Characterization and Chromosomal Localization of USP3, a Novel Human Ubiquitin-specific Protease*

Conjugation to the small eukaryotic protein ubiquitin can functionally modify or target proteins for degradation by the proteasome. Removal of the ubiquitin modification, or deubiquitination, is performed by ubiquitin-specific proteases and is an important mechanism regulating this pathway. Here we describe a novel human ubiquitin-specific protease, USP3, initially identified as a partial cDNA clone similar to one of two highly conserved sequence regions common to all ubiquitin-specific proteases. We have isolated a complete USP3 cDNA clone containing both of these conserved sequence regions. The USP3 gene appears to be single copy and maps to human chromosome 15q22.3. A USP3probe detects two mRNA transcripts, one of which corresponds in length to the cDNA. Both are expressed at low levels in all tissues examined, with highest expression in pancreas. The USP3 protein is a functional ubiquitin-specific protease in vitro, and is able to inhibit ubiquitin-dependent degradation of both an N-end Rule substrate and abnormal endogenous proteins in yeast. USP3 is also only the second known ubiquitin-specific protease capable of efficiently cleaving a ubiquitin-proline bond.

uitin as part of a precursor fusion protein, releasing free ubiquitin moieties, or cleave bonds conjugating ubiquitin (posttranslationally) to proteins. Based on sequence similarity to conserved amino acid regions observed in the first Ubps studied, numerous Ubps have been, and continue to be, identified in many species, with 16 known to exist in yeast. Because studies have shown that the presence of these motifs strongly correlates with Ubp activity, not all these Ubps have actually been functionally characterized as deubiquitinating enzymes. In humans, the ability to cleave ubiquitin bonds has been illustrated for only four of seven published Ubps; Unph, Tre-2, Hausp, and isopeptidase T (4 -13). Of the four characterized Ubps, each has its own characteristic substrate specificity and role in ubiquitindependent proteolysis, and it follows that novel deubiquitinating enzymes may also play a distinct role in regulating the ubiquitin proteolytic pathway.
Fundamental to their regulatory role in ubiquitin-dependent proteolysis is the ubiquitin cleavage activity exhibited by the Ubps. Moreover, the cleavage specificity of individual Ubps dictates at which point(s) they contribute to this regulation of ubiquitin-dependent proteolysis. This is reflected by the detrimental effects caused by overexpression of, or mutations to, the Ubps, which have been best studied in yeast. Triple null mutants of the yeast Ubp1p, Ubp2p, and Ubp3p are extremely sensitive to stress conditions, presumably because of a reduction in the rate of ubiquitin-dependent proteolysis (14). The yeast Doa4p (Ubp4p) cleaves peptide or isopeptide bonds linking multi-ubiquitin chains to peptide remnants following proteasomal degradation of the ubiquitin-targeted protein (9). The doa4 null mutant causes general inhibition of ubiquitin proteolysis and phenotype defects ranging from slow growth to defects in DNA repair. Overexpression of Doa4p increases the degradation rate of several substrates, indicating that Doa4p is one rate-limiting factor in their degradation by the ubiquitindependent proteolysis. When yeast Ubp14p (which disassembles unanchored polyubiquitin chains) is deleted, ubiquitin-dependent proteolysis is generally inhibited because of the accumulated free chains inhibiting the proteasome (15). However, overexpression of Ubp14p can inhibit the degradation of some proteins although not affecting others, possibly because of its effects on preassembled ubiquitin chains being attached to some substrates (15). Thus, the characteristic substrate specificity of a Ubp greatly influences its role in ubiquitin-dependent proteolysis.
In this study, we present the isolation and characterization of a novel human ubiquitin-specific protease. We have adopted the nomenclature suggested by the Human Genome Project nomenclature committee and have termed this enzyme USP3. It is demonstrated that USP3 is a functional Ubp, capable of cleaving the artificial ubiquitin-glutathione S-transferase (GST) fusion protein in vitro. We have localized the USP3 gene to chromosome 15q22.3, distinct from any known human Ubp. USP3 expression in yeast has a general inhibitory effect on ubiquitin-dependent proteolysis, as assayed with the degradation of L-␤-gal, a specific introduced N-end rule substrate of the ubiquitin-dependent proteolytic pathway, and on the degradation of abnormal endogenous proteins. In addition, USP3 is able to efficiently cleave the ubiquitin-proline bond. To date this cleavage activity has only been attributed to the mouse Ubp, Unp, and its human homolog, Unph (5).

