UBP41 Is a Proapoptotic Ubiquitin-specific Protease¹

The addition of ubiquitin-chains to proteins controls many biochemical processes including cell cycle progression, protein localization, and signal transduction (1, 2, 3). Most of these cellular activities are dependent on the subsequent degradation of targeted proteins by the 26S proteasome (4). This process is disturbed in many malignancies, some of which link the ubiquitin-proteasome system to tumorigenesis (5).

The attachment of ubiquitin to proteins is highly regulated. One level of regulation is accomplished at the elongation of the polyubiquitin chain. Whereas an interplay of different enzymes is adding ubiquitin, individual proteins, so-called deubiquitinating enzymes, act to counterbalance this reaction and can release ubiquitin from the polyubiquitin chain (6). These cysteine proteases comprise two families, the UBPs,³ also known as USP family and the UCH family. Whereas UCH members preferentially cleave ubiquitin from small adducts such as peptides and amino acids, UBPs target larger molecules, such as proteins. In contrast to the UCH family, the UBP family is highly diverse with >20 mammalian members. Relatively little is known about the cellular role of UPBs.

One class of UBPs is thought to be responsible for the maintenance of the monoubiquitin pool by releasing ubiquitin from free polyubiquitin chains, e.g., UBP14 (7), or from polyubiquitin chains still attached to degraded protein remnants, e.g., UBP4 (6). At the same time these UBPs appear, directly or indirectly, to promote degradation of certain protein substrates (6). On the contrary, another class of UBPs, such as the FAF protein and the PA700 Isopeptidase, is stabilizing proteins presumably by rescuing proteins from proteasome-mediated degradation (6). Importantly, whereas specific substrates of UBPs have been assumed to exist for a long time, only recently p53 was demonstrated to be such a specific target for the ubiquitin-specific protease USP7 (8). This and other substrates of UBPs are thought to be responsible for UBPs to play a role in various biological processes, such as growth control (9, 10, 11, 12) and transcriptional silencing (13).

Here we provide additional evidence for a direct involvement of UBPs in apoptosis. Using a screen for proapoptotic genes, we isolated Ubp41 that was described only recently in chicken (8). On overexpression, Ubp41 is able to induce apoptosis with all of the characteristic features. As the ubiquitin-proteasome system of protein degradation plays an important, but not fully understood, role in cell death (14), we have additionally characterized UBP41 for its involvement in apoptosis induction.

All of the fine chemicals were from Sigma Inc. unless specified otherwise. The proteasome inhibitor MG132 was from Calbiochem, and the pan-caspase inhibitor zVAD-fmk from Enzyme Systems Products.

Human embryonic kidney cells (293T), HeLa cells, and MCF-7 cells were cultured in DMEM (Sigma) supplemented with 5% (293T) or 10% FCS, respectively.

For cloning of hUbp41 (gi: 4759291), hUsp18 (gi: 7159736), and hUsp21 (gi: 7363365), total RNA was isolated from cultured human cells using the RNeasy kit (Qiagen), and cDNA was generated using the Superscript II RT kit (Life Technologies, Inc.). For reverse transcription-PCR, forward primers were used that contained the ATG start codon, a Kozak sequence (CCACC), and the restriction site (HindIII for hUBP41 or KpnI for hUsp18 and hUsp21), whereas the reverse primers were complementary to sequences downstream of the stop codon and contained a NotI restriction site. PCR fragments were cloned into the pcDNA3Δ vector, which was derived from the pcDNA3 vector (Invitrogen) by deletion of the neomycin resistance region. The obtained cDNA sequences were sequenced and compared with the above given GenBank sequence entries. The cDNA sequence obtained for hUbp41 was identical to that predicted by GenBank (gi: 4759291) with exception of the last four amino acids, which proved to be identical not to the human sequence but to that published for the mouse sequence (gi: 3386551). We also identified differences to the GenBank sequences for hUsp18 and hUsp21. Our sequences could be confirmed by cloning and sequencing of cDNA obtained from 293T, HeLa, and Jurkat cells. For construction of COOH-terminally HA-tagged proteins, the Usp sequences were inserted into a pcDNA3Δ vector carrying the HA-tag sequences between the NotI and XhoI restriction site. The active-site mutants hUbp41 C24A, hUsp18 C64A, and hUsp21 C37A were generated by site-directed PCR mutagenesis using overlapping primer pairs covering the intended mutation site and containing the appropriate nucleotide changes. Successful mutagenesis was confirmed by sequencing.

mRNA was isolated from 10-week-old CD1 mice and normalized as described previously (15). The cDNA was subcloned into a modified pcDNA3 vector in which the neomycin resistance gene was deleted. The screen for dominant apoptosis inducers was performed essentially as published (15), except that aliquots containing single bacteria clones were grown up. A novel 96-well DNA isolation method allowed a considerably higher throughput (16).

