Rapid turnover of c-FLIPshort is determined by its unique C-terminal tail.

The caspase-8 inhibitor c-FLIP exists as two splice variants, c-FLIP(L) and c-FLIP(S), with distinct roles in death receptor signaling. The mechanisms determining their turnover have not been established. We found that in differentiating K562 erythroleukemia cells both c-FLIP isoforms were inducibly degraded by the proteasome, but c-FLIP(S) was more prone to ubiquitylation and had a considerably shorter half-life. Analysis of the c-FLIP(S)-specific ubiquitylation revealed two lysines, 192 and 195, C-terminal to the death effector domains, as principal ubiquitin acceptors in c-FLIP(S) but not in c-FLIP(L). Furthermore the c-FLIP(S)-specific tail of 19 amino acids, adjacent to the two target lysines, was demonstrated to be the key element determining the isoform-specific instability of c-FLIP(S). Molecular modeling in combination with site-directed mutagenesis demonstrated that the C-terminal tail is required for correct positioning and subsequent ubiquitylation of the target lysines. Because the antiapoptotic operation of c-FLIP(S) was not affected by the tail deletion, the antiapoptotic activity and ubiquitin-mediated degradation of c-FLIP(S) are functionally and structurally independent processes. The presence of a small destabilizing sequence in c-FLIP(S) constitutes an important determinant of c-FLIP(S)/c-FLIP(L) ratios by allowing differential degradation of c-FLIP isoforms. The conformation-based predisposition of c-FLIP(S) to ubiquitin-mediated degradation introduces a novel concept to the regulation of the death-inducing signaling complex.

The caspase-8 inhibitor c-FLIP exists as two splice variants, c-FLIP L and c-FLIP S , with distinct roles in death receptor signaling. The mechanisms determining their turnover have not been established. We found that in differentiating K562 erythroleukemia cells both c-FLIP isoforms were inducibly degraded by the proteasome, but c-FLIP S was more prone to ubiquitylation and had a considerably shorter half-life. Analysis of the c-FLIP S -specific ubiquitylation revealed two lysines, 192 and 195, C-terminal to the death effector domains, as principal ubiquitin acceptors in c-FLIP S but not in c-FLIP L . Furthermore the c-FLIP S -specific tail of 19 amino acids, adjacent to the two target lysines, was demonstrated to be the key element determining the isoform-specific instability of c-FLIP S . Molecular modeling in combination with site-directed mutagenesis demonstrated that the C-terminal tail is required for correct positioning and subsequent ubiquitylation of the target lysines. Because the antiapoptotic operation of c-FLIP S was not affected by the tail deletion, the antiapoptotic activity and ubiquitin-mediated degradation of c-FLIP S are functionally and structurally independent processes. The presence of a small destabilizing sequence in c-FLIP S constitutes an important determinant of c-FLIP S /c-FLIP L ratios by allowing differential degradation of c-FLIP isoforms. The conformation-based predisposition of c-FLIP S to ubiquitin-mediated degradation introduces a novel concept to the regulation of the death-inducing signaling complex.
Death receptors are cell surface receptors that belong to the tumor necrosis factor receptor superfamily, the most well known members of which are tumor necrosis factor receptor 1, the CD95/Fas receptor, and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) 1 receptors DR4/TRAIL-R1 and DR5/TRAIL-R2 (for a review, see Ref. 1). Upon ligand-mediated oligomerization of CD95/Fas or TRAIL receptors, the Fas-associated death domain protein (FADD) attaches to the death receptor via homophilic death domain interactions. The death effector domain (DED) of FADD is in turn linked to the apoptotic machinery due to its affinity for the initiator caspases, procaspase-8 and procaspase-10. The close proximity of procaspase-8 molecules results in dimerization of the procaspases, the assembly of which forms an enzymatically active site (2,3). This enzymatic activity induces serial cleavages converting the zymogen into p10 and p18 fragments, forming the proteolytically active caspase-8 heterotetramer. In responsive cells, activated caspase-8 is then able to activate effector caspases thereby initiating apoptosis (for a review, see Ref. 4).
