A Ubiquitin-specific Protease That Efficiently Cleaves the Ubiquitin-Proline Bond*

Ubiquitin is a small eukaryotic protein that is synthesized naturally as one of several fusion proteins, which are processed by ubiquitin-specific proteases to release free ubiquitin. The expression of heterologous proteins as fusions to ubiquitin in either prokaryotic or eukaryotic hosts often dramatically enhances their yield, and allows the exposure of any amino acid following cleavage of ubiquitin. The single exception is when proline is the amino acid immediately following ubiquitin; the ubiquitin-proline bond is poorly cleaved by presently studied ubiquitin-specific proteases. We show that the mouse ubiquitin-specific protease Unp, and its human homolog Unph, can efficiently cleave the ubiquitin-proline bond in ubiquitin fusion proteins of different sizes. N-terminal sequencing of the cleavage products reveals that cleavage occurs precisely at the ubiquitin-proline junction. The biological significance of this cleavage activity is unclear, as ubiquitin-proline fusions do not occur naturally. However, it may indicate a different catalytic mechanism for these ubiquitin-specific proteases and/or that they can cleave ubiquitin-like proteins. Unp and Unph thus represent versatile ubiquitin-specific proteases for cleaving ubiquitin-fusion proteins in biotechnology and basic research, regardless of both the amino acid immediately following ubiquitin, and the size of the fusion partner.

yeast Saccharomyces cerevisiae, a family of 16 Ubp enzymes can be identified in its completely sequenced genome, based on sequence similarity to conserved sequence domains observed in the first three yeast Ubp enzymes to be studied (5). Ubp activity has been demonstrated for 15 of these recombinant enzymes (4 -6, 8, 9). 2 In addition, proteins from other species that contain these conserved sequence motifs have also been demonstrated to have Ubp activity (6,10,11), indicating that the presence of such motifs strongly correlates with Ubp activity. Yeast also contains an additional ubiquitin-cleaving enzyme termed Yuh1p (12), which is unrelated to the Ubp family by sequence, and which has several homologs in mammals (13,14). The presence of such a large family of Ubp enzymes in yeast is not required merely to cleave ubiquitin precursors, and is indicative of other regulatory roles in the ubiquitin pathway.
The unusual ubiquitin fusion structure, and the ability of Ubps to cleave the ubiquitin moiety, have been exploited in several ways. The relative insensitivity of Ubp-mediated cleavage to the amino acid residue immediately following ubiquitin allows the synthesis of a protein with any residue at its N terminus. This technique was used to generate of a set of otherwise identical ␤-galactosidase (␤-gal) proteins that differed only at their N-terminal residue, and subsequently led to the discovery of the N-end rule: the relationship between the half-life of a protein and the identity of its N-terminal residue (15)(16)(17)(18). One limitation observed during these experiments was that the ubiquitin-proline bond was cleaved very inefficiently, either in yeast or in mammalian cells (rabbit reticulocyte extract), presumably due to the unusual structure of proline. The rate of cleavage was estimated to be 20 times slower than all other ubiquitin-␤-gal fusions (17). Furthermore, the presence of the uncleaved ubiquitin moiety in a ubiquitin-proline-␤-gal (Ub-P-␤-gal) fusion in yeast, or an "uncleavable" Ub-␤-gal fusion engineered by mutation of the Ubp cleavage site, actually targeted the fusion protein for degradation via ubiquitin-dependent proteolysis (19,20). Thus uncleaved ubiquitin fusions are recognized as abnormal proteins and rapidly degraded.
The second feature of ubiquitin fusions is that they enhance (often dramatically) the yield of the protein to which ubiquitin is fused. This phenomenon was initially observed in yeast, where the natural ubiquitin-ribosomal protein fusion structure was found to enhance the yield of the ribosomal protein (21). Ubiquitin fusions have since been used to improve the yield of recombinant proteins in bacteria, yeast, and insect cells (reviewed in Ref. 22). That the yield enhancement occurs in bacteria, which lack ubiquitin and Ubp enzymes, indicates that this property is intrinsic to ubiquitin, and has been proposed to be due to a covalent chaperone effect of ubiquitin in allowing the fusion partner to fold and/or protect it from degradation * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) L00681 and U20657.