293T cells were transfected by the calcium phosphate method (17). The transfections were then added to 293T cells in a well of a six-well plate, containing 2 ml of DMEM (5% FCS) to which 2 μl 40 mm of chloroquine had been added 20 min before transfection. The calcium phosphate DNA precipitate was incubated for 6 h on the cells after which the medium was exchanged. Transfection efficiency could be controlled by fluorescence of the cotransfected pEGFP and was determined by FACS analysis to be usually ∼50%. HeLa cells were transfected using the Effectene transfection reagent (Qiagen). The transfection efficiency could be controlled by fluorescence of the cotransfected pEGFP and was controlled by FACS analysis to be generally ∼60%.

cells were harvested by trypsination, resuspended in PBS, and taken up in 3 volumes of hypotonic PI-buffer (20 μg/ml PI, 0.1% w/v Na-citrate, and 0.1% Triton X-100 in PBS). The resulting cell nuclei were analyzed by flow cytometry for sub-G₁ DNA content using the FACScalibur (BD) machine in FL-2 and subsequent evaluation using the CellQuest Software that calculated the percentage of cells containing fragmented DNA. If not indicated differently, the apoptotic cell population was determined by taking into account the percentage of transfected, i.e., GFP-positive cells, which was determined in parallel by FACS analysis in FL-1 using an aliquot of the corresponding cell sample. Each data point usually represents triplicates, and the result was confirmed in several independent experiments.

DNA fragmentation in 293T cells by PI-FACS analysis was determined applying a sensitive protocol, which includes a citrate-phosphate buffer extraction step essentially as described (18). The cells were then analyzed by FACS as described above for PI-FACS analysis of HeLa cells.

The caspase-3 activity assay was purchased from Roche and used as recommended. Each data point was measured at least in triplicates and was confirmed by several independent experiments. The specificity of this assay was confirmed by treating MCF-7 cells that do not express caspase-3 with tumor necrosis factor and doxorubicin. Both reagents were unable to elicit an activity in this assay, whereas both substances were active in HeLa cells.

Isolation and detection of low molecular weight DNA fragments was performed as described (15).

For detecting protein expression, cells were harvested by trypsinization, washed with PBS, and lysed in Triton X buffer [50 mm Tris-HCl (pH 8.0), 150 mm NaCl, and 1% Triton X-100] for 10 min on ice. After centrifugation at 14,000 × g in a microfuge, supernatants were obtained as cytoplasmic extracts, which were quantified for protein content using the Bradford reagent (Bio-Rad). Equal amounts of protein were loaded on a 12% SDS-polyacrylamide gel after boiling with 1 volume of 2× sample buffer [100 mm Tris-HCl (pH 6.8), 4% SDS, 10% mercaptoethanol, 20% glycerol, and 0.05% bromophenol blue], separated in an electric field of approximately 20 V/cm, and transferred to a polyvinylidene difluoride membrane in a semidry blotting device. The membrane was blocked for 1 h with 5% dry milk powder in TBS-Tween [10 mm Tris-HCl (pH 8.0), 150 mm NaCl, and 0.05% Tween] and probed overnight at 4°C with rat monoclonal anti-HA high affinity antibodies (Roche), rabbit polyclonal antiubiquitin antibodies (Sigma), mouse monoclonal anti-p21, p27, or cyclin B1 (all Promega). Membranes were washed three times with 5% dry milk powder in TBS-Tween, and were then probed for 1 h at room temperature with secondary horseradish peroxidase-conjugated antibodies (Roche). After washing four times with TBS-Tween, the blot was developed using Super Signal enhanced chemiluminescence reagents (Pierce) and by exposure to BioMax MR films (Kodak).