EXPERIMENTAL PROCEDURES
Isolation of the Full-length USP3-The cDNA clone Hsaaadqei (EBI accession number Z21167) was obtained from the UK-Human Genome Mapping Project as a NM1149 phage clone. The ϳ1.7-kb EcoRI insert was subcloned into the vector pRS316 (16) and, following restriction endonuclease mapping, completely sequenced (Sequenase version 2.0, U. S. Biochemical Corp.) to reveal an artifact of two fused cDNA clones (see "Results"). The Ubp-like portion was isolated as a SacI/EcoRI fragment, termed USP3, and used for subsequent experiments. The partial cDNA clone was used as a random primed probe (Megaprime; Amersham Pharmacia Biotech) for screening a gt11 phage human testis tissue cDNA library (CLONTECH) (17). After selecting positive clones derived from tertiary screening, phage DNA was isolated (Promega Wizard Preps purification system) and the inserts amplified using -specific primers (New England Biolabs, catalogue numbers 1218F and 1222R). The polymerase chain reaction products were digested with EcoRI and ligated into pBluescript and sequenced. Computer sequence data bases were searched using the BLAST algorithm (18); this identified two expressed sequence tag (EST) cDNAs, which confirmed the fragment order and sequence of the majority of the 2,324-bp sequence. The open reading frame was amplified by the polymerase chain reaction using specific oligonucleotide primers (USP3BamHI, 5Ј-d(TCGGATCCATGGAGTGTCCACACCTGAG); USP3HindIII, 5Ј-d(ATAAAGCTTGCAGCCTTGAGAGACAAGC)) and a proofreading DNA polymerase to reduce errors (Pfu-Turbo, Stratagene). The product was digested with BamHI/HindIII and ligated into protein expression vectors as required (see below).
Southern and Northern Hybridization Analyses-The probe used for Southern hybridization was a 760-bp ScaI-EcoRI fragment from the 3Ј end of the cDNA, whereas the probe for Northern hybridization was a 488-bp BspHI-XbaI fragment, extending from bp 605 to bp 1093 of the USP3 sequence. Probes were radioactively labeled with [␣-32 P]dATP using random priming (Megaprime; Amersham Pharmacia Biotech). Genomic DNA was isolated from human blood samples (19) and 10 g digested overnight at 37°C with BamHI, EcoRI, PstI, or XbaI. The digests were electrophoresed on 1ϫ TAE, 1% agarose gel, and capillary blotted (20) onto Hybond N ϩ (Amersham Pharmacia Biotech), with hybridization carried out as specified by the membrane manufacturer. Northern hybridization used a CLONTECH Premade Multiple Human Tissue Northern blot (catalog no. 7760), containing ϳ2 g of mRNA per lane, according to the manufacturer's instructions.
Chromosomal Localization-Radiation hybrid mapping data was searched at the National Center for Biotechnology Information. The probe used for localization of the USP3 gene by fluorescence in situ hybridization was the IMAGE cDNA clone 45276 obtained from Genome Systems (accession number H08387) in the vector Lafmid BA. The probe was nick-translated with biotin-14-dATP and hybridized in situ at a final concentration of 5 ng/l to metaphases from two normal males. The fluorescence in situ hybridization method was modified from that previously described (21). Chromosomes were stained before analysis with both propidium iodide (as counterstain) and 4,6-diamidino-2phenylindole for chromosome identification). Images of metaphase preparations were captured by a cooled CCD camera using the CytoVision Ultra image collection and enhancement system (Applied Imaging Int Ltd.).