In addition to the initiator caspases, the DED in FADD also interacts with the apoptosis modulator cellular FLICE-inhibitory protein (c-FLIP; Ref. 5), which exists as two splice variants, the long splice form of FLIP (FLIP L ) and the short splice form of FLIP (FLIP S ). Together with the activated death receptor, FADD, and caspase-8 and -10, the FLIP proteins form the core of the death-inducing signaling complex (DISC; Ref. 6), which also contains a number of other proteins, many of which seem to be involved in modulating death receptor signals (for a review, see Ref. 1). The DISC assembly with pro-and antiapoptotic proteins allows for control of death receptor signaling through a number of different routes. Consequently in addition to apoptosis, stimulation of the receptors may also lead to cell survival and proliferation. Modulation of the apoptotic pathway at the level of the activated receptor is required in situations when the extrinsic, but not the intrinsic, apoptosis pathway needs to be inhibited or when localized initiator caspase activity is required for specialized signaling functions. For example, caspase-8 activity is necessary for normal T cell development (7,8) and is probably acquired through complex regulation of anti-and proapoptotic proteins in the DISC. In addition, some malignant cells are able to convert death receptor stimulation into proliferative signals (9,10). Therefore, active regulation of the DISC proteins, both transcriptional and post-translational, determines the outcome of death receptor stimulation.
c-FLIP L and c-FLIP S have been characterized as specific inhibitors of death receptor-mediated apoptosis (for a review, see Ref. 11). c-FLIP L is homologous to caspase-8, consisting of two tandemly repeated DEDs and a catalytically inactive caspase-like domain. Although c-FLIP S shares most of its se-quence with c-FLIP L , it is considerably shorter, comprising only the two DEDs followed by a short C-terminal sequence and an isoform-specific C-terminal tail of 19 amino acids, the role of which is still unknown (for a review, see Ref. 11). Both c-FLIP isoforms act as inhibitors of caspase-8-mediated apoptosis through binding to DED in FADD (12)(13)(14). However, in addition to the well documented antiapoptotic activity of c-FLIP L and c-FLIP S , recent reports have provided evidence for isoformspecific regulatory functions. For example, c-FLIP L is capable of inducing the first cleavage of caspase-8 and itself, thereby leading to membrane-restricted caspase-8 activation, whereas c-FLIP S completely prevents procaspase-8 cleavage (12,13,15). Moreover, c-FLIP L , but not c-FLIP S , interacts with TRAF2 (tumor necrosis factor receptor-associated factor 2) and induces activation of the NF-B signaling pathway (16). Thus, although first considered to be similar inhibitor proteins, the c-FLIP isoforms have now been established as DISC molecules with distinct and even opposite functions.
During development and differentiation it is of great importance for cells to be able to regulate their apoptotic sensitivity. The ability of c-FLIP L and c-FLIP S to modulate caspase-8 activity makes them highly adaptable regulators of death receptor signaling. Consequently disturbances in c-FLIP expression have been implicated in certain malignancies. For example, high levels of c-FLIP have been found in some cancer cells, including melanoma and colonic adenocarcinoma (5,17,18), and an elevated expression of c-FLIP has been shown to result in the escape of tumors from immune surveillance (19,20). c-FLIP expression is carefully regulated at different levels. The transcriptional regulation is linked to a number of growthand survival-promoting signaling pathways, including NF-B (21,22), mitogen-activated protein kinase/extracellular signalregulated kinase (23), and Akt (24,25). In addition to gene expression, it has been recently reported that the turnover of c-FLIP is actively regulated by ubiquitin-mediated degradation (26,27). However, very little is known about the mechanisms underlying this regulation or whether there could be differences in isoform stability. In this study, we show that although both c-FLIP isoforms can be degraded via a common inducible pathway, they display clearly distinct half-lives. Our mutational analyses showed that the principal ubiquitin acceptor lysines are different in c-FLIP L and c-FLIP S and that the unique C terminus of c-FLIP S possesses a destabilizing function. In addition, molecular modeling of the C-terminal segment of c-FLIP S revealed structural features compatible with its preferred ubiquitylation. Although c-FLIP S lacking the C terminus was equally efficient in suppressing TRAIL-induced apoptosis as the wild type protein, the antiapoptotic activity

EXPERIMENTAL PROCEDURES
Cell Culture and Treatments-Human K562 erythroleukemia cells and WM35 melanoma cells were cultured in a humidified 5% CO 2 atmosphere at 37°C in RPMI 1640 medium supplemented with 10% fetal calf serum, antibiotics (penicillin and streptomycin), and 2 mM L-glutamine. For WM35 cells, the medium was supplemented with 5 g/ml insulin. K562 cells stably overexpressing c-FLIP S and c-FLIP L isoforms (2G11 and 1E5, respectively; Ref. 12) were maintained in RPMI 1640 medium containing G418 (500 g/ml, Calbiochem). HeLa cervical carcinoma cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum with antibiotics (penicillin and streptomycin) and 2 mM Lglutamine. The proteasome inhibitor epoxomicin (Calbiochem) was used at 200 nM for 14 h. The protein synthesis inhibitor cycloheximide (Sigma) was used at either 5 or 50 M concentration for the indicated time periods. For K562 cells, hemin (Sigma) was added to a final concentration of 30 or 40 M. For WM35 cells, cisplatin (Sigma) was added to a final concentration of 7 g/ml. Apoptosis was induced by adding 100 ng/ml FLAG-tagged TRAIL (Alexis) together with 2 g/ml cross-linking M2 anti-FLAG antibody (Sigma).

