0.05 unit of creatinine phosphotransferase, 1.5 mg/ml bovine serum albumin, and 10-20 pmol of [2,8-3H]ATP. The E , .3H]AMP.Ub complex formed during the reaction was precipitated with 20% trichloroacetic acid and washed once with 5% trichloroacetic acid. The 3H contained in the pellet, dissolved in 100 p1 of 0.1 N NaOH, was measured by liquid scintillation spectrometry.
Stimulation of Ubiquitin-dependent 32PPi-ATP Exchange Reaction with E,-The conditions employed for this assay have been described (6). The [32P]ATP formed during the reaction is measured after adsorption to acid-washed charcoal.

Partial Purification and Properties of Ub-Ubme and Ub-
Xase-Ion exchange and affinity chromatography separated the Ub-Ubase from the Ub-Xases activities ("Materials and Methods"). Fraction-I1 after ion exchange chromatography gave a profile of absorbance at 280 nm (Fig. lA). Two peaks of Ub-Ubase activity were observed; hydrolysis to monoubiquitin was detected by the decrease in the 40-kDa pentaubiquitin band and the simultaneous appearance of a 5-kDa band of monoubiquitin on 18% SDS-PAGE (Fig. 1B). This band was shown to be monoubiquitin by immunoblots (data not shown) in addition to the El assays (Table I). The column fractions with Ub-Ubase activity are indicated as Peak-1 and Peak-2 in Fig. 1, A and B. The column fractions were also assayed for Ub-Xase activity. With Ub-Mt as substrate, there was only one major peak of Ub-Mt hydrolysis (data not given). This peak of Ub-Xase activity corresponded to Peak-2 in Fig.  1, A and B. Peak-1 (Fig. 1, A and B ) when further purified on Green Dye-Ligand-Sepharose hydrolyzed pentaubiquitin to monoubiquitin (Ub-Ubase) but did not hydrolyze Ub-Mt or HUBCEP-52 (Ub-Xases) (Fig. 2B). Peak-2 contained Ub-Xase activities and small amounts of Ub-Ubase activity (Fig.  1, A and B, and Fig. 2C). Peak-2 was purified further by adsorption on a ubiquitin affinity column, which separated the HUBCEP-52-hydrolyzing enzyme from the Ub-Mt-hydrolyzing enzyme and Ub-Ubase (Fig. 2 0 ) . Most of the Ub-Ubase and the Ub-Mt-hydrolyzing enzymes were found in the unadsorbed column breakthrough and the column wash fractions (data not given), indicating that ubiquitin affinity chromatography enriched the HUBCEP-52-hydrolyzing enzyme. The Ub-Ubase and the Ub-Xases had molecular mass ranges of -150 and -200 kDa, respectively, when chromatographed on G-200 and Superose-12 columns, indicating that they are separate proteins. The Ub-Ubase and the Ub-Xase activities were inhibited by -5 ~L M Iodoacetamide. Time Course for the Hydrolysis of Pentaubiquitin, with (a-NH-ubiquitin)Protein Endoprotease-To estimate the hydrolysis of pentaubiquitin, Ub-Mt, or HUBCEP-52 with the (a-NH-ubiquitin)protein endoproteases a reaction mixture (20 pl) was incubated for 30 min at 37 "C ( Fig. 3). A 5-kDa band migrated in a gradient SDS-PAGE with the mobility of monoubiquitin (Fig. 3). The absorbance of Coomassie Brilliant Blue R-stained substrate (pentaubiquitin, 40-kDa band) decreased with time and a 5-kDa band with the mobility of monoubiquitin simultaneously appeared and increased. The estimated rate for pentaubiquitin hydrolysis was -5-10 ng/ min/pg Fraction 11.