Expression of USP3-Saccharomyces cerevisiae BWG1-7A and BBY45 strains were cultured and transformed as described elsewhere (22). To construct a GST-USP3 fusion expression plasmid, the complete USP3 open reading frame was cloned as a BamHI/HindIII fragment into the yeast pRS316-based (16), single copy shuttle vector pRD56 (gift of Dr. Danesh Moazed) to form pKT7. pKT7 expresses GST-USP3 from the GAL10 promoter. A second plasmid for the expression of GST-USP3 in yeast, pKT8, was constructed by transferring the pGAL10/GST-USP3 insert from pKT7 into the high copy vector YEplac195 (23). The pRD56 plasmid, which expresses GST alone, was used as a negative control.
In Vitro Ubiquitin Cleavage Assay-Preparation of the [ 35 S]-labeled Ub-GST fusion substrate required for in vitro ubiquitin cleavage assays has been described previously (24). An extract from S. cerevisiae BWG1-7A cells expressing either GST-USP3 (pKT8) or GST alone was prepared using a large scale version of an established method (17), except cell lysis was achieved using a RF-I RIBI Cell Fractionator (Sorvall). A portion of crude extract was retained as a positive control. The remaining extract was incubated with goat anti-GST antibody (Amersham Pharmacia Biotech) for 2 h at 4°C, then 20 l of 50% slurry of protein G-Sepharose (Amersham Pharmacia Biotech) added and incubated for a further 30 min. Following extensive washing of the pellet with TN buffer (50 mM Tris, 150 mM NaCl, pH 7.4), 15 l of TN buffer was added to the Sepharose beads, to which the GST-USP3 protein was still attached. This Sepharose/protein slurry and the crude extract as the positive control, were both used for in vitro ubiquitin cleavage assays. This involved adding 5 l of [ 35 S]-labeled Ub-GST fusion substrate, incubation at 37°C for 1 h, electrophoresis on denaturing 12% SDS-polyacrylamide gel (25), and fluorography as described elsewhere (26).
Canavanine Sensitivity Assay-BBY45 yeast cultures transformed with the plasmids expressing GST or GST-USP3 (see above) were grown for 48 h in galactosidase ϩ Gro-medium at 30°C as described previously (30), subcultured into SD-Ura/Gal, and grown overnight at 30°C until they reached A 600 ϭ 0.5-1.0. The yeast cultures were serially diluted and plated in triplicate on selective media (SD-uracil, -arginine; Ref. 17) containing 0, 0.5, 1, or 1.5 g/ml canavanine. Sensitivity to canavanine was determined by expressing the number of colonies present at each concentration of canavanine as a percentage of those present at 0 g/ml canavanine (control).

RESULTS
Completion of the Hsaaadqei/USP3 cDNA Clone Sequence-We identified an EST cDNA clone arising from the UK Human Genome Mapping Project (accession number Z21167, clone name Hsaaadqei) by similarity to the conserved Ubp Cys-box originally identified in the yeast Ubp1p, Ubp2p, and Ubp3p enzymes (14). Sequencing of the Hsaaadqei cDNA clone, followed by sequence comparison to computer data bases revealed that this clone was a fusion of two clones, a partial Ubp-like clone containing the conserved Cys-box, and a second cDNA clone derived from the human thyroid receptor-interacting protein gene Trip1 (HUMTRIP1). These were joined together by their 5Ј ends, producing the ϳ1.7-kb Hsaaadqei cDNA clone presumably as an artifact of cDNA library construction. Using the partial Ubp-like portion (ϳ770-bp, renamed USP3) as a probe for screening a human testis library, several positive clones, including C2 and C3 (clone 2 and clone 3) were obtained. The products of EcoRI digests of these clones were sequenced. Clone C2 produced two EcoRI fragments, C2.1 and C2.2 (Fig. 1A). Clone C3 produced three EcoRI fragments, C3.1, C3.2, and C3.3 (Fig. 1A). The sequence of the C2.1 and C3.2 fragments confirmed the 770-bp USP3 sequence, including a 64-bp extension 5Ј to the SacI site used to obtain the partial 770-bp SacI/EcoRI USP3 cDNA clone from Hsaaadqei. The sequences from C2.2 and C3.1 were not similar to USP3 as they presumably represented adjoining EcoRI fragments. The necessary overlaps were found through subsequent detection of the EST clone H08387 by data base searches with the 770-bp USP3 sequence. The known sequence of the USP3 cDNA was thus extended to 2.32-kb, containing both the