SDS-PAGE and Western
Blotting-For Western blot analysis, cells were harvested by centrifugation and washed once with PBS. Cells were lysed either in the Laemmli SDS sample buffer or in lysis buffer (30 mM Tris, pH 7.5, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, Complete miniprotease inhibitor mixture (Roche Applied Science)). Resulting Triton X-100 lysis buffer lysates were centrifuged to remove insoluble material, and the protein concentrations were determined by Bradford assay. Each lysate containing 30 -50 g of protein was loaded and resolved by SDS-PAGE and transferred to nitrocellulose membrane (Protran nitrocellulose, Schleicher & Schuell) by using a semidry transfer apparatus (Bio-Rad). Western blotting was performed using antibodies against c-FLIP (NF6 FLIP antibody, kindly provided by Peter Krammer, German Cancer Research Center, Heidelberg, Germany; also available from Alexis) and Hsc70 (SPA-815, StressGen). Horseradish peroxidase-conjugated secondary antibodies were purchased from Amersham Biosciences and Southern Biotechnology. The bands were visualized using the enhanced chemiluminescence method (ECL, Amersham Biosciences).
Plasmid Constructs-The FLAG-tagged c-FLIP L and c-FLIP S were a kind gift from Dr. Jü rg Tschopp (Institute of Biochemistry, University of Lausanne, Lausanne, Switzerland). c-FLIP S and c-FLIP L point mutations were made using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by sequencing. The FLAG-tagged c-FLIP S deletion mutant was constructed by PCR and cloned into the EcoRI and XhoI sites in-frame with the N-terminal FLAG tag in pCR3-Met-Flag. The HA-tagged ubiquitin was a kind gift from Dr. Dirk Bohmann (University of Rochester, Rochester, NY).
Transient Transfections and Stable Cell Lines-For transfections, 5ϫ10 6 K562 cells were centrifuged and resuspended in 0.4 ml of Opti-MEM (Invitrogen), and 20 or 30 g of plasmid DNA was added. The expression levels of the mutants were titrated to be comparable because expression levels were noticed to affect ubiquitylation. Cells were subjected to a single electric pulse (220 V, 975 microfarads), in 0.4-cm gap electroporation cuvettes (BTX) using a Bio-Rad Gene Pulser electroporator followed by dilution to 5 ϫ 10 5 cells/ml in RPMI 1640 medium with 10% fetal calf serum and antibiotics. Cells were incubated at 37°C for 24 -36 h prior to the experimental treatments. The stable neomycin-resistant c-FLIP S ⌬203-221 cell lines were selected by G418 (500 g/ml, Calbiochem) for 2 weeks, the resistant pool was serially diluted on a 96-well plate in the presence of G418, and the single cell clones were upscaled and screened for c-FLIP expression by Western blotting.