¶ Supported by a QEII Fellowship from the Australian Research Council and a grant from the Australian Capital Territory Cancer Society. To whom correspondence should be addressed. Tel.: 61 2 6249 3824; Fax: 61 2 6249 4712; E-mail: Rohan.Baker@anu.edu.au. 1 The abbreviations used are: Ub, ubiquitin; Ubp, ubiquitin-specific protease, deubiquitinating enzyme; Unp, mouse ubiquitin-specific protease; Unph, human homolog of Unp, GST, glutathione S-transferase; ␤-gal, ␤-galactosidase; Ub-X-␤-gal, ubiquitin-X-␤-gal fusion, where X is one of the 20 amino acids; PCR, polymerase chain reaction; PAGE, polyacrylamide gel electrophoresis. 2 R. T. Baker, unpublished data. (21). The yield enhancement also occurs when a Ubp is coexpressed in bacteria with the fusion protein, which is rapidly cleaved co-translationally, indicating that the protective effect of ubiquitin is exerted in the small period of time that the fusion is intact (23). We have recently been studying a mouse protein termed Unp, that is an oncoprotein and causes tumors when overexpressed in nude mice (24,25). Unp, and also its human homolog Unph, both contain the conserved Ubp domains identified in the yeast Ubp family (5), and thus are potential Ubps (6,26). During the characterization of the Ubp activity of Unp and Unph expressed in yeast and bacteria, we noted that it exhibited significant cleavage activity of the ubiquitin-proline bond. In this report, we characterize this novel cleavage activity.

EXPERIMENTAL PROCEDURES
Plasmids Expressing Ubps, Unp, and Unph-Escherichia coli was cultured and transformed with plasmids according to standard procedures (27). Clones containing Unp and Unph cDNAs have been described previously (24,26). The 5Ј portion of the Unp open reading frame from the initiation codon to the internal XbaI site was amplified by the polymerase chain reaction (PCR) using oligonucleotides UNP-DAG (5Ј-d(CTCGAATTCGAGATGGCGGAAGG)) and UNPXBA (5Јd(GGCTTCTTCTTTACGCGGTTCAG)). The product was digested with EcoRI and XbaI, ligated into pBluescript, and the insert of one clone was sequenced (ThermoSequenase kit, Amersham Corp.) in full to reveal no PCR errors. This fragment and the remainder of the Unp clone 3Ј of the XbaI site were ligated into the bacterial expression plasmid pKK261 (a derivative of pKK223-3) (23) to reconstruct the Unp open reading frame under control of the tac promoter (28), yielding plasmid pCG53. An analogous procedure was used to amplify and reconstruct Unph from a human Unph cDNA clone. Plasmids pCG43 and pCG42 were also initially constructed by analogous procedures, that expressed truncated Unp and Unph proteins, lacking residues 1-168 (Unp⌬1-168) and 1-55 (Unph⌬1-55), respectively. These truncated proteins were derived from deduced incomplete open reading frames due to sequencing errors in the original Unp and Unph sequencing reports (24,26). Corrected sequences and full details will appear elsewhere 3 and in accession nos. L00681 (Unp) and U20657 (Unph). Plasmids expressing Ubp1p (pJT70), Ubp2p (pRB105, pRB173), Ubp3p (pRB143, pRB175), and Yuh1p (pKKYUH1) have been described previously (4,5). A plasmid for the expression of Unp in yeast was constructed by ligation of the EcoRI-HindIII fragment from pCG53 downstream of the constitutive ADH1 promoter (29) in the high copy vector YEplac181 (30).
Plasmids Expressing Ubiquitin-Proline Fusion Proteins-Plasmid pUb23-P expresses a Ub-P-␤-gal fusion protein from a hybrid CYC-GAL10 promoter on a yeast high copy plasmid (15). We have previously described a pKK223-3-based plasmid, pRB269, containing a single ubiquitin-coding region that contains a SacII site at the 3Ј end of ubiquitin to enable ligation of other gene fragments (5). A ubiquitinproline-GSTP1 fusion (Ub-P-GSTP1) was constructed in pRB269 analogously to the ubiquitin-methionine-GSTP1 fusion described previously (5). The wild-type mature N terminus of human GSTP1 (Pro-Pro-Tyr) was mutated to Pro-Ala-Tyr, and a SacII restriction site added, using the forward primer UBPGST (5Ј-d(GTGCCGCGGTGGTCCGGCGTAT-ACCGTGG)), the reverse primer 3EX2 (5Ј-d(CTTAAGCTTCCCTCACT-GTTTCCCG) (5), a GSTP1 cDNA clone (5), and the PCR. The PCR product was digested with SacII and HindIII, ligated into pBluescript, sequenced in full to reveal no PCR errors, and ligated into pRB269 digested with SacII and HindIII to yield pRB481. For coexpression of Ub-P-GSTP1 in E. coli with a Ubp-expressing plasmid, the tac promoter/Ub-P-GSTP1-rrnB terminator region of pRB481 was excised with EagI and ScaI, and ligated into pACYC184 digested with EagI and EcoRV to yield pRB486.