Cys-and the His-conserved sequence boxes common to all Ubps, an open reading frame (ORF) extending from a start codon at base pair 100 to a stop codon at bp 1,666, and a polyadenylation signal and poly(A) tail downstream of this ORF (Fig. 1B). The 3Ј end of the sequence contains a well defined mammalian polyadenylation signal AATAAA (bp 2,263-2,268), followed by a diffuse GU-rich (GT-rich in cDNA) sequence spanning approximately 28 bp until the actual site of polyadenylation (bp 2,297) (31)(32)(33). At the 5Ј end of the sequence, located 5Ј to the start codon, is a GC-rich region, consistent with a 5Ј-untranslated region. The flanking nucleotide sequence around this initiation codon, containing a G at Ϫ3 and ϩ4 (Kozak consensus sequence; Ref. 34) is consistent with the ATG at bp 100 being the actual initiation codon. However, as there are no in-frame stop codons upstream of the start codon at bp 100, we cannot rule out a further upstream initiation codon. The present ORF would produce a protein of approximately 59 kDa, and the cDNA is consistent in length with subsequent Northern blot hybridization analysis (see below). Alignment of multiple USP3 ESTs detected through data base searches with the USP3 sequence in Fig. 1 confirmed the latter was the consensus. While this manuscript was under review, several EST clones appeared in the data base that confirm the 5Ј end of the cDNA, and two ESTs suggest that the sequence in Fig. 1B may be preceded by GGCCGAGCGC (accession numbers AI525838, AA371059, AA356126, AA361954).
Alternative Splicing in USP3-One EST identified through data base searches with the USP3 sequence contained an additional insert of 138 bp. This clone, GenBank TM accession number H93896, deviates from the USP3 sequence at bp 383, for 138 bp, then matches USP3 sequence identically from bp 384 (Fig. 1A). This raises the possibility that alternative splicing of the USP3 gene occurs. The only other EST clone that spans this region of USP3 (accession number AI525838) does not contain this additional sequence, and the 138-bp sequence was not found to be similar to any other EST clone apart from itself. As this extra sequence introduces stop codons in all three frames, it is most likely a splicing error rather than a normally alternatively spliced exon, as it would produce a severely truncated protein.
Sequence comparisons between the amino acid sequence deduced from the ORF of the complete USP3 cDNA clone and other known Ubps did not identify any strong sequence iden- tities or similarities aside from a number of conserved regions, which include the Cys-, His-, Asp-and KRF-boxes (Figs. 1A and 2). Among the 16 yeast Ubps, USP3 is most similar to Ubp8p (27% identity/40% similarity over the whole 471 residues of the latter; Fig. 2). The function of Ubp8p is not known (2).
Southern and Northern Hybridization Analysis-Southern blot analysis using a fragment spanning bp 1564 to 2324 of the USP3 cDNA and restriction-enzyme-digested human genomic DNA produced one or at most two hybridizing bands with each enzyme, consistent with a single copy gene ( Fig. 3; see also Fig.  1). The presence of two bands in the EcoRI digest is presumably due to at least one intron occurring within the region covered by the probe. If any closely related gene to USP3 was present, it must have an identical restriction map, at least with the enzymes used here.
Northern hybridization analysis of total RNA did not detect specific bands (data not shown). Hybridization of a multiple human tissue poly(A) ϩ RNA Northern blot (CLONTECH) with a USP3 probe produced a band representing a 2.45-kb transcript and a more weakly hybridizing band of 5.8 kb (Fig. 4A). The 5.8-kb transcript, and a third very weakly hybridizing 4.0-kb transcript, were more evident on longer exposures (not shown). The 2.45-kb mRNA corresponds to the 2.3-kb cDNA sequence, the difference accounted for by additional poly(A) tail and possibly additional 5Ј-untranslated region not present in the cDNA. The functional significance of the two major transcripts is not known, but the 5.8-kb transcript may be because of either the presence of a gene highly homologous to USP3, or an alternatively polyadenylated/spliced transcript of USP3 (see "Discussion").
