The Human UNP Locus at 3P21.31 Encodes Two Tissue-Selective

The human UNP locus at 3p21.31 encodes two tissue-selective, cytoplasmic isoforms with deubiquitinating activity that have reduced expression in small cell lung carcinoma cell lines

Amy Frederick, Mark Rolfe and M Isabel Chiu

Mitotix Incorporated, One Kendall Square, Building 600, Cambridge, Massachusetts 02139, USA

The human Unp gene at 3p21.3 has sequence similarity The transfer of ubiquitin to the target substrate is a to ubiquitin proteases and has been suggested to play a multi-step process. Ubiquitin is ®rst activated in an role in carcinogenesis of the lung (Gray et al., 1995). To ATP-requiring step by an E1 enzyme, transferred to investigate this possibility, we isolated cDNAs from one of a small family of conjugation enzymes (E2s, or several human tissue libraries and found evidence for two UBCs) and ®nally onto the substrate. The requirement major isoforms, encoding proteins predicted to either for an additional activity, E3, has been convincingly contain an internal 47 amino acid segment or not. Both demonstrated in a few instances, the ®rst being the are functional in deubiquitination assays, and mutation recognition component of the N-end-rule pathway in of a critical conserved cysteine residue to alanine the yeast S. cerevisiae encoded by the UBR1 gene abolished activity. Unp speci®es two closely-migrating (Varshavsky, 1996). Two well-characterized examples transcripts whose relative abundance varies among in mammalian cells suggest that E3s comprise a human adult tissues. Antibodies speci®c to UNP con®rm structurally diverse group. In one case, the degrada- the presence of at least two endogenous protein isoforms tion of p53 in cells infected with human papilloma of approximately 105 ± 110 kDa in cell lysates, as virus (HPV) is mediated by a complex between the predicted from the cDNA sequences. Cellular fractiona- viral E6 and a cellular protein, E6AP (Beer-Romero et tion and immunocytochemistry revealed UNP expression al., 1997; Rolfe et al., 1995; Schener et al., 1995). The localized primarily in the cytoplasm. When we examined latter is the founding member of a family of related a panel of lung-derived cell lines for both UNP mRNA enzymes sharing a motif that contains the catalytic and protein expression, we found reduced levels of UNP cysteine involved in thioester formation with ubiquitin. protein in all four small cell lung carcinoma cell lines In the second case, the E3 activity required for the tested. These ®ndings directly contradict and oer degradation of mitotic cyclins (known as the cyclosome alternative interpretations to a number of previously or APC) comprises eight subunits, none of which published observations on Unp. appear to form a covalent linkage with ubiquitin (King et al., 1995). Thus, E3s promise to be the major Keywords: ubiquitin protease; small cell lung cancer; determinants of speci®city for the ubiquitination of a 3p21 given target. Furthermore, there is compelling evidence that ubiquitin conjugation of many proteins is also regulated by phosphorylation events (Chen et al., 1996; Hicke and Riezman, 1996; Musti et al., 1997). Introduction Another class of enzymes postulated to have potential substrate selectivity are the deubiquitination Ubiquitin-mediated proteolysis is responsible for the enzymes (DUBs). These can be further grouped by turnover of select cellular regulators. In this pathway, sequence similarity into two main families, the multi-ubiquitin chains are covalently attached to target ubiquitin-carboxy-terminal hydrolases (UCHs), and substrates which are then degraded by the 26S the ubiquitin-speci®c proteases (UBPs). The latter proteasome. Recent evidence has extended the list of comprise a large and diverse group with at least 16 targets controlled by ubiquitination to include, among members in the budding yeast alone and more others, the tumor suppressors p53 and RB, the cell estimated in humans from the expressed sequence tag cycle regulators cyclin A, cyclin B, cyclin E and p27, (EST) databases. They share two signature motifs (The the Ser/Thr kinase c-mos, the transcription factors Cys box and the His box) that presumably constitute E2F, NF-kB/IkB, the proto-oncogenes c-myc,c-jun the catalytic active site (Papa and Hochstrasser, 1993). and c-fos, and membrane proteins such as the cystic Although the physiological substrates for these ®brosis transmembrane conductance regulator (CFTR) enzymes have yet to be identi®ed, they can all cleave and the IgE surface receptor (for reviews, see ubiquitin from an engineered ubiquitin-protein fusion Hochstrasser, 1995; Pahl and Baeurele, 1996; Rolfe et substrate in a bacterial assay. Their best characterized al., 1997). Aberrations in the regulated destruction of member, isopeptidase T, can remove ubiquitins such key cellular substrates can have profound processively from a multi-ubiquitin chain in vitro, physiological consequences for the cell and may thereby functioning to recycle free ubiquitin, and is underlie certain pathological conditions. expected to have a general, rather than a speci®c eect on cellular targets (Wilkinson et al., 1995). One reason for the excitement over these enzymes, however, comes Correspondence: M Chiu from their possible role as target-speci®c regulators. Received 8 July 1997; accepted 2 September 1997 Many UBPs have been identi®ed in genetic screens Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 154 whose mutant phenotypes or patterns of expression sequence homology to the UBP family, in particular implicate cellular processes as diverse as eye develop- that of the tre oncogene (Gupta et al., 1993). In ment in fruit¯y (Fischer-Vize et al., 1992; Huang et al., another study, Unp was found to specify two related 1995), transcriptional silencing (Moazed and Johnson, mRNAs that are ubiquitously expressed. Evidence for 1996), stability of the yeast MATa2 transcription nuclear localization and transforming potential of factor (Papa and Hochstrasser, 1993), cytokine UNP in athymic mice was also published (Gupta et response (Zhu et al., 1996, 1997), and oncogenesis al., 1994). Subsequently, the human cDNA homolog (Nakamura et al., 1992). Most recently, a member of of Unp was isolated (Unph) and found to hybridize the UCH family was implicated in long-term facilita- to chromosome band 3p21.3, a locus frequently tion of neuronal synapses in Aplysia (Hegde et al., rearranged in lung cancers. The authors then 1997). Whether such expectation will be ful®lled will surveyed a panel of lung tumors as well as lung require identi®cation of the physiological substrates for tumor cell lines and concluded that levels of Unp these enzymes and characterization of their selectivity mRNA expression are consistently elevated in small (reviewed in Hochstrasser, 1995; Kalderon, 1996). cell tumors and adenocarcinomas of the lung (Gray et In this paper we present molecular characterization al., 1995). Intrigued by its possible role in the of a deubiquitinating enzyme, human UNP. Unp was pathology of small cell lung cancer, we began the originally identi®ed during a survey of mouse genes characterization of human Unp. Here we present near the Mpv20 retroviral insertion site, and named results that suggest alternative interpretations to a for what the authors considered a ubiquitous nuclear number of the previously published observations on protein. Analysis of the mouse cDNA revealed Unp.

a Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 155 positions 96 and 2708, respectively (Table 1). The Results carboxy-terminal sequence dierence was con®rmed in all eight B lymphocyte cDNA clones as well as the fetal Identi®cation of Unp cDNA isoforms spleen EST cDNA clone. Conservation between the We isolated overlapping human Unp cDNA clones deduced amino acid sequence of human UnpEL and from three adult tissue sources: B lymphocyte, T the published mouse Unp sequence (Mm Unp) lymphocyte and retinal libraries (see Materials and supports the existence of the carboxy-terminal exten- methods), and used them to assemble a nucleotide sion (Figure 1b). Such comparison was not possible sequence containing the longest uninterrupted reading with the amino-terminal extension, because it falls frame for Unp (Figure 1a). We also determined the outside the boundary of the published mouse clone. sequence of a fetal spleen cDNA clone from the EST Notably, the region encoding the amino-terminal database derived from Unp (clone 72923). When the extension is within the published human Unph sequences were compared to one another and to the sequence, but was considered upstream of the initiator published human Unph sequence (Gray et al., 1995; methionine. accession number U20657), we found numerous Second, minor dierences within the coding region dierences, summarized in Table 1 and Figure 1. To were also found (Table 1), resulting in single amino avoid confusion, we refer to the prototype sequence acid substitutions at residues 324, 372 and 744 of the shown in Figure 1a, as UnpEL (extended, long deduced protein sequence, a pentapeptide substitution isoform), to the published sequence as Unph, and to at residues 401-405, and two silent polymorphisms. the locus encoding all cDNA derivatives, Unp. Third, two independent B cell cDNA isolates have First, the open reading frame of UnpEL predicts a an in-frame internal deletion of 141 nucleotides leading protein that is longer at both the amino- and carboxy- to a deduced protein shorter by 47 amino acids, or termini than the published Unph sequence by a total of 5 kDa, which we designate UnpES (Table 1). This 109 amino acid residues, and 13 kDa dierence in could be either a tissue-speci®c isoform, presumably a predicted molecular mass. The dierences were entirely product of alternative splicing, or a re¯ection of accounted for by single nucleotide insertions at individual variations in the population. For instance,

Figure 1 Characterization of human Unp cDNAs. (a) Composite nucleotide and deduced amino acid sequence of human UnpEL cDNA. Numbers for nucleotide positions are shown on the left (NT) and predicted amino acid positions on the right (AA). The signature ubiquitin-protease domains encompassing the conserved cysteine residue at position 311 (Cys box) and the two histidines at positions 873 and 881 (His box) are boxed. The underlined segment, nucleotides 696 ± 836, is absent from two independently isolated cDNAs, encoding an isoform that is 47 amino acids shorter (UnpES). Asterisks above nucleotides 96 and 2708 denote two key dierences with the published human Unph sequence leading to extensions at both the amino and carboxy termini of the predicted protein sequence. Other dierences are detailed in Table 1. Nucleotides +1 to +3 encode the presumed initiator methionine. For reference, the previously designated initiator methionine in GenBank sequence U20657 (amino acid 56 here) is underlined. The sequence of human UnpEL and UnpES have been deposited in the database (GenBank accession numbers AF017305 and AF017306 respectively). (b) Structure of Unp cDNAs. UnpEL and UnpES represent the two isoforms (long and short) identi®ed from cDNA libraries in this study. The structures of the published human Unp sequence (Gray et al., 1995) and mouse Unp sequence (Gupta et al., 1993) are based on GenBank accession numbers U20657 and L00681 respectively. UnpDL and UnpDS encode derivative constructs lacking the amino-terminal extension. Predicted molecular weights for each open reading frame were determined using DNASTAR software. Boxes represent aligned open reading frames. The amino- and carboxy- terminal extensions are marked by solid and white boxes at each end. Diagonal hatches, 47 amino acid fragment absent in short isoforms, presumably resulting from alternative splicing; horizontal hatches, carboxy-terminus predicted from U20657. The corresponding regions in mouse Unp are shown in dierent shading to denote homology, not identity Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 156 Table 1 Nucleotide sequence dierences found in various human Unp cDNA isolates Tissue origin of cDNA Nucleotide Cortex Adult Adult Adult Fetal Predicted amino acid positiona (U20657b) B cellc retina T cell spleend dierence 96 absent n.a. n.a. G n.a. 55 aa N-term extension 591 AAG AAA AAG AAG n.a. no dierence at Lys197 Long or short forms 696 ± 836 present absent present present n.a. diering by 47 aa 970 ± 972 CAC GCA GCA GCA n.a. His or Ala at residue 324 1101 GAC GAT GAC GAC n.a. no dierence at Asp367 1117 ± 1119 TGG CGC CGC CGC n.a. Trp or Arg at residue 372 1202 C absent absent absent n.a. PSRWI or LLDGL at residues 401 ± 405 1215 absent G G G n.a. 2232 AGA AGT n.a. AGT AGT Arg or Ser at residue 744 change in frame resulting in 61 aa C-term extension 2708 absent C n.a. n.a. C rather than 6 aa n.a., not applicable because region of interest falls outside of clone boundary. aas shown in Figure 1a for UnpEL cDNA. bGenBank sequence published by Gray et al. (1995). cVeri®ed from two independent cDNA isolates with dierent cloning junctions. dEST database cDNA clone 72923, sequence veri®ed by us