Oligonucleotide-directed Mutagenesis-The cysteine residue of Unp that lies within the conserved Ubp Cys domain (C311) was mutated to alanine using the oligonucleotide UNPCA (5Ј-d(CAGAGTTCATAAAG-GCAGTGTTCCCC), an M13 subclone of the Unp 5Ј EcoRI-XbaI fragment, and a dut ung mutagenesis kit (Bio-Rad). This oligonucleotide also destroys an RcaI (BspHI) site in this region. Mutagenized clones were screened by RcaI digestion, and then sequenced to confirm the correct mutation. The EcoRI-XbaI fragment containing the C311A mutation was used to replace the corresponding wild-type fragment in pCG53 to yield pCG54, which expresses UnpC311A.
Ubiquitin Cleavage Assays-Cleavage of Ub-P-␤-gal fusions in yeast was assayed by pulse-labeling with [ 35 S]methionine, extraction, immunoprecipitation with a monoclonal antibody to ␤-gal (Promega), electrophoresis in a 6% SDS-polyacrylamide gel (31), and fluorography exactly as described elsewhere (32). For in vitro cleavage assays, Ub-P-␤-gal was immunoprecipitated from yeast cells containing plasmid pUB23-P, the protein A-agarose pellet washed with TN buffer (50 mM Tris, pH 7.4, 150 mM NaCl) to remove SDS, and portions of the resuspended pellet slurry were mixed with extracts of E. coli expressing different Ubps. Where required, ␤-gal protein-containing bands were excised from dried gels, and the amount of 35 S in the bands was determined. Bacterial extracts were prepared from exponentially growing cultures that had been induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. Cells from 1.5-ml cultures were collected, resuspended in 0.2 ml of TN buffer containing 1 mM dithiothreitol and 0.25 mg lysozyme/ml, incubated on ice 10 min, sonicated for two 5-s bursts, and centrifuged for 5 min at 4°C. The supernatant was collected as the soluble extract. Assays consisted of mixing an extract containing a Ubp (usually 15 l) with an extract containing the substrate (5 l) and incubating at 37°C for 60 min. An antiserum against Ub-M-GSTP1 was raised in rabbits using purified Ub-M-GSTP1 (23). This antisera was used at 1:1000 dilution against immunoblots of cleavage reactions involving Ub-P-GSTP1 resolved in a 12.5% SDS-polyacrylamide gel. Where required, GSTP1-containing proteins were purified by affinity chromatography on GSH-agarose (Sigma) exactly as described previously (23). Protein sequencing was done at the Biomolecular Resource Facility, Australian National University.

RESULTS
Unp and Unph Are Ubps-Unp and Unph are mouse and human proteins, respectively, that contain conserved sequence motifs originally identified in the yeast ubiquitin-specific proteases Ubp1p, Ubp2p, and Ubp3p (5,6,24,26). These domains contain a conserved cysteine residue, and two histidine residues, respectively, that presumably form part of the active site of these thiol proteases (5) (Fig. 1). Mutation of the corresponding cysteine and/or histidine residues in both the yeast Doa4p(Ubp4p), Drosophila fat facets Ubp, and yeast Ubp2p abolish their Ubp activity (6, 10). 2 To confirm that the presence of these conserved domains in Unp and Unph correlates with Ubp activity, we expressed these enzymes in E. coli and assayed crude extracts for Ubp activity against an artificial linear ubiquitin-GST fusion protein that has a methionine residue immediately following ubiquitin (UbMGSTP1) (23). While E. coli extracts alone exhibit no cleavage of the Ub-M-GSTP1 fusion protein, extracts containing either Unp or Unph cleaved a substantial proportion of this ubiquitin-fusion protein, demonstrating that they are bona fide Ubps (data not shown). Furthermore, mutation of the conserved cysteine residue in Unp to alanine (UnpC311A) abolished this cleavage activity (data not shown; see below), consistent with a role for this residue in the active site of the enzyme. A full characterization of the Ubp activity of Unp and Unph will be presented elsewhere. 4 Unp Efficiently Cleaves the Ubiquitin-Proline Bond in a Ubiquitin-␤-Galactosidase Fusion Both in in Vivo and in Vitro-Ubiquitin-X-␤-galactosidase (Ub-X-␤-gal) fusion proteins have been routinely used as model substrates of ubiquitindependent proteolysis in yeast (e.g. see Refs. 15,19,20,and 32).