Immunoprecipitation-For immunoprecipitation of ubiquitylated c-FLIP, the cell pellet from transiently transfected cells was resuspended in 75 l of boiling 1% SDS in PBS, and the resulting lysate was heated at 100°C for 5 min. The lysates were suspended 1:10 in 1% Triton X-100 in PBS. DNA was sheared by sonication, and the particulate material was centrifuged for 15 min at 15,000 ϫ g. Samples were taken from the cleared lysates for input control. The lysates were further diluted 1:1 with 1% Triton X-100, 0.5% bovine serum albumin in PBS and incubated with anti-HA (Santa Cruz Biotechnology) antibody and 15 l of a 50% slurry of protein G-Sepharose under rotation for 2 h. After incubation, the Sepharose beads were washed four times with 1% Triton X-100 in PBS, and the immunoprecipitated proteins were run on an 8 or 10% SDS-polyacrylamide gel, transferred to nitrocellulose membrane

FIG. 2. c-FLIP S has a shorter halflife than c-FLIP L . Western blot analysis of c-FLIP protein in K562 (A) and HeLa (C) cells treated with cycloheximide (CHX, 5 M) for indicated time periods is
shown. Samples from 1% Triton X-100 lysates are shown. Hsc70 was used as a loading control. The level of c-FLIP isoforms after cycloheximide treatment of K562 (B) and HeLa (D) cells was quantitated from two to four different experiments and is shown relative to the levels of c-FLIP in untreated cells. E, the cisplatin (CISP.)-mediated down-regulation of c-FLIP S is prevented by the proteasome inhibitor epoxomicin (EPOX). WM35 cells were treated with 7 g/ml cisplatin for 12 h followed by treatment with 200 nM epoxomicin for 4 h. c-FLIP protein was detected with Western blot analysis. Samples from 1% Triton X-100 lysates are shown. Hsc70 was used as a loading control. F, Western blot analysis of c-FLIP protein in WM35 cells treated with cycloheximide (5 M) for the indicated time periods. Samples from 1% Triton X-100 lysates are shown. Hsc70 was used as a loading control. G, the level of c-FLIP isoforms after cycloheximide treatment of WM35 cells was quantitated and is shown relative to the levels of c-FLIP in untreated cells. A representative result from three independent experiments is shown.
Fluorescence-activated Cell Sorter Analysis of Cellular Caspase-3 Activity-After the treatments, the cells were washed once with ice-cold PBS, and the caspase-3 activity was analyzed with phycoerythrinconjugated monoclonal active caspase-3 antibody apoptosis kit 1 (BD Pharmingen) according to the manufacturer's protocol.
TRAIL-R Immunoprecipitation and DISC Analysis-To stimulate TRAIL receptors, 4 ϫ 10 7 K562 cells/sample were pelleted (500 ϫ g for 7 min) and resuspended in 1 ml of prewarmed RPMI 1640 medium. Thereafter 1 g of FLAG-tagged TRAIL (Alexis) and 2 g of anti-FLAG monoclonal M2 antibody (Sigma) were added to the cell suspension. The cells were incubated at 37°C for 20 min, and the reaction was stopped by adding 10 ml of ice-cold PBS. After washing, the cells were lysed in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10% glycerol, 0.2% Nonidet P40, and Complete protease inhibitor mixture (Roche Applied Science)) for 30 min on ice. The cell debris was removed by centrifugation at 15,000 ϫ g for 15 min at 4°C. The amount of protein was determined by the Bradford assay, and an equal amount of protein from each sample was precleared with 50 l of Sepharose CL-4B for 2 h at 4°C. A total of 2.5 g of monoclonal anti-DR5 and 2.5 g of monoclonal anti-DR4 (Alexis) were added to samples and immunoprecipitated with 15 l of protein G beads (Amersham Biosciences) for 2.5 h at 4°C. The beads were washed six times with 1 ml of lysis buffer, resuspended in 3ϫ Laemmli sample buffer, and finally boiled for 3 min. The immunoprecipitated samples and corresponding cell lysates were analyzed by 11% SDS-PAGE. Western blot analysis was performed with anti-DR5 (Alexis), anti-FADD (BD Transduction Laboratories), caspase-8 (C15 caspase-8 antibody, a kind gift from P. Krammer, German Cancer Research Center; also available from Alexis), and anti-FLIP (Dave-2, Alexis) as described above.
Molecular Modeling of c-FLIP S -All models were built using the Modeler computer program (salilab.org/modeler/modeler.html). The sequence of c-FLIP S contains two conservative DEDs between Ser 2 (numbering according to Ref  (35), and nnpredict (36). The loop between the DED2 and the c-FLIP S C-terminal segment has been taken as seen in the structure of PEA-15, the only currently known representative structure covering this region (28). Surface construction and lipophilic potential calculations were done using SYBYL (Tripos Inc., St. Louis, MO). Fig. 8 was produced with MolScript v2.1 (37) and Raster3D v2.4b (38).

Both c-FLIP Isoforms Are Degraded via the Ubiquitin-Proteasome Pathway-
The antiapoptotic functions of the c-FLIP isoforms are tightly associated with their expression levels, and down-regulation of c-FLIP is an important mechanism to sensitize cells to receptor-mediated apoptosis (for a review, see Ref. 11). Previously we have shown that both c-FLIP isoforms are down-regulated in K562 cells undergoing hemin-mediated erythroid differentiation, which sensitizes the cells to TRAILmediated apoptosis (12). These results prompted us to examine whether the inducible c-FLIP down-regulation is due to proteasome-mediated degradation. For this purpose, we incubated differentiating K562 cells with the proteasome inhibitor epoxomicin, which efficiently prevented hemin-induced downregulation of c-FLIP L and c-FLIP S (Fig. 1A). This suggested that the down-regulation of c-FLIP isoforms would be due to degradation through the ubiquitin-proteasome pathway.