Both the 2.45-and 5.8-kb USP3 transcripts were expressed in all the human tissues examined in approximately the same proportion to each other, including the heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas tissues. Abundance of both transcripts was reduced in brain tissue (relative to the ␤-actin control), whereas in both pancreas and, to a lesser extent, kidney tissue, increased expression is apparent.
Chromosomal Localization of USP3-An USP3 EST (Gen-Bank TM accession number H08388) had been mapped by radiation hybrid panel analysis to the chromosomal region 15q22-15q22.3, marker interval D15S159-D15S125, 58.8 -63.8cM. To confirm this assignment, we obtained the same clone from Genome Systems (IMAGE cDNA clone 45276), which spans bp 600 -2,324 of the USP3 cDNA (Fig. 1A), and used it as probe for fluorescence in situ hybridization. Twenty metaphases from a normal male were examined for fluorescent signal. All of these metaphases showed signal on one or both chromatids of chromosome 15 in the region 15q22-15q22.3; 70% of this signal was at 15q22.3 (data not shown), confirming the RH mapping result, and indicating that no closely related sequences were present elsewhere in the genome. There was a total of 12  4. Expression of USP3 in human tissues. A, a Northern blot of multiple human tissue poly(A) ϩ mRNA (CLONTECH) was probed with a radioactively labeled USP3 fragment (Fig. 1A). Sizes of RNA markers are shown on the left in kilobases, whereas tissues are listed above the lanes. The sizes of the hybridizing fragments are indicated on the right. B, hybridization with ␤-actin control, used to normalize USP3 tissue expression. Two forms of ␤-actin are present in the heart and skeletal muscle. nonspecific background dots observed in these 20 metaphases. A similar result was obtained from hybridization of the probe to 15 metaphases from a second normal male (data not shown).

USP3 Exhibits Ubiquitin Cleavage in Vitro-
The USP3 protein contains conserved sequence motifs found in all active Ubps (Fig. 2). These motifs contain a conserved cysteine residue and two histidine residues, respectively, that are thought to form part of the active site of these thiol proteases (14). To confirm that the presence of these conserved regions in USP3 correlates with Ubp activity, we initially attempted to express the USP3 protein in Escherichia coli. Unlike all previous Ubps expressed in E. coli, plasmids expressing USP3 could not be maintained in this bacteria, presumably because of toxicity (data not shown). USP3 was therefore expressed in Saccharomyces cerevisiae. Because S. cerevisiae has 16 Ubps, it was necessary to isolate the USP3 protein to assay its ubiquitin cleavage ability. USP3 was, therefore, expressed as a GST-USP3 fusion and isolated by immunoprecipitation using anti-GST antibody. (Attempts to purify GST-USP3 by glutathione affinity chromatography were unsuccessful (data not shown).) It should be noted that GST-USP3 was expressed in yeast at very low levels, barely detectable by Western immunoblot analysis with an anti-GST antibody (data not shown). The assay was conducted using the GST-USP3⅐antibody⅐protein G-Sepharose complex. The USP3 ubiquitin cleavage activity was assayed against an artificial linear 35 S-labeled ubiquitin GSTfusion protein (Fig. 5) (5, 35). Whereas the isolated GST protein alone (control) exhibited no cleavage of the Ub-GST fusion protein, an increase in cleavage of this fusion by the GST-USP3 protein was apparent, although at a very low level. This result is reinforced by evidence for USP3 cleavage of ubiquitin-␤galactosidase fusions in vivo (see below).