During initial experiments where Unp was coexpressed in yeast with Ub-P-␤-gal to ascertain any effect of Unp on ubiquitindependent proteolysis, we noticed that substantially more cleavage of the ubiquitin-proline bond occurred in cells containing Unp than those without. This effect was further quantitated in a pulse-chase assay ( Fig. 2A). In the presence of Unp, substantially more of the P-␤-gal species is present at the zero time point than in control cells (compare lanes 1 and 4). The P-␤-gal thus produced is a stable protein, in agreement with the N-end rule (15), and is not degraded over the 30-min chase. Unp does not have an isopeptidase activity in these experiments, in that the multiubiquitin chain assembled on Ub-P-␤gal (33) is not affected by the presence of Unp ( Fig. 2A).
One explanation for the apparent enhanced cleavage at the ubiquitin-proline bond is that Unp may have a relaxed or degenerate cleavage specificity, and may be cleaving at an incorrect site other than the predicted glycine-proline junction. To test this, we used a Ub-X-␤-gal fusion protein, Ub V76 -V-␤gal, where the Gly 76 residue of ubiquitin had been mutated to valine, and the X residue was also valine. This mutation prevents cleavage by all known yeast Ubp enzymes, because when this fusion is expressed in yeast, no ubiquitin cleavage is observed, and the intact fusion is targeted for ubiquitin-dependent degradation via a pathway that recognizes noncleavable ubiquitin fusions (19,20). Yuh1p, a different yeast ubiquitinspecific protease, is also unable to cleave the ubiquitin-proline bond in Ub-P-␤-gal fusions (12). The presence of Unp did not result in any cleavage of this G76V mutant ( Fig. 2A), indicating that Unp also requires the Gly 76 residue for cleavage, and does not appear to have a relaxed cleavage site specificity.
To test whether Unp was capable of cleaving Ub-P-␤-gal only when co-expressed with it in vivo, we performed an in vitro cleavage assay, whereby Ub-P-␤-gal was immunoprecipitated from yeast cells grown with [ 35 S]methionine and then incubated with extracts of E. coli expressing different Ubps. As shown in Fig. 2B, an E. coli extract alone had no effect on Ub-P-␤-gal, whereas extracts containing Unp resulted in 63% of the Ub-P-␤-gal-containing species (including the multiubiquitinated species) being converted to P-␤-gal (lane 5). In wildtype yeast, only 9% of the 35 S in ␤-gal-containing species was present as P-␤-gal (lane 1). Given that Unp does not have an isopeptidase activity (see above), it appears to be able to cleave the ubiquitin-proline bond even when a large multiubiquitin chain is attached to the ubiquitin moiety. The yeast Ubps Ubp1p (lane 3), Ubp2p (lane 4), or Ubp3p (not shown) were not able to cleave the ubiquitin-proline bond in this assay, consistent with the poor cleavage of Ub-P-␤-gal in yeast, and indicative of the unique cleavage specificity of Unp. (The Ubp1p and Ubp2p extracts used here were able to cleave a Ub-M-GSTP1 fusion protein in vitro (data not shown) (5). Ubp3p does not appear to have Ubp activity in vitro (5).) Notably, this assay revealed that Ubp1p has an apparent isopeptidase activity, in that it removed the bulk of the high molecular weight ubiquitin conjugates on Ub-P-␤-gal (lane 3). This activity is being further investigated.
Unp and Unph Precisely Cleave at the Ubiquitin-Proline Junction in Vivo and in Vitro-We have previously used the ubiquitin fusion technique to express human glutathione Stransferase GSTP1 (23). While the mature N terminus of GSTP1 is proline, we were unable to use the ubiquitin fusion technique to expose proline at the N terminus of GSTP1 due to the poor efficiency of cleavage at this bond by known Ubps, and so retained the initiator methionine codon in the fusion (23). Given the ability of Unp to cleave this bond in Ub-P-␤-gal, we constructed a Ub-P-GSTP1 fusion and tested its cleavage with Unp and Unph (see "Experimental Procedures"). For the purposes of this experiment, we mutated the wild type N terminus of GSTP1 (Pro-Pro-Tyr) (34,35) to Pro-Ala-Tyr, so that the subsequent N-terminal sequencing to determine the cleavage site would not be ambiguous.