Estimation of the Ubiquitin Formed during the Endoproteolysis with El-The 5-kDa protein migrating as ubiquitin (Figs. 1-3), which appeared only when pentaubiquitin, Ub-Mt and HUBCEP-52 were hydrolyzed, was confirmed to be a functional ubiquitin molecule by the El assays. These assays estimate only the 76-amino acid ubiquitin released during endoproteolysis. Specificity of the El reaction was demonstrated by showing that mutants of ubiquitin were not substrates (Table I). The inability of E l to activate Ub-al, Ub-Met, Ub-A-76, Ub-Mt, HUBCEP-52, Ub treated with trypsin, or pentaubiquitin shows the specific requirements of E1 for ubiquitin with the proper COOH terminus. The results of the El assays (Table I)  amino terminus in the conjugate and that the first amino acid was methionine. Thus Ub-Mt and HUBCEP-52 isolated from E. coli was a contiguous protein and contained no isopeptide bonds. T o confirm that the bond hydrolyzed in HUBCEP-52 was Gly-76-X the CEP-52 released after hydrolysis of HUBCEP-52 by Fraction I1 was isolated and subjected to amino-terminal analysis. The first two amino acids were isoleucine indicating that the endoproteolytic hydrolysis of HUBCEP-52 was specific. Ub-Ubase and Ub-Xases present in Fraction I1 hydrolyzed pentaubiquitin, Ub-Mt, or HUBCEP-52 as shown by the appearance of 5-kDa monoubiquitin band with increasing time of incubation with all the substrates (Figs. 1-2). However, these endoproteases did not hydrolyze Ub-Gly-7bAla-Mt (where Gly-76 of Ub-Mt was altered to alanine) as no 5-kDa monoubiquitin band appeared with increasing time of incubation (Fig. 2 A ) .
To understand further the role of ubiquitin in the recognition of the substrates we attempted to inhibit the Ub-Xases with various ubiquitin mutants. Ub-Mt was incubated with Fraction-I1 and in the absence or presence of various ubiquitin mutants (Fig. 4). Except for Ub-a1 (where the band of Ub-Mt is not hydrolyzed), attempts to inhibit the Ub-Xase with ubiquitin mutants, methylated ubiquitin, human or yeast ubiquitin, were not successful (recognized by the loss of 16-kDa Ub-Mt band) (Fig. 4). Other results indicate that molar equivalents of Ub-a1 (compared to Ub-Xase in this case) were  Fig. 1) (10 pg) after Green-Sepharose chromatography was incubated with substrates according to conditions described above. The panel shows that Peak-1 hydrolyzed pentaubiquitin to monoubiquitin but did not hydrolyze Ub-Mt or HUBCEP-52. Panel C, Peak-2 ( Fig. 1) (10 pg) was incubated with the substrates according to the conditions described above. This fraction hydrolyzed all three substrates (pentaubiquitin, Ub-Mt, and HUBCEP-52). Panel D, pH 9 eluate, 1 pg (Peak-2 purified further on ubiquitin affinity column), was incubated with the substrates according to the conditions described above. This fraction hydrolyzed HUBCEP-52 but did not hydrolyze pentaubiquitin or Ub-Mt.
sufficient to inhibit the hydrolysis of Ub-Mt. Preliminary observations indicate that this inhibition by Ub-a1 may be irreversible. Similar results were obtained when pentaubiquitin was used as a substrate (data not shown) which indicates that Ub-a1 also inhibits Ub-Ubase. Thus any alteration of the Gly-76 residue of ubiquitin may alter the ability of the conjugate to be a substrate for both Ub-Ubase and the Ub-Xases ( Figs. 2A and 5). These results along with those showing the multiplicity of (a-NH-ubiquitin)protein endoproteases suggest that the endoproteases recognize sequences both in ubiquitin and in the conjugated proteins.

DISCUSSION
We have reported that pentaubiquitin expressed in E. coli could support the degradation of lZ5I-bovine serum albumin in the reticulocyte lysate but that isolated pentaubiquitin was not activated by E , (10). Thus, pentaubiquitin appeared to have been processed to monoubiquitin in the reticulocyte lysate by endoprotease(s), which presumably cleaved either the peptide bond between ubiquitin-ubiquitin and/or between ubiquitin and asparagine of the terminal repeat of pentaubiquitin (10, 13). A number of other observations suggest the existence of cellular (a-NH-ubiquitin)endoproteases. In studies with the ubiquitin-P-galactosidase (Ub-@Gal) fusion proteins in yeast, the expression of various Ub-@Gal fusions resulted in the release of monoubiquitin (28). In vitro translation of diubiquitin and truncated diubiquitin mRNAs yielded monoubiquitin, detected by immunoprecipitation or SDS-PAGE (29). Ubiquitin-metallothionein (Ub-Mt) or some of its mutant fusion proteins can be processed to ubiquitin in yeast or the reticulocyte extracts (26,30). Finally HUBCEP-52 and -80 are processed to ubiquitin and CEP-52 or -80, with the reticulocyte extracts (17). However, to date multiple