For hybridization of the multiple tissue Northern blot (Clontech), the coding sequence of hUbp41 was excised from the expression plasmid and labeled with 5′-[α³²P]dCTP using the RediPrime random prime labeling kit (Amersham). The blot was prehybridized for 3 h at 65°C in 40 ml of hybridization buffer (6× SSC, 5× Denhardt’s reagent, 0.5% SDS, and 0.1 mg/ml salmon sperm DNA),, then 10 ml of a 50% dextrane sulfate solution was added, and finally the labeled probe (∼5 × 10⁷ cpm) was added to the hybridization solution and incubated at 62°C for 16 h. The blot was washed with wash buffer 1 (4× SSC and 0.5% SDS) for 1 h at 62°C and then for 5 min with wash buffer 2 (2× SSC and 0.5% SDS). The blot was exposed to a Fuji BAS 2500 phosphoimager screen.

By applying a direct functional screen for the isolation of dominant apoptosis-inducing genes (15, 19), we identified several mouse cDNAs, which, on overexpression, induce cell death in the human kidney cell line 293T. Among the isolated genes were known apoptosis mediators that serve as internal positive controls for the screen.⁴

One of those cDNA clones not only produced the typical morphological changes of apoptosis but also generated oligonucleosomal DNA fragmentation, a hallmark of apoptosis (Fig. 1). Sequence analysis using BLAST (20) identified this cDNA as the mouse orthologue of the ubiquitin-specific protease UBP41 (gi:7949157; also denominated USP2; Ref. 21), which had been isolated recently in chicken skeletal muscle (22). PCR primers were designed for reverse transcription-PCR of human Ubp41 with the forward primer containing the start ATG and the reverse primer downstream of the stop codon. The obtained PCR product confirmed the published sequence with the exception of a few residues at the extreme COOH terminus, which was determined to be identical to the homologous mouse amino acid sequence (Fig. 2,A). Therefore, the PCR product was regarded to be the human homologue of Ubp41, subsequently designated hUbp41, and its coding sequence was cloned by adding a COOH-terminal HA-tag. A Northern blot comprising different tissue samples revealed a strong expression of Ubp41 in skeletal muscle, heart muscle, and in the kidney (Fig. 2,B). Some USPs are differentially localized in the cell, which might have an impact on their substrate range and, hence, their activity. Therefore, we constructed GFP fusion proteins of wild-type hUBP41. Fluorescence microscopy showed a uniform distribution of hUBP41-GFP in HeLa cells. UBPs contain a catalytic center consisting of a cysteine, an aspartic acid, and a histidine that is highly conserved. Site-directed mutagenesis of the active-site cysteine to alanine generated the enzymatically inactive form, hUBP41-C24A, that was also fused to GFP. The hUBP41-C24A fusion protein with GFP displayed additional spots of higher fluorescence intensity, possibly representing protein aggregates (data not shown). Transfection of HA-tagged hUBP41 and hUBP41-C24A into 293T cells revealed that the human UBP41 (hUBP41-HA) induces apoptosis as judged by phenotype and DNA laddering, whereas the active site mutant did not induce significant cell death (Fig. 3, A and B). hUBP41 also induced apoptosis in the human cervix carcinoma HeLa cells (Fig. 3,C). When we transfected 400 ng of the Ubp41 expression plasmid and scored the cells after 38 h, >40% of apoptotic cells could be observed (Fig. 3,D). DNA fragmentation could be inhibited by the pan-caspase inhibitor zVAD-fmk (Fig. 3 E) and by cotransfection with Bcl-XL (data not shown), supporting an apoptotic mode of cell death.

We isolated the two most homologous human family members of Ubp41: hUsp18 (25% identity to hUPB41) and hUsp21 (45% identity to hUBP41). They were cloned as COOH-terminally HA-tagged proteins. We also constructed their corresponding active site mutants hUSP18-C64A-HA and hUSP21-C37A-HA. Transient transfection into 293T cells demonstrated that only the expression of hUbp41-HA resulted in a significant DNA fragmentation and caspase-3 activity (Fig. 4, A and B), as well as apoptotic phenotype changes (data not shown).

On the basis of the described activity of UBPs to release ubiquitin from the polyubiquitin chain attached to proteins, we performed an antiubiquitin immunoblot analysis against total lysates from 293T cells transfected with hUBP41. They showed the disappearance of a broad range of ubiquitinated protein bands when compared with extracts from control transfected cells, or cells transfected with the homologous genes or the inactive-site mutant hUBP41-C24A (Fig. 5,A). Thus, hUBP41-HA is apparently able to deubiquitinate a broad range of proteins when overexpressed, whereas its homologues hUSP18-HA and hUSP21-HA did not reveal such an activity. The same deubiquitinating ability of hUBP41-HA could be observed in HeLa cells (data not shown). To interfere with the ubiquitination status by other means, we expressed the ubiquitin double-mutant Ub-K48R, G76A in 293T, which did not result in apoptosis induction (Fig. 5,B). This dominant-negative ubiquitin mutant cannot be expected to induce deubiquitination comparable with that observed for hUBP41-HA (Fig. 5 A), but rather has been shown to influence the dynamics of polyubiquitin chain formation by generating shorter chain lengths (23).