USP3 Inhibits Degradation of L-␤-Gal and Cleaves the Ubiquitin-Proline Bond in Vivo-The
Ub-X-␤-galactosidase (Ub-X-␤-gal) fusion proteins are commonly used as model substrates of ubiquitin-dependent proteolysis in yeast (e.g. see Refs. 26,27,36,37). We investigated the effect of USP3 expression on ubiquitin-dependent degradation of L-␤-gal, an N-end rule substrate, and Ub-P-␤-gal, a non-N-end rule substrate, in yeast. Yeast Ubps efficiently cleave Ub-L-␤-gal to L-␤-gal, an N-end rule substrate, whereas yeast Ubp activity against the ubiq-uitin-proline bond of Ub-P-␤-gal is 20-fold slower than for other amino acids (27,38,39). At present, only the mouse Ubp, Unp, and its human homolog Unph have been found to efficiently cleave the ubiquitin-proline bond (5). In yeast, L-␤-gal has a half-life of ϳ3 min, whereas the uncleaved Ub-P-␤-gal is degraded with a half-life of ϳ7 min (27). Measurement of steadystate ␤-gal activity showed that expression of GST-USP3 from the high copy plasmid (pKT8) increased the activity of L-␤-gal and Ub-P-␤-gal 2-3-fold and 16-fold, respectively (Fig. 6A). This suggested that USP3 might be inhibiting the ubiquitindependent degradation of both these proteins and/or, in the case of Ub-P-␤-gal, cleaving the Ub-proline bond to generate the stable P-␤-gal protein. The inhibitory effect of USP3 on the degradation of these proteins was examined further through pulse-chase assays (Fig. 6B). In the presence of GST-USP3, expressed from both low copy (data not shown) and high copy (Fig. 6B) plasmids, there was a clear increase in levels of L-␤-gal protein at later time points compared with the control. Quantification of the half-life over the 0 -10 min chase period revealed that the half-life increased from ϳ3 min in the control, to ϳ6.5 min in the high copy GST-USP3 expressing cells. The simplest explanation for this would be that USP3 possessed a "trimming" isopeptidase activity, and was shortening the multi-ubiquitin chain on L-␤-gal, thus reducing its efficiency as a proteasome targeting signal. Although no change is apparent from Fig. 7B, a small difference in high molecular weight conjugates may account for the small change in half-life observed.  1-6) or Ub-P-␤-gal (lanes 7-12), and GST protein (lanes 1-3, 7-9), or GST-USP3 from a high copy plasmid pKT8 (lanes 4 -6, 10 -12) were pulse-labeled with [ 35 S]methionine for 5 min, and chased in the presence of unlabeled methionine and cycloheximide for either 0, 10, or 30 min, as indicated below the lanes. Extracts were immunoprecipitated with a monoclonal antibody to ␤-gal, resolved by SDS-polyacrylamide gel electrophoresis, and fluorographed (see "Experimental Procedures"). Bands representing cleaved X-␤-gal, uncleaved Ub-X-␤-gal, and Ub-X-␤-gal species bearing a multi-ubiquitin chain (Ub n -X-␤-gal) are indicated on the left. An arrow indicates an ϳ90-kDa degradation product of ␤-gal (26).
Co-expression of GST-USP3 with Ub-P-␤-gal resulted in a substantial increase in levels of a protein of the size of P-␤-gal (Fig. 6B). Because P-␤-gal is a stable protein in the yeast N-end rule (38), this accounts for the substantially increased steadystate ␤-gal levels observed (Fig. 6A). The proportion of cleavage was higher when GST-USP3 was expressed from the high copy plasmid than the low copy plasmid (data not shown), consistent with a dose effect. It is hard to judge whether USP3 has an isopeptidase activity against the multi-ubiquitin chain attached to Ub-P-␤-gal, as the amount of Ub-P-␤-gal available for multi-ubiquitination diminishes as more is cleaved to P-␤-gal. Because similar results were reported recently for the mouse Unp and human Unph, together with a demonstration that this cleavage occurs precisely at the ubiquitin-proline junction (5), it can be inferred that USP3 is also cleaving this bond.