Because the 26 S proteasome acts on proteins destined to be degraded by polyubiquitin conjugation, we investigated whether c-FLIP isoforms are inducibly ubiquitylated upon hemin treatment. K562 cell lines 1E5 and 2G11 (12), expressing ectopic FLAG-tagged c-FLIP S and c-FLIP L , respectively, were treated with hemin and epoxomicin. The hemin treatment did not result in down-regulation of the ectopically expressed c-FLIP in the tested cell lines, indicating that the degradation machinery might be saturated. Hemin treatment, however, caused a prominent smear above the FLAG-FLIP S band that was further enhanced by epoxomicin, whereas in FLAG-FLIP Lexpressing cells, a prominent slower migrating smear was induced only by treating the cells with both hemin and epoxomicin (Fig. 1B). To verify that the high molecular weight smear indeed represented hemin-induced polyubiquitin conjugation, we transiently coexpressed HA-tagged ubiquitin and FLAGtagged c-FLIP. The transfected cells were treated with 30 M hemin for 16 h, and ubiquitin conjugates were immunoprecipitated with HA antibodies. Western blotting with c-FLIP-specific antibodies revealed the ubiquitylation of c-FLIP S and c-FLIP L , which was further enhanced by hemin-mediated differentiation (Fig. 1C). In conclusion, these results show that both c-FLIP isoforms are degraded by the ubiquitin-proteasome pathway, a process that can be activated through a common signal.
c-FLIP S Has a Markedly Shorter Half-life than c-FLIP L -Although both c-FLIP isoforms were inducibly ubiquitylated upon hemin treatment, the ubiquitylation of c-FLIP S appeared to be more prominent (Fig. 1B). To study whether the degradation of the c-FLIP isoforms is differentially regulated, we compared the half-lives of c-FLIP L and c-FLIP S by treating K562 cells with the protein synthesis inhibitor cycloheximide for the indicated time periods. Despite the similarity of the isoforms, c-FLIP S appeared to have a markedly shorter half-life than c-FLIP L , which does not totally disappear even after 24 h of cycloheximide treatment ( Fig. 2A). This result was verified by a quantitation of c-FLIP protein levels, which showed that the half-life of c-FLIP S is ϳ40 min, whereas the half-life of c-FLIP L is almost 2 h (Fig. 2B). This differential stability of the c-FLIP isoforms is not limited only to K562 cells as a similar difference in the half-lives could be observed in HeLa cells (Fig.  2, C and D). To test whether the different half-lives of the c-FLIP isoforms could contribute to changes in their stoichiometry also in other physiologically relevant cellular backgrounds, we studied WM35 melanoma cells, which have been shown earlier to become sensitized to TRAIL upon cisplatin treatment by specific down-regulation of c-FLIP S (39). In agreement with the earlier study (39), cisplatin treatment downregulated c-FLIP S more efficiently than c-FLIP L (Fig. 2E). Notably this down-regulation could be completely prevented by the proteasome inhibitor epoxomicin (Fig. 2E), indicating that proteasomal degradation is needed for the specific down-regulation of c-FLIP S . Moreover cycloheximide chase experiments revealed that also in WM35 melanoma cells c-FLIP S is significantly less stable than c-FLIP L (Fig. 2, F and G). Taken together, these findings imply that the stability of c-FLIP isoforms is regulated in an isoform-specific manner and that the differential stability can contribute to an altered stoichiometry between the isoforms.