Polyubiquitinated proteins are degraded by the 26S proteasome. Therefore, we asked whether the deubiquitinating activity of hUBP41-HA can result in the stabilization of a GFP-fusion protein Ub-G76V-GFP that is NH₂-terminally ubiquitinated and, therefore, rendered unstable (24). In HeLa cells, as shown in Fig. 5 C, hUBP41-HA is able to stabilize this GFP fusion protein. The same effect could also be observed in 293T cells (data not shown).

These data suggest that hUBP41 induces apoptosis by stabilizing a broad range of proteins. However, when we investigated p27, p21, and cyclin B, proteins that have been shown to be substrates of the proteasome (9, 25, 26, 27), we did not detect a differential regulation of the protein level after hUBP41 transfection (Fig. 5D). In contrast, the proteasome inhibitor MG132 led to a marked increase of these proteins. If a broad range inhibition of the degradation of proteins is the cause for apoptosis induction of UBP41, one would expect that the inhibition of the proteasome would also cause apoptosis. The treatment of cell lines with proteasome inhibitors results in apoptosis induction, although in some cell types proteasome inhibitors appear to protect those cells from apoptosis induced by various factors (28). Hence, we tested the proteasome inhibitor MG132. Fig. 5 E shows that this chemical at 1 μm was able to induce a cell cycle arrest in G₂/M, an activity that was not observed with hUBP41. In addition, in comparison with UBP41, MG132 caused apoptosis less efficiently. MG132 at a concentration of 0.25 μm still led to a cell cycle arrest followed by apoptosis induction, with both effects being equally reduced (data not shown).

These results indicate that hUBP41 mediates cellular changes that are distinctly different from a general block of proteasomal activity.

As part of the ubiquitin-proteasome system, deubiquitinating enzymes (UBPs) act by enzymatically shortening or removing the polyubiquitin chain from targeted proteins. This can lead to both the rescue of distinct protein substrates from degradation as well as their targeted destruction (6). As a consequence, a variety of biological processes such as development, growth, and transcription are altered. There is accumulating evidence for an involvement of the ubiquitin-proteasome system in apoptosis by regulating the stability and activity of several factors including the Bcl-2 family of proteins, the IAPs, p53, and inhibitor of nuclear factor-κB, as well as inhibitor of nuclear factor-κB kinase (14). However, thus far there have been only few reports implicating UBPs in apoptotic processes.

We report here that transfection of the ubiquitin-specific protease UBP41 induces cell death in the human cell lines 293T and HeLa. The apoptotic mode of cell death is supported by the detection of oligonucleosomal DNA fragmentation (Fig. 3,A), activation of caspase-3 (Fig. 4,B), and by the fact that cell death can be inhibited by the caspase inhibitor zVAD-fmk (Fig. 3 E).

Thus far, evidence for an involvement of UBPs in apoptosis has been mostly indirect, such as with the fat facets gene (faf) in Drosophila or the related Fam protein in mouse, which both appear to be crucial in development of the fly eye (29) and the interdigital webs (30), respectively. It also has been reported that an active-site mutant of UBP-M leads to apoptosis induction when transfected into mammalian cells (31). Overexpression experiments have also been described with Dub-1 (12), Usp21 (32), and UbpY (10), but in all three of the cases no apoptosis has been observed. Instead, Dub-1 overexpression was reported to result in growth suppression by arresting cells in G₁, and transfected Usp21 or UbpY also caused a profound growth-inhibitory effect arguing for a specific effect of UBPs on the cell. These differences could be explained by UBPs having a specific range of substrates.