USP3 Disrupts Ubiquitin-dependent Proteolysis-One role of ubiquitin-dependent proteolysis is to selectively eliminate abnormal proteins, protecting against their toxic effects (28,29,40). To evaluate the involvement of USP3 in ubiquitin-dependent proteolysis, we examined the effect of GST-USP3 expression in cells grown on canavanine, an analogue of arginine. When incorporated into proteins, canavanine results in abnormal proteins that are degraded by the ubiquitin pathway. Disruption of this process renders cells more susceptible to the toxic effects of misfolded proteins. Expression of GST-USP3 from the low copy number plasmid pKT7 produced a negative (but not statistically significant) trend in viability with increasing canavanine concentration, indicating that USP3 was interfering with this process (Fig. 7). When GST-USP3 was expressed from the high copy plasmid pKT8, it caused a significant decrease in the number of viable cells at high concentrations of canavanine (1-1.5 g/ml), confirming this trend and consistent with USP3 causing an inhibition of general ubiquitin-dependent proteolysis. The more trivial explanation, that the USP3 protein is toxic in its own right, does not apply, because there was no effect of GST-USP3 expression on cell viability in the absence of canavanine (raw viable cell data from 0 g/ml canavanine lanes, Fig. 7; data not shown).

DISCUSSION
Through this study the complete cDNA of a new and previously uncharacterized human Ubp has been isolated. According to a systematic nomenclature proposal for human Ubps (41), we have named this enzyme USP3, and the new names for other human Ubps are presented parenthetically in the following discussion. The USP3 amino acid sequence contains two highly conserved sequence regions, the Cys-and His-boxes, containing a conserved cysteine residue and two conserved histidine residues, respectively. These regions were originally proposed to contain the active site Cys and His residues of these thiol proteases (14), and mutagenesis studies have confirmed this in several Ubps (e.g. Refs. 5, 9, 15, and 42-44). USP3 also contains two other conserved Ubp regions, the Asp-, or Gln-box, and the KRF-box, both of presently unknown function (3,8,10). Unlike all Ubps studied to date, which have been successfully expressed in E. coli (48), expression of the USP3 protein in E. coli was unsuccessful, presumably because of toxicity, and thus an alternative expression system was required. Expression of USP3 in yeast as a GST fusion was successful, albeit at very low levels. The ubiquitin cleavage activity exhibited by USP3 is consistent with the presence of the Cys-and His-boxes, containing conserved residues that in other Ubps have been found to be functionally important (5,9,15,(42)(43)(44).
Southern hybridization analysis of the USP3 gene suggests that it is a single copy gene. The USP3 gene mapped uniquely to human chromosome 15q22.3 by both radiation hybrid and fluorescence in situ hybridization mapping, with no secondary hybridization signal detected, and it can be inferred that no closely related sequences exist elsewhere in the genome. Most Ubp genes have been mapped, and in humans these include DFFRX(USP9X) and DFFRY(USP9Y), which localize to the X and Y chromosomes, respectively (11,12); UNP(USP4), which localizes to 3p21.3 (4); ISOT-1/2(USP5) to 12p13 (45); and ISOT-3(USP13) to 3q26.2 (46). At present, only the mouse DUB family of Ubps forms a cluster of at least 4 genes on chromosome 7 (43,44). While this manuscript was in preparation, Fujiwara et al. (52) reported the identification and chromosomal localization of USP1, a novel human USP of 785 amino acids/88 kDa, that localizes to chromosome 1p31.3-p32.1. USP1 is clearly distinct to USP3 in size, sequence, and chromosomal location, and notably was also identified by a random cDNA sequencing program.
A USP3 probe hybridizes to a transcript of 5.8 kb in addition to the expected 2.45-kb mRNA in all tissues examined. Similar large variation in transcript sizes have been observed in some other Ubps. Most nonyeast Ubps, including Unp, Unph, Hausp, fam, and faf, also have two RNA transcripts (4,6,(47)(48)(49)(50). The two Unp transcripts (3.5 and 3.7 kb) result from alternate polyadenylation sites at the 3Ј end of the sequence (4), whereas the mouse fam (11.5, 10, and 8.5 kb) and faf (8.1 and 8.2 kb) mRNAs also vary only in their 3Ј-untranslated regions. In addition, both Unph and ISOT-1/2 give rise to two transcripts because of alternate splicing of small coding region exons 8, 10, 48). We did identify one EST clone that included an extra 138 bp within the USP3 coding region, but this presumably represents a splicing error, because it would disrupt the reading frame. ISOT-3 gives rise to three transcripts of 3.5 kb (the expected size from the cDNA), 5.5 and 8.5 kb, presumably due to alternate polyadenylation sites (46). The USP3 5.8-kb transcript may thus also be due to different lengths of the 3Јuntranslated region. Further analysis of this 5.8-kb transcript, together with sequencing of the USP3 gene, should determine the reason for this large variation in transcript size, and provide a better understanding of its functional significance.