the two silent polymorphisms observed from our study which was isolated from a cortex library (Gray et al., (Table 1, Lys197 and Asp367) were also harbored by 1995). the same two cDNA clones. As described below, our results support the former scenario. Detection and analysis of UNP protein isoforms in human cell lines Selective expression of Unp mRNA isoforms in human We asked whether protein isoforms of Unp are present tissues in human cell lines. Two UNP-speci®c antisera were We looked at Unp mRNA expression from a panel of generated: anti-ICP1, which recognizes both long and adult human tissues to test whether the heterogeneity short isoforms of Unp, and anti-ICP3, which was in cDNAs re¯ects endogenous expression (Figure 2). raised against a GST-fusion with the 47 residue stretch Indeed, Unp-speci®c transcripts appear as a closely- absent from UnpES and recognizes only the long form migrating cluster. Within the limits of resolution are at (see Materials and methods). The antisera were least two major isoforms approximately 3.5 and 3.7 kb characterized for their speci®city by their ability to in size, similar to what was observed for mouse Unp in recognize recombinant GST-UNP fusions of varying a smaller set of tissues (Gupta et al., 1994). The same sizes and by competition with peptides (unpublished). pattern of expression was also present in human cell Indeed, all cell lines tested showed some expression lines (Figure 7a). of UNP protein, often present as a doublet, sometimes It is also apparent that whereas Unp mRNA was as a single species, as was observed with Unp detectable in all tissues tested, their overall abundance transcripts (Figure 3a, UNP, top panel). In cell lines was not high, particularly in those with proliferating expressing two forms, incubation with anti-ICP3 cells (excepting testis). The medium-to-low level of (Figure 3a, UNP-Long, bottom panel) detected only expression on the Northern blots is consistent with the a single band, demonstrating that the slower-migrating representation of cDNA clones found in our libraries species does indeed contain the 47 residue stretch, and (about one in 105 in the dT-primed library). Unp the faster-migrating species does not. The presence of a mRNA expression was highest in skeletal muscle, doublet was also observed in a panel of lung cell lines heart, pancreas and testis. (Figure 7b). Lastly, the relative abundance of Unp isoforms Notably, the endogenous UNP proteins had the varies from tissue to tissue: in brain, the slower- expected mobility (105 ± 110 kDa) consistent with the migrating species predominates, while in testis, most of open reading frames predicted by the cDNAs. This the Unp-speci®c signal came from the faster-migrating contrasts sharply with the results obtained by Gupta et species, and in skeletal muscle, both are equally al. (1994) with mouse UNP: their antibody detected a represented. In light of this tissue-selective pattern of band of 180 kDa from NIH3T3 cells, far larger than expression, it's not surprising that dierent isoforms predicted by either mouse or human Unp cDNAs. were isolated from the various libraries. We speculate The sequences of human UnpEL and Unph encode that the longer form most likely corresponds to open reading frames that dier, among other changes, UnpEL, and the shorter form UnpES, for the by the presence or absence or a 55 amino-terminal following reasons. First, the two cDNAs dier by leader (Figure 1b). To test whether the endogenous 141 bp, consistent with the observed size dierence Unp proteins can be distinguished from either form, we between the pair of transcripts. Second, if the compared the mobility of proteins produced by three correlation held, the tissue Northern result predicts transfected recombinant cDNA constructs (UnpEL, that any cDNAs derived from brain, which expresses UnpES and UnpDL) with that of the endogenous Unp predominantly the longer isoform, should contain the (Figure 3b). Protein lysates from human diploid lung 141 bp encoding the 47 amino acid insert, and indeed ®broblasts (IMR90) expressing recombinant myc- such was the case with Unph, the published sequence, tagged UnpEL and UnpES from transient transfec- Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 157 Spleen Thymus Prostate Testis Ovary Small Intestine Colon Peripheral Blood Leukocyte Heart Brain Placenta Lung Liver Skeletal Muscle Kidney Pancreas 7.5 kb —

4.4 kb — — Unp

2.4 kb —

2.0 kb — β-actin 1.8 kb

Figure 2 Expression of human Unp mRNA in adult tissues. Northern blots of poly(A)+ RNA (2 mg per lane) isolated from 16 human tissues were hybridized with a 32P-labeled probe derived from a fragment of Unp that recognizes both long and short isoforms (510 bp fragment from HpaI at 1841 to SacI at 2351), and exposed to autoradiography for 4 days. The same blots were subsequently hybridized to a b-actin probe to control for loading, and exposed for one day. Non-proliferative tissues (left) and proliferative tissues (right) were run on separate blots. Molecular weight standards in kb are marked on the left. Signal from autoradiographs were analysed as described in Materials and methods. Unp-speci®c signal from each lane was calibrated with the corresponding b-actin signal, expressed quantitatively as fold increase relative to the lowest signal in each blot and plotted. The columns representing expression in heart and skeletal muscle show two values, obtained by normalizing to only the standard 2 kb form of tubulin (black), or to both it and the 1.8 kb isoform (grey) tions were analysed alongside mock-transfected cells recognized by anti-ICP3, middle panel) migrate closely and MCF7 cells as well as lysates from cells transfected with myc-UnpEL and myc-UnpES, but more slowly with an engineered expression construct (myc-tagged than myc-UnpDL (Figure 3b, anti-ICP1, top panel). UnpDL). This construct is identical to myc-UnpEL These results suggest that endogenous Unp likely except that it lacks the 55 residue amino-terminal contains the 55 amino-terminal leader, consistent with extension, and employs the internal initiation methio- our assignment of the initiator methionine (Figure 1a). nine predicted from the Unph sequence. The myc-epitope extends the size of all three Endogenous UNP proteins are primarily cytosolic recombinant proteins equally, by 10 amino acid residues, or approximately 1 kDa. When transfected We monitored the sub-cellular localization of endo- and endogenous UNP were directly compared on a genous UNP in nuclear and cytosolic fractions of Hela denaturing gel, the endogenous UNP proteins in cells (Figure 3c) using the UNP-speci®c antibody anti- IMR90 cells (comprising at least a long isoform ICP1 on Western blots. Again, endogenous UNP in Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 158 a bc L ∆ Hela myc-UnpEL myc-UnpES IMR90 H345 293 Hela MCF7

MCF7 Mock myc-UnpEL myc-UnpES myc-Unp L N C C M T BRCA1 — UNP — UNP —

UNP — UNP — UNP __ Long Long c-MYC —

c-MYC — α-TUBULIN —

Figure 3 Characterization of endogenous cellular Unp protein. (a) Detection of endogenous cellular UNP proteins. Western blot analysis of cellular protein extracts (150 mg per lane, 4 ± 20% SDS ± PAGE) prepared from ®ve human cell lines reacted with either an anity-puri®ed antiserum to both isoforms of UNP (anti-ICP1, top panel) or an antiserum speci®c to the long isoform of UNP (anti-ICP3, bottom panel). Transfected, myc-tagged Unp constructs control for antisera speci®city. These results appear again as a subset of Figure 7b. (b) Comparison of endogenous cellular Unp with recombinant myc-tagged Unp proteins. Whole-cell extracts from IMR90 cells (100 mg per lane, 8% SDS ± PAGE) transfected with myc-tagged UnpEL, UnpES and UnpDL expression plasmids were analysed alongside extracts from mock-transfected cells and MCF7 cells. The immunoblot was reacted with an anti-Unp antiserum (anti-ICP1, top panel), an antiserum speci®c for Unp long isoforms (anti-ICP3, middle panel) and an anti-c-myc antibody which recognizes only the recombinant Unp proteins (bottom panel). (c) Detection of endogenous Unp in the cytosol. Fractionated Hela extracts (100 mg per lane, 4 ± 20% SDS ± PAGE) were analysed by immunoblotting with antisera speci®c for BRCA1 (a nuclear marker), a-tubulin (a cytoplasmic marker), c-myc (a nuclear marker), and Unp (anti-ICP1). L, whole cell lysate; N, nuclear fraction; C, cytosolic fraction; M, membrane fraction. The two cytosolic samples loaded were collected from the top and bottom (close to pellet) halves of the tube. Whole cell lysate (50 mg) from transfected cells expressing recombinant myc-UnpDL was included for comparison (T)