enzymes responsible for processing (a-NH-ubiquitin)protein conjugates have not been identified or biochemically characterized. The present results demonstrate the existence of multiple (a-NH-ubiquitin)protein endoproteases (Ub-Ubase and Ub-Xase) and suggest that there may be different enzymes which hydrolyze various ubiquitinated conjugates (Ub-Mt-hydrolyzing enzyme, ubiquitin carboxyl-terminal hydrolyzing enzyme) (Fig. 2). The result that only one amino terminus was observed in Ub-Mt and HUBCEP-52 showed that the multiplicity observed in these reactions were not due to isopeptidases. The correct amino terminus of CEP-52 released after endoproteolysis of HUBCEP-52 and the quantitative release of monoubiquitin from the substrates during the endoproteolytic reaction (Table I), demonstrate that the hydrolysis catalyzed by the (a-NH-ubiquitin)protein endoproteases is specific. The separation of Ub-Ubase and Ub-Xases activities by ion exchange (Fig. 1) and molecular size columns suggest that they are separate enzymes. The elution of Ub-Ubase both at Peak-1 and Peak-2 may be due to protein-protein interactions because the behavior of the Ub-Ubase from Peak-1 or Peak-2 on Green-Sepharose or the sizing columns was similar. Further purification of the Ub-Xases on ubiquitin Sepharose (pH 9 fraction, Fig. 2) indicated that HUBCEP-52-hydrolyeing enzyme can be separated from the Ub-Mt-hydrolyzing enzyme. This result can be explained by the selective inactivation of the Ub-Mt-hydrolyzing enzyme during its purification on ubiquitin-Sepharose or the separation of the enzymes that hydrolyze Ub-Mt and HUBCEP-52 by the affinity matrix. The results show that the Ub-Mt hydrolyzing enzyme was found primarily in the unadsorbed column breakthrough and wash fractions whereas the HUBCEP-52-hydrolyzing enzyme was found in the pH 9 eluate fractions. Thus they are separate proteins.
Attempts to inhibit the endoproteolytic reaction with various mutants of ubiquitin were unsuccessful (Fig. 4). However, results indicate that (Ub-al) a t concentrations equal to those of Ub-Ubase and Ub-Xases irreversibly inhibits both endoproteases (Fig. 4). The likely mechanism for the inhibition of the endoprotease activity is the binding of the substrate analogue (Ub-al) to the substrate binding site on the enzymes. The endoproteases may also be specific for a particular ubiquitin conjugate and the inability of the enzymes to hydrolyze Ub-Gly-7bAla-Mt (Fig. 2 A ) , indicate their specificity for a Gly-X peptide bond. This hypothesis is supported by the result that yeast with plasmids expressing Ub-Gly-7bAla-Mt grow a t a reduced rate compared to yeast containing plasmids expressing Ub-Mt (30).
Our results suggest that several (a-NH-ubiquitin)protein endoproteases are involved with the maintenance of the cellular content of ubiquitin. Ub-Ubase may be induced when cells are subjected to stress conditions, such as heat shock or starvation. This hypothesis is supported by the previous observation in yeast that deletion of the polyubiquitin gene results in marked susceptibility to heat, starvation, sporulation, or conditions for spore germination (1,18). The enhanced susceptibility of these cells may have been due to the increased accumulation of structurally and/or functionally abnormal proteins. The polyubiquitin gene in normal yeast induced under similar stress conditions could provide the cells with quantities of ubiquitin needed to complex with and target the abnormal proteins for energy-dependent proteolysis. Factors regulating the induction of the polyubiquitin gene may, in addition, regulate the amounts of Ub-Ubase in cells. Normal cell growth is not affected by deleting the yeast polyubiquitin gene (1,18). Thus cells may have other enzymes, for example, Ub-Xase(s), ubiquitin carboxyl-terminal hydrolases, ubiquitin-dependent proteases, or isopeptidases to maintain homeostasis between ubiquitin and ubiquitin conjugates under normal growth conditions (19).
Ub-Xase(s) not only help in maintaining appropriate levels of ubiquitin in cells but may play a major role in cell regulation. For example, the (a-NH-ubiquitin)CEP conjugates appear to be regulated by post-or co-translational processing to give ubiquitin and CEPs in cells (19,20). Genetic analyses with the yeast natural fusion protein genes (UBI-1, UBI-2, UBI-3 coding for the yeast ubiquitin carboxy extension proteins) show that mutations in any of the three UBI loci lead to a prolonged GI phase in the cell cycle and that a double mutation of UBI-l/UBI-2 is lethal (1). Purification and characterization of the individual (a-NH-ubiquitidprotein endoproteases will help in understanding further the cellular roles of these enzymes.