In line with this, we found that overexpression of hUsp21-HA and hUsp18-HA, two homologous genes to Ubp41, did not result in apoptosis induction in 293T cells (Fig. 4), although we cannot totally rule out that this is explained by the fact that the expression levels of hUsp18 and hUsp21 were lower than those of hUbp41 (Fig. 4,C). The protein levels of these UBPs could not be raised to the same level as UBP41 by using more plasmid DNA in the transfections, possibly reflecting different protein stabilities (data not shown). Indeed, we recognized that appropriately high expression levels of hUbp41-HA (as shown in Fig. 4) are necessary to obtain a reasonable degree of apoptosis induction in transient transfections. Therefore, if hUBP41-mediated sensitization for apoptosis induction plays a role under physiological or pathological conditions, tissues with strong hUbp41 expression levels such a kidney or skeletal muscle (Fig. 2 B) are most likely to be affected. As stable transfections yield lower expression levels compared with transient transfections, we attempted to generate 293T clone pools stably expressing hUbp41-HA, hUsp18-HA, and hUsp21-HA. However, only in case of hUbp41-HA we were not able to recover any surviving cell clones, suggesting that hUBP41-HA sensitizes cells to apoptosis even at lower protein concentrations.

Expression of hUbp41-HA resulted in dramatic deubiquitination of many substrates (Fig. 5A), which should severely interfere with the ubiquitin-proteasome system. One possible consequence could be the increase in the pool of free monoubiquitin leading to enhanced ubiquitination and degradation of certain proteins. But this appears not to be likely, because in antiubiquitin immunoblots we observed very high basal levels of monoubiquitin in untreated 293T and HeLa cells, which did not recognizably change in hUbp41-HA transfected cells (Fig. 5 A).

Another more likely scenario would be that deubiquitination of a variety of proteins would lead to their rescue from proteasomal degradation and, thus, accumulation of a multitude of factors normally regulated by the ubiquitin-proteasome system. Surprisingly, we found this not to be the case. The cell cycle proteins p21, p27, and cyclin B have been well described to be ubiquitinated and degraded by the proteasome. Accordingly, we observed that these proteins accumulate in the presence of the drug inhibitor of the proteasome, MG132, whereas overexpression of hUbp41-HA did not stabilize those proteins (Fig. 5,D). While we did not find any tested endogenous proteins to be stabilized by hUBP41-HA, we could show the accumulation of a test substrate, an artificially ubiquitinated and, thus, destabilized GFP protein (Fig. 5 C). Therefore, it is likely that despite the apparent broad-range deubiquitination effect of hUBP41-HA, only distinct, usually short-lived substrates are actually stabilized and as proapoptotic factors elicit the observed cell death effect. Notably, after overexpression of Usp21 (32) and UbpY (10) a comparably strong effect on the ubiquitination status of proteins was reported, but in both cases not apoptosis but a growth inhibitory effect was observed. Interestingly, a similar decrease of overall ubiquitination levels and concurrent apoptosis induction has been reported for cells rendered deficient in the ubiquitin-activating enzyme E1 (33). Deficiency in E1 enzyme resulted in accumulation of p21, p27, cyclin D1, and p53, but not Bax. The authors proposed that as yet unidentified short-lived proteins might be stabilized, accumulate, and act as death effectors in the induction of apoptosis. It should be mentioned that E1 deficiency resulted in broad range deubiquitination and apoptosis, but in contrast to hUBP41-HA mediated apoptosis, this type of cell death was caspase-independent (33).

We checked the response of HeLa cells to the proteasome inhibitor MG132 and found that MG132 primarily leads to a cell cycle arrest in G₂/M and presumably only subsequently results in moderate apoptosis induction (Fig. 5,E). In contrast, overexpression of hUbp41-HA did not arrest cells in the cell cycle but seems to drive cells directly into apoptosis (Fig. 5 E). Thus, although a strong interference with the ubiquitin-proteasome system must be assumed on hUbp41-HA overexpression, the resulting cellular changes are clearly distinct from a general inhibition of proteasome activity by the common drug MG132.

Proteasome inhibitors have been introduced recently into clinical studies because of their potential capability to enhance the proapoptotic effect of chemotherapeutical drugs toward otherwise resistant tumor cells (34). We found that expression of Ubp41 induces apoptosis with distinct differences to a common proteasome inhibitor. Hence, apart from drugs that target proteasomes directly, interfering with the ubiquitination status might present an additional pharmacological approach to treat cancer cells (35). Consequently, we think that our findings warrant additional study, as investigating the multiple roles of the ubiquitin-proteasome system in the regulation of apoptosis will give more insight into the complex signaling networks involved but might also lead to novel therapeutic approaches for some human diseases that are linked to dysregulated apoptosis, especially cancer.

We thank Ursula Cramer for invaluable support in the screening procedure and Elisabeth Buergelt for technical support. We also thank N. Dantuma for allowing us to use the Ub-G76V-GFP construct.