both the cytoplasmic fractions and in whole cell lysate transient transfection assay, and detected by immuno- has the expected mobility, slightly slower than cytochemistry using a monoclonal antibody raised recombinant myc-UnpDL. Endogenous UNP immu- against the FLAG epitope (Eastman Kodak, New noreactivity co-fractionates with a-tubulin, a cytosolic Haven, CT). The staining pattern was indistinguishable marker, and not with BRCA1 or c-myc, both nuclear from that of myc-UnpEL (data not shown). markers. These results show that UNP is localized The same results were obtained whether visible (a, d) primarily in the cytosol, and not in the nucleus. or ¯uorescent (b, e) detection was used. Taken together, the endogenous fractionation and immunostaining of myc-UNP indicate a primarily cytoplasmic Recombinant UNP is localized primarily in the localization for both UNP isoforms, which is in direct cytoplasm and is not exclusively nuclear disagreement with published reports of mouse UNP as We could not visualize the localization directly in cells a uniquely nuclear protein (Gupta et al., 1994). by immunocytochemistry because our anti-UNP We also asked whether myc-tagged UNP expression antibodies stained cells poorly. Instead, we tested constructs harboring either the short isoform UnpES immunostaining of transfected myc-UNP in IMR90 or mutant Unp derivatives (UnpDL, UnpEL-C311A, cells using anti-myc antibody (Figure 4a ± f). Expres- UnpES-C264A) led to any changes in intracellular sion of both myc-UnpEL and myc-UnpES is observed localization (Figure 4, Table 2). No major dierences throughout the cell, with frequent examples of were observed: all showed whole-cell localization with convincing nuclear-exclusion (d, e ± f), indicating a no evidence of uniquely-nuclear staining. Exclusively primarily cytoplasmic localization (Table 2). We favor cytoplasmic staining was observed at only 1% in myc- the possibility that at least some of the whole-cell UnpDL. Because of the caveat of overexpression, we staining seen with Unp is due to masking of the can only interpret these results with caution. Qualita- nucleus by the strong cytoplasmic signal, and that tively, we conclude that neither the internal 47 residue unambiguous nuclear-exclusion was evident only in stretch, nor the amino-terminal 55 amino acids, nor the fortuitously well-positioned nuclei. The related ques- conserved cysteine appear to play a major role in tion of whether the apparent additional expression in determining the speci®city of compartmentation of the nucleus has any physiological basis, perhaps related UNP. to the cell cycle, or is an artifact of overexpression remains to be determined. For instance, under the Unp isoforms encode functional deubiquitination enzymes same conditions, transfected myc-wee1 localized both exclusively in the nucleus, as expected (Heald et al., An obvious question is whether both isoforms of Unp 1993), and throughout the cell in other instances (Table encode functional deubiquitination enzymes. Because 2), presumably due to overexpression. To further both contain the signature motifs for these enzymes, ensure that the cytoplasmic localization of myc- namely the conserved Cys and His boxes, we expected UnpEL is independent of the chosen epitope, another them to be active at least in the standard bacterial construct, FLAG-UnpEL, was similarly tested in a assay for deubiquitination that de®nes other members Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 159

Figure 4 Cellular localization of recombinant human Unp cDNAs. IMR90 cells transfected with plasmids expressing myc-tagged UnpEL (a,±f)ormyc-tagged Wee1 (g±i) were analysed by immunostaining with an anti-c-myc monoclonal antibody. Staining was detected either by immunoperoxidase activity (a, d, g, brown) or immuno¯uorescence (b, e, h, red) followed by Hoechst staining to identify nuclear material (c, f, i, blue). The immuno¯uorescent and Hoechst stains are shown in pairs of identical frames visualized through dierent ®lters (b±c, e±f, h±i). Both whole-cell staining (a±c) and distinctly cytoplasmic, non-nuclear staining pro®les (d± f) were observed with myc-UnpEL (Table 2). Nuclear staining of myc-Wee1 served as control (g±i)

Table 2 Intracellular localization of transfected myc-Unp in IMR90 suggest that the internal 47 amino acid stretch that cells distinguishes UnpEL from UnpES is not indispensable Whole cell Cytoplasmic Nuclear to the cleavage reaction itself. Whether it may function Construct (%) (%) (%) in vivo to govern substrate selectivity in dierent tissues myc-UnpEL 330 (67) 165 (33) 0 (0) is an open question. myc-UnpES 200 (100) 0 (0) 0 (0) We also examined the deubiquitinating activity of myc-UnpEL,C311A 150 (100) 0 (0) 0 (0) UnpDL and UnpDS and found both capable of myc-UnpES,C264A 150 (100) 0 (0) 0 (0) myc-UnpDL 168 (99) 2 (1) 0 (0) removing ubiquitin from the substrate (Figure 5a), myc-wee1 322 (41) 0 (0) 467 (59) demonstrating that the amino-terminal 55 residues are Cells transfected with each construct were stained with DAB as not absolutely required for catalysis. We note that described in Materials and methods and visualized by light although the cleavage reactions appears less ecient microscopy. The staining in each category was recorded. For each than those of UnpEL and UnpES, the current results construct, the number of cells scored and the percentage of total in are best interpreted qualitatively, because the ratios of each staining pattern are listed. See Figure 4 substrate to enzyme are not exactly controlled in this in vivo bacterial system. As with the internal 47 amino acid stretch, whether the amino-terminal 55 residues of the family (Papa and Hochstrasser, 1993). In this play any role in the selection of biological substrates assay, an arti®cial fusion linking the last amino acid of remains to be determined. mammalian ubiquitin (glycine 76) to bacterial b- galactosidase serves as the substrate, and a GST- Unp maps to human chromosome 3p21.31 UNP fusion (Figure 5b) provides the enzymatic activity in question. Presence of deubiquitinating activity is Intrigued by the many dierences between the reported evidenced by the appearance of a faster-migrating form properties of Unp and those observed in this study, we of b-galactosidase, without the attached ubiquitin decided to map the chromosomal position of UnpEL. moiety (Figure 5a). Our results show that indeed, A genomic clone of UnpEL was obtained from both UnpEL and UnpES isoforms cleaved ubiquitin screening a PAC1 library which was then used as a from the fusion substrate eciently, while their ¯ourescent probe to label human metaphase chromo- corresponding active-site mutants (cysteine-to-alanine) somes (Figure 6). Speci®c hybridization to the mid did not, demonstrating the critical role of that residue short arm of a group A chromosome (Figure 6a) was (Cys 311 in UnpEL and Cys 264 in UnpES) in the subsequently identi®ed as 3p21.31 when double-labeled deubiquitination reaction. Presumably, it is the site for with a probe speci®c to the centromere of chromosome ubiquitin thioester formation. These results also 3 (Figure 6b). Notably, the labeling was speci®c to a Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 160 single chromosomal region and not to multiple positions as expected if Unp were a member of a near-identical gene family. Since the map position obtained by us is in complete agreement with the published location of Unph, the possibility that Unph and UnpEL derive from dierent genes on dierent chromosomes is remote.