To assay in vivo cleavage of Ub-P-GSTP1, various Ubps were co-expressed in E. coli with Ub-P-GSTP1, cell extracts were resolved by SDS-PAGE, and an immunoblot was performed with a rabbit antiserum raised against purified Ub-M-GSTP1 (23). This antiserum detects uncleaved Ub-P-GSTP1, P-GSTP1, and (weakly) free ubiquitin. As shown in Fig. 3A, approximately 40% of the Ub-P-GSTP1 fusion was cleaved when co-expressed with full-length Unp (lane 8), while the truncated Unp (Unp⌬1-168) and Unph (Unph⌬1-55) enzymes also exhibited substantial cleavage (lanes 5 and 6). To assay in vitro cleavage of Ub-P-GSTP1, extracts of E. coli expressing Ub-P-GSTP1 were mixed with extracts of E. coli expressing different Ubps, and following incubation at 37°C for 1 h, electrophoresis and immunodetection were performed as above. Similar results to the in vivo cleavage assays were observed (Fig. 3B), with full-length Unp cleaving approximately 40% of Ub-P-GSTP1 (lane 4), while the truncated Unp and Unph enzymes exhibited partial activity (lanes 6 and 7). To determine the fidelity of Unp and Unph-mediated cleavage of the ubiquitin-proline bond, the GSTP1-containing proteins arising from scaled-up cleavage reactions of Unp⌬1-168 cleavage of Ub-P-GSTP1 in vivo (Fig. 3A, lane 5), Unph⌬1-55 cleavage of Ub-P-GSTP1 in vivo (Fig. 3A, lane 6), and fulllength Unp cleavage of Ub-P-GSTP1 in vitro (Fig. 3B, lane 4) were purified by glutathione affinity chromatography on GSHagarose, eluted with 5 mM GSH, and resolved by SDS-PAGE. A representative purification is shown in Fig. 4 (full-length Unp cleavage of Ub-P-GSTP1 in vitro). The Coomassie-stained P-GSTP1-sized bands were excised from the gel (Fig. 4, lanes 5  and 6), eluted onto a polyvinylidene difluoride membrane, and  Table I. In all three cases, proline was the only residue detected at the N terminus, and the sequence obtained matched the expected sequence exactly. Neither glycine nor alanine was observed in the first cycle, which would be expected if cleavage had occurred either one residue N-terminal or C-terminal, respectively, of the glycine-proline bond. From this analysis, we conclude that at least 95% of the cleavage had occurred at the glycine-proline bond. Thus in contrast to other Ubps, Unp and its human homolog Unph exhibit very efficient and precise cleavage of the ubiquitin-proline bond in ubiquitin fusion proteins.

DISCUSSION
The ubiquitin-fusion technique has proven to be a versatile method of increasing the yield of recombinant protein expression, and as a way of synthesizing proteins with any desired N terminus, both for biotechnological and research applications (reviewed in Ref. 22). One limitation in this technique has been the extremely inefficient cleavage of the ubiquitin-proline bond, and thus proteins with an N-terminal proline residue have been excluded. For example, the status of N-terminal proline in the E. coli N-end rule has not been determined due to the inability of Ubp1p to deubiquitinate Ub-P-␤-gal (36). Our demonstration that the mouse Ubp enzyme Unp, and its human homolog Unph, can efficiently and precisely cleave the ubiquitin-proline bond either in vitro or in vivo, allows circumvention of this problem, and improves the versatility of the ubiquitin fusion technique. Unp and Unph exhibit this activity irrespective of the size of the fusion partner (1045 residues in Ub-P-␤-gal, 209 residues in Ub-P-GSTP1), and do not appear to be affected by the residue immediately following proline in the P-2Ј position (His in Ub-P-␤-gal, Ala in Ub-P-GSTP1). This activity is unique to Unp and Unph among known Ubps; cleavage of Ub-P-␤-gal by Ubps present in yeast or in rabbit reticulocyte extract is very inefficient (15,17), and recombinant yeast Ubps Ubp1p, Ubp2p, Ubp3p, and Yuh1p lack this activity (see "Results"). In addition, recombinant yeast Ubps Doa4p/Ubp4p, Ubp6p, Ubp12p, and Ubp15p cannot cleave the ubiquitin-proline bond (data not shown). Ubp12p is the most closely related yeast Ubp to Unp and Unph, and its inability to cleave this bond implies that yeast lack an efficient functional homolog of Unp/Unph with respect to ubiquitin-proline bond cleavage. While proline is only one of the 20 amino acids that may occur at the N terminus of a protein or peptide, several important proteins do have an N-terminal proline residue, including GSTP1, which is strongly induced in early stages of hepatocarcinogenesis in rats (reviewed in Hayes and Pulford (37)). Several biologically active peptides also have an N-terminal proline, such as angiogenesis inhibitor platelet factor 4 (38), basic fibroblast growth factor (39,40), neuropeptide F (41), and neuropeptide Y derivatives (42). While organic synthesis is efficient for the production of short peptides in vitro, the ubiquitin fusion technique provides an efficient alternative for the synthesis of peptides of any length in vivo (43), and the ability of Unp and Unph to cleave the ubiquitin-proline bond enhances this technique.