The low level of USP3 expression, emphasized by an inability to detect USP3 transcripts on Northern blots of total RNA, reflects the situation of another human Ubp, Hausp(USP7): transcripts of Hausp, like USP3, were also detected only on blots of poly(A) ϩ RNA (6). USP3 is expressed at approximately equal levels in heart, placenta, lung, liver, kidney, and skeletal muscle, at relatively high levels in the pancreas, and at much lower levels in brain. Mouse Unp, and human Unph, DFFRX, FIG. 7. Effects of USP3 expression on ubiquitin-dependent proteolysis of abnormal proteins. Yeast expressing the GST protein (control), the GST-USP3 protein from a low copy plasmid (pKT7), or the GST-USP3 protein from a high copy plasmid (pKT8), were grown on media containing various concentrations of canavanine (x axis). Cell viability was expressed as a percentage of the cells growing on no canavanine (y axis). The means Ϯ S.D. of viable cells were calculated from triplicate platings. DFFRY, are expressed in all tissues studied (4, 11, 12, 48 -50). In contrast, ISOT-1/2 and ISOT-3 show marked tissue-specific expression; the former strong in brain, and the latter strong in testes and ovary (45,46). Different expression levels in different tissues may reflect different abundance of the substrate(s) whose ubiquitination state is regulated by these Ubps.
An unusual activity exhibited by USP3 is the ability to cleave the ubiquitin-proline bond. USP3 is only the second characterized Ubp able to cleave the glycine 76 -proline in the Ub-P-␤-gal fusion. To date, only the mouse Unp and its human homolog, Unph, have demonstrated this ability. The significance of this activity is not known at present; however, a naturally occurring ubiquitin-like protein with a C-terminal proline residue linking it to a 590-amino acid protein (known together as An1p) has been identified in Xenopus laevis (51). The ability of USP3 to cleave this protein is currently being investigated.
The ability of USP3 to cleave linear ubiquitin fusion proteins suggests that USP3 could participate in the generation of free ubiquitin molecules from ubiquitin precursor fusions. This may not, however, be its primary function, because the pulse-chase analysis of L-␤-gal degradation indicates that USP3 can inhibit ubiquitin-dependent proteolysis of this protein, the simplest explanation being a trimming isopeptidase cleavage of ubiquitin moieties from the multi-ubiquitin chain targeting it for degradation. The ability of USP3 to inhibit ubiquitin-dependent proteolysis in yeast is further emphasized by its effect on the degradation of abnormal, canavanine-containing proteins.
An alternate explanation is that USP3 may have an isopeptidase-T/Ubp14p-like activity in cleaving unanchored ubiquitin chains, because overexpression of Ubp14 in yeast has been observed to also stabilize the N-end rule substrate L-␤-gal (but not other substrates; Ref 15). As also suggested for Ubp14p, USP3 may form aberrant complexes with other components of the ubiquitin system and cause a general impairment of function (15).
It is difficult to suggest what the specific deubiquitinating function of USP3 may be at present. Our observed minor disruption to general ubiquitin-dependent proteolysis suggests that USP3 may participate in regulation of ubiquitin-dependent proteolysis of key substrate(s) of the ubiquitin pathway, rather than in a general regulatory role. The orthologue of the key substrate(s) whose ubiquitin-dependent degradation USP3 may help regulate in human cells may actually be absent from yeast. Complete analysis of the USP3 protein in the future will require study of USP3 in human tissue and cell lines, because only in its natural physiological environment will the impact of USP3 on ubiquitin-dependent pathways be accurately assessed. Nevertheless, this initial characterization of USP3 has provided an insight into its role and significance as a potential regulatory component of the ubiquitin-dependent proteolytic pathway.