Variable expression of Unp in a panel of lung tumor cell lines Studies of allele loss in lung tumors and cell lines have de®ned chromosomal region 3p21.3 as frequently deleted in lung cancers, speci®cally small-cell cancer (SCCL) (Hibi et al., 1992; Whang-Peng et al., 1992), leading to the expectation that the region encompasses a tumor suppressor gene. We compared the abundance of Unp mRNA (Figure 7a) in a panel of eight human lung tumor cell lines to those in normal adult lung and normal diploid lung ®broblasts (IMR90). We chose to include cell lines derived from each class of lung tumors: four small-cell (H345, H69, H82, H446), two squamous (H520, SK-MES1), one large cell (H460), and one unspeci®ed tumor of the lung (A549). Unp levels in each sample were normalized to those of b- actin, a ubiquitous transcript, and expressed as fold increase over IMR90. One sample, H345, had only trace amounts of poly(A)+ mRNA and was excluded from subsequent analysis. We observed no signi®cant c misregulation of Unp expression in any of the cell lines when compared to the non-tumor samples, either in

a GST fusions L S L C256A S C209A ∆ ∆ ∆ ∆ vector Unp EL Unp ES Unp EL C311A Unp ES C264A control Unp Unp Unp Unp

Ub-β-gal β-gal

b — GST-Unp

Figure 5 Human Unp encodes a functional deubiquitinating enzyme. The ability of various GST-Unp constructs to deubiquitinate a ubiquitin-b-galactosidase fusion substrate (Ub- Figure 6 Mapping of human Unp by ¯uorescence in situ b-gal, encoded by plasmid pAF6) in bacterial cells was tested. hybridization of human metaphase chromosomes. Normal Extracts are from MC1061 cells alone (control), cells co- metaphase chromosomes derived from PHA-stimulated periph- transformed with pAF6 and pGEX empty vector (vector), or eral blood lymphocytes were hybridized to a Unp probe labeled with pAF6 and the designated pGEX-Unp cDNA plasmids with digoxigenin dUTP. All samples were counterstained with encoding either wild type or mutant derivatives where the DAPI (blue). (a) Single labeling. Speci®c hybridization to the mid conserved cysteine has been replaced by alanine. Unp EL, ES, short arm of a group A chromosome was visualized by DL and DS are as represented in Figure 1b. Bacterial lysates were incubation with ¯uoresceinated anti-digoxigenin antibody analysed by Western blotting and anti-b-galactosidase immunore- (arrows, green). (b) Double labeling of chromosome 3. activity (a). The fusion substrate (Ub-b-gal) and the cleaved Chromosomes were co-hybridized with the Unp-speci®c probe product (b-gal) are distinguishable by their dierent mobility. (b) (open arrows, green) and a biotin-labeled probe speci®c to the To detect expression of each GST-Unp fusion protein, a tenfold centromere of chromosome 3 (closed arrows, red). (c), Two increase in protein extracts (8 ± 10 mg) were loaded on a SDS ± ideograms illustrating the chromosomal position of Unp at PAGE gel and immunoblotted with an anti-Unp antiserum (anti- 3p21.31 (International System for Human Cytogenetic Nomen- ICP1; see Materials and methods) clature 1995) Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 161

a Small Squamous b Small Squamous adult lung IMR90 H345 H446 H69 H82 A549 H460 H520 SKMES MCF7 myc-UnpEL myc-UnpES IMR90 H345 H446 H69 H82 A549 H460 H520 DKMES 293 Hela