Although the ability of Unp and Unph to cleave the ubiquitin-proline bond is a versatile research and biotechnology tool, the biological significance of this cleavage activity is unclear, as no natural ubiquitin fusions containing the ubiquitinproline bond have been observed. It is likely that this cleavage activity reflects a unique property of Unp and Unph, in that they can access and cleave the C terminus of ubiquitin in fusions that other Ubps cannot. Presumably this reflects the structure of the natural substrate(s) of Unp/Unph, which may require a special catalytic mechanism of the Ubp that cleaves them. The unique structure of proline, with the side chain cyclized onto the backbone nitrogen atom, may sterically shield the bond to be cleaved from all Ubps other than Unp/Unph. It is also possible that Unp may cleave some or all of the ubiquitin-like proteins that are processed from linear precursors and/or post-translationally formed conjugates, such as the ubiquitin-like protein fused to ribosomal protein S30 (44) that is apparently conjugated to an immune suppressor factor (45), and the small ubiquitin-like modification (SUMO) of the Ran-GTPase-activating protein (46,47). An ability to cleave ubiquitin-like proteins may signify a "relaxed" active site conformation, which may be able to tolerate the proline residue, whereas other Ubps cannot. In this respect, we note that both Unp and Unph have a glycine positioned six residues on the N-terminal side of the active site cysteine within the conserved Cys domain, whereas no other known Ubp from any species has a glycine in this position, having instead much bulkier residues (Fig. 1B) (8) (data not shown). Although we have no information on the structure that this domain adopts, the small glycine residue may allow toleration of the proline residue in the P-1Ј position. One testable prediction of this model is that mutation of this Unp/Unph glycine to a non-glycine residue should greatly reduce its ability to cleave the ubiquitin-proline bond.
It is also noteworthy that in any ubiquitin-fusion where the P-1Ј residue is not proline, be it a linear fusion or an isopeptide linkage, there is a proton on the nitrogen atom of the P-1Ј residue, whereas when the P-1Ј residue is proline, this proton is absent. It is possible that this proton is required for the catalytic mechanism of most Ubps, which cleave very inefficiently if this proton is absent, and that Unp and Unph have an alternate mechanism and/or can utilize a different proton, and thus can cleave fusions where proline is in the P-1Ј position. In the model thiol protease papain, it has been proposed that the NH of the P-1Ј leaving group is involved in hydrogen bonding with a carbonyl group of the papain peptide backbone, to aid in distortion of the scissile C-N bond to increase its electrophilicity and thus facilitate attack by the thiol group (48). The nitrogen of the P-1Ј leaving group also has a role in the breakdown of the tetrahedral intermediate linking papain to its substrate, becoming protonated in the transition state (49). If the Ubps operate by an analogous mechanism, then the unusual structure of proline in the P-1Ј position may not allow either of these roles to operate. In either case, it would appear that Unp and Unph have a different catalytic mechanism than other Ubps. The exploitation of this mechanism to cleave the ubiquitin-proline bond should prove fruitful for both biotechnology and research applications. Pro-Ala-Tyr-Thr-Val-Val * Cleavage occurred either in vivo (co-expression of Ub-P-GSTP1 and Ubp in E. coli) or in vitro (mixing of an extract of E. coli expressing Ub-P-GSTP1 and an extract of E. coli expressing Unp).