4.4 kb — — UNPEL == Unp 2.4 kb — 1.4 kb — == UNP

β 2.4 kb — — -actin — α-tubulin 1.4 kb —

Figure 7 Abundance of human Unp mRNA and protein in a panel of cell lines. (a) Northern blot analysis of poly(A)+ RNA (5 mg per lane) isolated from eight human lung carcinoma cell lines, a normal diploid lung ®broblast line (IMR90), a non-lung carcinoma line (MCF7, breast adenocarcinoma) and normal adult lung tissue. The lung carcinoma cell lines comprise four small cell carcinoma (NCl-H345, H446, H69, H82), two squamous carcinoma (NCl-H520, SK-MES1), one non-small- cell carcinoma (H460), and one unspeci®ed lung tumor line (A549). The blot was hybridized with a 32P-labeled probe derived from Unp (as described in Figure 2) and exposed for 60 h. The b-actin control blot was exposed for 24 h. Molecular weight standards in kb are shown at left. Signal from autoradiographs were analysed as described in Materials and methods. Unp-speci®c signal from each lane was calibrated with the corresponding b-actin signal, expressed quantitatively as fold increase relative to the signal for IMR90 and plotted. (b) Western blot analysis of cellular protein extracts (150 mg per lane) prepared from the same panel of human cell lines described above. For comparison, extracts from IMR90 cells transfected with recombinant myc-tagged Unp plasmids were also included. The same immunoblot was reacted sequentially with an antiserum speci®c to the long isoform of UNP (anti-ICP, top panel); an anity- puri®ed antiserum to both isoforms of UNP (anti-ICP1, middle panel); and an anti-tubulin antiserum (bottom panel), with stripping between each antiserum incubation. Unp-speci®c signal (middle panel) from each lane was calibrated with the corresponding a- tubulin signal (bottom), expressed quantitatively as fold increase relative to the signal for IMR90 and plotted abundance or size. Variations in abundance observed Unp locus. Nevertheless, the reduced expression among samples were subtle (less than twofold), no observed in the limited set of small cell carcinoma greater than the dierence observed between the adult cell lines ful®lls one criterion for candidate tumor lung sample and the lung ®broblast sample. suppressor genes at 3p21.3. However, when the comparison was made with UNP protein expression normalized to a-tubulin, a tentative correlation between abundance and tumor Discussion origin was noted (Figure 7b). All four lines of small cell lung origin as well as a squamous carcinoma cell In light of the results presented in this study, we line (H520) had reduced expression levels, ranging propose that the human Unp locus at chromosome from approximately one third (H69), two thirds (H345 3p21.31 encodes at least two isoforms UnpEL and and H446) to four-®fths (H82 and H520) of the levels UnpES, distinguished by the presence or absence of an in IMR90. The remaining three lung tumor cell lines internal 47 amino acid stretch, whose relative had levels equal to or slightly higher than IMR90 abundance varies from tissue to tissue. The proteins (A549, H460, SKMES1). Endogenous UnpEL is have a molecular mass of approximately 105 ± 110 kDa present in all the cell lines as judged by immuno- and can be found either exclusively in the cytoplasm or reactivity with anti-ICP3 (top panel). Likewise, the throughout the cell. Both isoforms possess a deubiqui- presence of a smaller, closely-migrating species, tinating activity that requires the presence of a critical presumably UnpES, is seen in most cell lines in the cysteine residue, the presumed thiol-forming active site anti-ICP1 immunoreactive blot (middle) that is absent in the conserved Cys box unique to this family of in the anti-ICP3 blot (top). The presence of UNP ubiquitin proteases. Finally, although expression of protein in all samples tested demonstrates that each cell Unp transcripts appear unaltered when a panel of lung line contains at least one chromosomal copy of the cancer cell lines was compared to normal ®broblasts or Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 162 normal lung tissue, the abundance of UNP protein is need not be mutually exclusive; for instance, TGFb slightly reduced in all four small cell carcinoma cell exerts opposing eects on dierent target cells (for lines tested. review, see Kingsley, 1994; Massague et al., 1994), and Although some of the above results apparently the tumor suppressor p53 was originally identi®ed as a contradict published observations about Unp, we proto-oncogene (Linzer and Levine, 1979). suggest alternative interpretations to the published The question of what physiological role the family of data that may reconcile these dierences. The deubiquitination enzymes plays remains open. Do they variations in cDNA sequence resulting in a dierent act soley to debranch ubiquitin from multichains for predicted amino acid sequence can easily be explained recycling purposes, or do they have a more subtle by errors in DNA sequence determination. The regulatory function in determining the stability of designation of UnpEL coding sequence further speci®c target substrates in the cell? Ultimately, the eliminates the small open reading frames whose answer lies in elucidating the biological targets for each unusual presence in the 5' untranslated sequence of member of the family, including Unp, and in under- Unph required explanation. The carboxy-terminal standing the mechanisms specifying both catalysis and extension in UnpEL is conserved in mouse Unp, and target selection. the amino-terminal extension neatly explains why UnpDL protein (using the internal methionine initiator predicted in Unph) appears smaller than endogenous Materials and methods UNP. The unaccountably large 180 kDa nuclear protein described from NIH3T3 cells is most probably Cell culture not mouse UNP, but rather an unrelated cross-reacting All cell lines were obtained from American Type Culture species recognized by the antiserum. We have observed Collection. The lung panel includes cell lines derived from a similarly-sized band unrelated to Unp which is human small cell lung carcinoma (NCI H345, NCI H69, recognized by our crude anti-Unp antisera but not by NCI H82, NCI H446), large cell carcinoma (NCI H460), the anity-puri®ed antisera (unpublished observa- squamous cell carcinoma (NCI H520, SK-MES-1), an tions). We think it unlikely that the discrepancy be unspeci®ed human lung carcinoma (A549), and human attributed solely to species dierences, given the high fetal diploid lung ®broblasts (IMR90). Human cell lines conservation generally observed between mouse and derived from other tissues include transformed embryonic human proteins. Last, unlike the ®ndings of Gray et al. kidney ®broblasts (293S), cervical carcinoma (HeLa), and (1995), we found no evidence for overexpression of breast adenocarcinoma (MCF7). Cells were maintained Unp transcripts in cell lines derived from small cell according to recommendations in either RPMI 1640 medium (Life Technologies) or Dulbecco's modi®ed lung carcinomas. This could be due to either the Eagle's medium (DMEM) supplemented with 10% fetal dierent cell lines tested, or dierences in procedures calf serum, 2 mM glutamine, 100 U/mL penicillin and 1 mg/ used for normalization and quanti®cation. We have mL streptomycin. also not tested primary tumor samples. The tumor suppressor gene in 3p21.3 remains Isolation of human Unp cDNAs elusive. Recent advancements in positional cloning coupled with the ever-growing list of DNA sequences Database searches for Unp clones were performed using the in databases have helped identify several candidates BLAST network service of the NCBI. The dbEST clone that map to the region, including MLH1, encoding the number 72923, encoding a 1.3 kb fragment from the 3' end of human homolog of the E. coli DNA mismatch repair Unp cDNA, was from a human fetal spleen library and was purchased from Research Genetics. Using part of the insert enzyme mutL, and SEAM, a gene encoding a as probe, 57 positives were obtained from screening 26106 semaphorin that appears deleted in primary tumors plaques of a Epstein-Barr virus transformed human B (Roche et al., 1996). Reports of genes mapping to lymphocyte, poly dT-primed cDNA library (Durfee et al., 3p21.3 with elevated expression in lung cancer cell lines 1993), of which eight were plaque-puri®ed, excised, and their or tumors include the alpha subunit of intergrin (Hibi junction DNA sequence determined. The two longest clones et al., 1994), the multifunctional protein 37LRP/p40 (pIC644 and pIC648) were sequenced in their entirety. A (Jackers et al., 1996) and Unph (Gray et al., 1995). Our retinal cDNA fragment (from nucleotide 505 ± 1236 of observation that levels of UNP protein are reduced, UnpEL) was obtained by PCR ampli®cation of a human rather than elevated, relative to normal lung ®broblasts retinal cDNA library (Nathans and Hogness, 1984) and in all four small cell lung tumor lines tested leaves open subcloning into pUC118 (pIC643). Last, a cDNA clone was isolated from a human T cell library (random and poly dT the possibility that Unp may be a candidate for the primed, Stratagene) that contained more 5' sequence than all tumor suppressor locus implicated in the pathogenesis the others (pAF36). Full-length UnpEL, UnpES, and of lung tumors attributed to 3p21.3. Further clarifica- UnpDL were reconstructed by replacing the 5' end of tion awaits investigation of protein levels in direct pIC644, which contains full-length UnpDS, with appropri- tumor samples. ate 5' fragments from the other clones. The DNA sequences It is formally conceivable that Unp may yet possess of all four full-length cDNA clones, pertinent regions of all oncogenic potential of its own. Since details of the the parent cDNA clones along with that of the dbESTclone initial identi®cation of Unp are unpublished, it is not 72923, were determined on both strands (ABI) and analysed clear whether the original retroviral insertion around using the Sequencher software. the Unp locus in mice resulted in any transforming phenotype. Reports of the transforming potential of Construction of the cysteine to alanine mutation Unp constructs in athymic mice await con®rmation The conserved cysteine at the Cys box motif was changed to with larger sample sizes than those initially reported by alanine by oligonucleotide mutagenesis. Coding and non- Gupta et al. (1994). Theoretically, a locus with both coding strand oligonucleotides containing the mutation transforming potential and tumor suppressing activity (MIC156, 5'-GGAAACCTGGGAAACACCGCCTTCAT- Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 163 GAACTCCGCTTTG-3' and MIC157, 5'-CAAAGCG- speci®cally recognizes the long isoforms of UNP. The GAG TTCATGAA GGCGGTGTTTCCCAGGTTTCC - 3') rabbit anti-b-galactosidase antibody was from Cappel were each paired with separate oligonucleotides (MIC157 (Organon Teknika) and used at 1 : 3000 dilution. The with MIC130 and MIC156 with MIC139) to PCR amplify mouse monoclonal antibodies speci®c for c-myc and two overlapping fragments. The latter were subsequently BRCA1 were both from Oncogene Research Products combined in a ®nal PCR reaction with oligonucleotides (Calbiochem) and used at 1 : 100. The anti-mouse or anti- MIC130 and MIC139, digested with EcoRI and XbaI, cloned rabbit HRP-linked IgG antibodies were from Amersham into plasmids with UnpEL, UnpES, UnpDL, and UnpDSto and used at 1 : 4000 dilution. replace the wild type fragment. The resulting clones were sequenced on both strands to con®rm the presence of the desired mutation. Band quanti®cation and anlaysis Autoradiograms of Northerns and Westerns were analysed Northerns using the Gel Doc 1000 (Biorad) image analysis system and Molecular Analyst/Macintosh software (Biorad). Signals The human adult multiple tissue Northern blots were used for the analyses were in the linear range; overexposed purchased from Clontech and hybridized as below. For the autoradiograms with saturated signals were avoided. lung panel Northern blots, poly(A)+ RNA was isolated from approximately 16108 cells using the Invitrogen Fast Track 2.0 kit. RNA yield was determined by absorbance at Transfection of recombinant Unp clones 260 nm. 5 mg of mRNA was loaded per lane and The myc-tagged-Unp cDNA expression constructs were electrophoresed through a 1.2% agarose gel containing driven by the human CMV promoter in a pcDNA3 formaldehyde at a ®nal concentration of 2.2 M.The (invitrogen) vector backbone derived from the plasmid mRNA was transferred onto a Nytran+ membrane CMV-myc-Wee1, a kind gift from Dr Frank McKeon (Schleicher and Schuell) by capillary action. The mem- (Heald et al., 1993). IMR90 cells were transfected using brane was hybridized overnight at 428C with a 32P-labeled the Transit LT1 polyamine reagent (Mirus) according to the probe (446108 c.p.m./mg DNA) derived from either manufacturer's instructions. Brie¯y, cells were seeded at nucleotides 1874 ± 3112 of UnpEL cDNA or human b- 40% con¯uence in 10 cm plates with 0.2% gelatin-coated actin cDNA (Clontech) in 56SSC, 56Denhardt's solu- coverslips. The next day, 110 ml of Transit LT1 was pre tion, 30% formamide, 1% SDS, and 200 mg/mL denatured incubated in 500 ml of Opti-Mem reduced serum media (Life salmon sperm DNA. The ®lter was washed at 658Cin Technologies), followed by addition of 10 mg of each DNA 0.26SSC and 0.1% SDS, and exposed to X-ray ®lm (Fuji) (myc tagged versions of UnpEL, UnpES, UnpEL-C311A, with an intensifying screen at 7808C. UnpES-C264A, UnpDL, or Wee1) and incubation for 10 min. Cells were rinsed then replaced with Opti-MEM medium. The DNA mixture was added to the cells and Westerns incubated for 8 h at 378C, 5% CO2. The medium was Unless speci®ed otherwise, whole-cell extracts were prepared removed and replaced with DMEM supplemented with 10% as described (Tam et al., 1994), in lysis buer (50 mM Tris- fetal calf serum. Cells were harvested by scraping 40 h later. HCl (pH 7.4), 0.25 mM NaCl, 1% Triton-X-100, 1 mM

EDTA, 50 mM NaF, 1 mM DTT, 0.1 mM Na3VO4) with protease inhibitors (0.1 mM phenylmethylsulfonyl ¯uoride, Cell fractionation 1 mg/ml leupeptin, 10 mg/ml L-1 chlor-3-(4-tosylamido)-4- The protocol for cell fractionation was adapted from Lee phenyl-2-butanone, 1 mg/ml aprotinin, 10 mg/ml L chlor-3- et al., (1987). Hela cells (four 15 cm plates *66107)were (4-tosylamido)7-amino-2-heptanone hydrochloride). After rinsed twice in PBS, collected by scraping in PBS/0.5 mM incubation on ice for 60 min, the samples were centrifuged EDTA, and pelleted at 1500 r.p.m. for 5 min (Beckman, at 14 000 r.p.m. in an Eppendorf microfuge for 10 min at GS-6KR). All subsequent steps were at 48C. After 48C to recover the supernatant. The amount of protein was resuspension in cold hypotonic RSB buer (10 mM determined using Bio-RAD protein assay. Extracts were HEPES pH 6.2, 10 mM NaCl, 1.5 mM MgCl2 and separated by 8% SDS polyacrylamide gel (Novex) electro- proteinase inhibitors), one quarter of the cells were phoresis and transferred onto a nitrocellulose membrane pelleted and resuspended in RIPA (50 mM Tris, pH 8.0/ (Novex). Membranes were blocked in 5% milk in PBS/0.1% 150 mM NaCl/0.1% SDS/0.1% sodium deoxycholate/1% Tween (PBST) for 1 h, incubated with primary antibody in NP40 with phosphatase and protease inhibitors) lysis 1% milk in PBST for 2 h, then with the appropriate buer (Whole cell lysate). The remaining cells were rinsed horseradish peroxidase-linked secondary antibody, and again in RSB, lysed by 20 strokes in a tight-®tting Dounce detected by enhanced chemiluminescence (Pierce) according homogenizer in 3 ml volume, and centrifuged at to the manufacturer's instructions. Membranes that were 1500 r.p.m. for 15 min (Beckman, GS-6KR). The pellet reblotted were ®rst incubated in 0.2 N NaOH for 10 min at was rinsed once more in RSB, and then resuspended in room temperature and washed in PBST. Incubations were at RIPA lysis buer (Nuclear fraction). The supernatant was 258C. further fractionated into cytosolic (supernatant) and membrane (pellet) components by centrifugation at 35 000 r.p.m. for 30 min (Beckman SW40Ti). After Antibodies determining protein concentrations, samples were resus- The rabbit antiserum anti-ICP1 was raised against the pended in Laemmli buer. peptide CHFSKADTIATIEK derived from residues 154 ± 167 from the predicted sequence of UNP, and anity puri®ed with a CNBr-activated sepharose column (Phar- Immunocytochemistry macia) coupled with the same peptide. Anity-puri®ed Transfected cells on coverslips were treated with ice-cold anti-ICP1 was used at 0.3 mg/ml for Westerns, and detects acetone:methanol (1 : 1) for 3 ± 5 min, rinsed in PBS, both long and short isoforms of UNP. The rabbit reacted with 5% normal horse serum to block nonspeci®c antiserum anti-ICP3 was raised against the GST-fusion binding, then with anti-myc monoclonal antibody (Onco- peptide derived from residues 233 ± 279 from the predicted gene Research Products) at 10 mg/ml for 1 h. Treatment sequence of UNP, and used at 1 : 5000 dilution. It with biotinylated-secondary-IgG (1 h) was followed by Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 164 incubation with avidin-biotin-peroxidase complex and 8% SDS ± PAGE, and analysed by immunoblotting as visualization of immunoperoxidase activity by incubation described. For detection of b-galactosidase in the with diaminobenzidine (DAB, 0.5 mg/ml) and hydrogen deubiquitination assay, 50 ± 800 ng of total extracts peroxide (0.015%) as recommended by the manufacturer suced, depending on the success of induction. To (Vector Labs). For immuno¯uorescence staining, the ®xed monitor the presence of GST-UNPs, approximately coverslips were ®rst blocked in DMEM, 10% fetal calf tenfold amounts (8 ± 10 mg) of proteins were loaded. serum, 0.1% azide (DFA) for 10 min, incubated with anti- myc (1 : 10 in DFA) for 1 h, then treated for 30 min with Chromosomal localization biotinylated anti-mouse IgG (1 : 50 in DFA) and subsequently, Texas Red conjugated streptavidin (1 : 100 in Chromosomal mapping of human Unp by ¯uorescence in DFA) for 30 min. All incubation steps were at 258C. To situ hybridization (FISH) was performed by Genome visualize nuclear DNA, Hoechst reagent (1 : 10 000 in PBS) Systems (St Louis, Missouri). A 2.9 kb Unp cDNA was added to each coverslip for 2 min. Coverslips were fragment (XhoI fragment of pIC653) was used to screen a mounted onto slides with 90% glycerol/PBS. human genomic Pac library containing an average insert size of 120 kb. A single positive clone was obtained which following DNA puri®cation was labeled with digoxigenin Deubiquitination assay dUTP by nick translation and used as a probe for FISH. GST-Unp fusion constructs were made in pGEX-4T Labeled probe was combined with sheared human DNA vectors (Pharmacia). The ubiquitin-b-galactosidase fusion and hybridized to normal metaphase chromosomes derived (pAF6) was constructed in two steps. First, coding and from PHA-stimulated peripheral blood lymphocytes in a noncoding strand oligonucleotides engineered with SalI solution containing 50% formamide, 10% dextran sulfate and NcoI sites wre used to PCR amplify the human and 26SSC. Speci®c hybridization signals were detected ubiquitin cDNA using dbEST clone 37797 (Research by incubating the hybridized slides in ¯uoresceinated anti- Genetics) as a template, and the digested product was digoxigenin antibodies followed by counterstaining with cloned into pLacF vector containing the E. coli lacZ gene DAPI, with or without cohybridization with biotin-labeled, (gift from J Peschon). The ubiquitin-b-galactosidase fusion chromosome-speci®c probes. A total of 80 metaphase cells was then digested with SalIandBglII and subcloned into were analysed with 69 exhibiting speci®c labeling. pACYC184 (New England Biolabs) which harbors the p15A origin of replication and the gene for chloramphenicol-resistance. The GST-Unp fusion constructs were cotransformed with pAF6 into MC1061 cells. Ampicillin Acknowledgements plus chloramphenicol resistant transformants were assayed We are grateful to Frank McKeon for the myc-Wee1 for deubiquitination activity. 50 mL of ech bacterial expression plasmid, Jeremy Nathans for the human retinal

transformant were grown at 378C until A600=0.4, then cDNA library, Jacques Peschon for placZ, Judith Whittier induced with 1 mM IPTG at 258C with vigorous shaking for DNA sequencing, Lou Lamphere for help with for 4 h. Cells were harvested at 4200 r.p.m. for 20 min at densitometry measurements, Chuck Weitz for help with 48C, then lysed in 1% Triton/PBS by 4615 s sonications graphics, Peter Strack, Peggy Romero and Lou Lamphere on ice. The supernatant was recovered by centrifuging at for critical reading of the manuscript, and Guilio Draetta 14 000 r.p.m. for 10 min at 48C. Extracts were resolved by for helpful discussions.

References

Beer-Romero P, Glass S and Rolfe M. (1997). Oncogene, 14, Kalderon D. (1996). Curr. Biol., 6, 662 ± 665. 595 ± 602. King RW, Peters J-M, Tugendreich S, Rolfe M, Hieter P, et Chen ZJ, Parent L and Maniatis T. (1996). Cell, 84, 853 ± al. (1995). Cell, 81, 279 ± 288. 862. Kingsley DM. (1994). Genes Dev., 8, 133 ± 146. Durfee T, Becherer K, Chen P-L, Yeh S-H, Yang Y, et al. Lee W-H, Shew J-Y, Hong FD, Sery TW, Donose LA, et al. (1993). Genes Development, 7, 555 ± 569. (1987). Nature, 329, 642 ± 645. Fischer-Vize JA, Rubin GM and Lehmann R. (1992). Linzer DIH and Levine AJ. (1979). Cell, 17, 43 ± 52. Development, 116, 985 ± 1000. Massague J, Attisano L and Wrana JL. (1994). Trends Cell Gray DA, Inazawa J, Gupta K, Wong A, Ueda R, et al. Biol., 4, 172 ± 178. (1995). Oncogene, 10, 2179 ± 2183. Moazed D and Johnson AD. (1996). Cell, 86, 667 ± 677. Gupta K, Chevrette M and Gray DA. (1994). Oncogene, 9, Musti AM, Treier M and Bohmann D. (1997). Science, 275, 1729 ± 1731. 400 ± 402. Gupta K, Copeland NG, Gilbert DJ, Jenkins NA and Gray Nakamura T, Hillova J, Mariage-Samson R, Onno M, DA. (1993). Oncogene, 8, 2307 ± 2310. Huebner K, et al. (1992). Oncogene, 7, 733 ± 741. Heald R, McLoughlin M and McKeon F. (1993). Cell, 74, Nathans J and Hogness DS. (1984). Proc. Natl. Acad. Sci. 463 ± 474. USA, 81, 4851 ± 4855. Hegde AN, Inokuchi K, Pei W, Casadio A, Ghirardi M, et al. Pahl HL and Baeuerle P. (1996). Curr. Opin. Cell Biol., 8, (1997). Cell, 89, 115 ± 126. 340 ± 347. Hibi K, Takahashi T, Yamakawa K, Ueda R and Sekido Y. Papa F and Hochstrasser M. (1993). Nature, 366, 313 ± 319. (1992). Oncogene, 7, 445 ± 449. RocheJ,BoldogF,RobinsonM,RobinsonL,Varella- Hibi K, Yamakawa K, Ueda R, Horio Y, Murata Y, et al. Garcia M, et al. (1996). Oncogene, 12, 1289 ± 1297. (1994). Oncogene, 9, 611 ± 619. Rolfe M, Chiu MI and Pagano M. (1997). J. Molecul. Med., Hicke L and Riezman H. (1996). Cell, 84, 277 ± 287. 75, 5 ± 17. Hochstrasser M. (1995). Curr. Opin. Cell Biol., 7, 215 ± 223. Rolfe M, Romero P, Glass S, Eckstein J, Berdo I, et al. Huang Y, Baker RT and Fishcer-Vize JA. (1995). Science, (1995). Proc. Natl. Acad. Sci. USA, 92, 3264 ± 3268. 270, 1828 ± 1831. Schener M, Nuber U and Huibregtse J. (1995). Nature, 373, Jackers P, Minoletti F, Belotti D, Clausse N, Sozzi G, et al. 81 ± 83. (1996). Oncogene, 13, 495 ± 503. Human UNP encodes two cytoplasmic ubiquitin proteases AFredericket al 165 Tam SW, Shay JW and Pagano M. (1994). Cancer Res., 54, Wilkinson KD, Tashayev VL, O'Connor LB, Larsen CN, 5816 ± 5820. Kasperek E, et al. (1995). Biochemistry, 34, 14535 ± 14546. Varshavsky A. (1996). Proc. Natl. Acad. Sci. USA, 93, Zhu Y, Carroll M, Papa F, Hochstrasser M and D'Andrea A. 12142 ± 12149. (1996). Proc. Natl. Acad. Sci. USA, 93, 3275 ± 3279. Whang-Peng J, Bunn Jr P, Kao-Shan C, Lee E, Carney D, et Zhu Y, Lambert K, Corless C, Copeland N, Gilbert D, et al. al. (1982). Cancer Genet. Cytogenet., 6, 119 ± 134. (1997). J. Biol. Chem., 272, 51 ± 57.