The Unique C-terminal Tail of c-FLIP S Is Indispensable for Ubiquitylation-Both c-FLIP isoforms have a common N terminus consisting of two tandemly repeated DEDs. In c-FLIP L , the DEDs are followed by a caspase-like domain, whereas c-FLIP S has a unique C-terminal tail sequence composed of 19 amino acids (Fig. 3A). To investigate whether the unique C-terminal tail of c-FLIP S contributes to its preferred ubiquitylation, the ⌬203-221 mutant lacking the c-FLIP S -specific tail was transiently transfected together with HA-tagged ubiquitin into K562 cells. The effects were analyzed by Western blotting of HA-immunoprecipitated samples with a c-FLIP antibody. Intriguingly whereas wild type c-FLIP S was prominently ubiquitylated, the ⌬203-221 deletion mutant exhibited no detectable ubiquitin conjugation (Fig. 3B), demonstrating that the unique C terminus is indispensable for the ubiquitylation of c-FLIP S . The effect was not cell type-specific because when the ⌬203-221 mutant was transfected into HeLa cells we could not observe any ubiquitylation of the mutant protein (data not shown).
c-FLIP S C Terminus Is Not Needed for Antiapoptotic Activity-Although it is well known that c-FLIP S is recruited to the DISC and inhibits procaspase-8 activation, the molecular mechanisms underlying this function are not finally established. As it has been demonstrated that ubiquitylation of the Drosophila inhibitor of apoptosis (DIAP1) is needed for its antiapoptotic activity (40), we wanted to examine whether ubiquitylation per se is required for the antiapoptotic function of c-FLIP S . To this end, we generated stable cell lines overexpressing the non-ubiquitylatable c-FLIP S ⌬203-221 deletion mutant and analyzed their sensitivity to TRAIL during hemin treatment (Fig. 4, A and B). Similarly to cells overexpressing wild type c-FLIP S , all cell lines expressing the c-FLIP S ⌬203-  ⌬203-221 (5F4). G stands for sample with lysate, isotype-specific antibody, and protein G-beads; C stands for control; and T stands for TRAIL-treated sample. The cells were treated with TRAIL and M2 for 20 min and then lysed. TRAIL-R DISC was immunoprecipitated by using protein G-Sepharose and DR4-and DR5-specific antibodies. The specificity of immunoprecipitation was controlled by incubating the sample with isotype-matching antibody. The immunoprecipitates were analyzed by Western blotting using anti-FADD, anti-caspase-8, anti-c-FLIP, and anti-DR5 antibodies. The presence of the proteins in cell lysates before immunoprecipitation is shown in the right-hand panels. The migration positions of the proteins are indicated. IP, immunoprecipitate. 221 deletion mutant were strongly protected against TRAILinduced apoptosis after hemin treatment (Fig. 4B). This result demonstrates that the ⌬203-221 mutant harbors antiapoptotic activity similar to c-FLIP S and that the unique c-FLIP S C terminus is purely a determinant of ubiquitylation. It should be noted that, in contrast to endogenous c-FLIP S , ectopically expressed wild type c-FLIP S is not significantly down-regulated upon hemin treatment (Fig. 1B and Ref. 12), possibly due to saturation of the degradation machinery.
DISC immunoprecipitation analyses of the 5F4 clone showed that c-FLIP S ⌬203-221 was avidly recruited to the DISC, and it inhibited procaspase-8 cleavage similarly to wild type c-FLIP S (Fig. 4C). These observations also validated the results from the ubiquitylation experiments as the mutant protein was obviously fully functional, implying that the decreased ubiquitylation of the c-FLIP S ⌬203-221 deletion mutant is not due to abnormal protein conformation, localization, or other problems that could arise from aberrant functions of a mutated protein.
Taken together, this analysis shows that the antiapoptotic activity and the regulation of ubiquitylation of c-FLIP S are two distinct functions that reside at physically separated domains. The far C terminus of c-FLIP S is a determinant of ubiquitylation with no direct effects on the antiapoptotic properties of c-FLIP S .
Lys 192 and Lys 195 Are the Principal Ubiquitin Acceptors in c-FLIP S but Not in c-FLIP L -One possible explanation for the absolute requirement of the c-FLIP S -specific C terminus for its ubiquitylation could be that the tail contains the target lysine residue(s) for ubiquitin. The only lysine (Lys 214 ) in the unique C terminus of c-FLIP S (Fig. 5A) was mutated to arginine to examine whether the mutation of this potential target residue interferes with ubiquitylation of c-FLIP S . As shown in Fig. 5B, the K214R mutation did not reduce ubiquitylation of c-FLIP S , demonstrating that the unique C terminus is not a direct target for ubiquitin conjugation.
For mapping the sites of ubiquitin conjugation in c-FLIP S , we mutated all lysine residues in c-FLIP S in four subsequent FIG. 5. Lys 192 and Lys 195 are ubiquitylated in c-FLIP S but not in c-FLIP L . A, schematic presentation of the lysine residues in c-FLIP S . To find out the target residues for ubiquitylation, in addition to single, double, and triple mutants, four lysine to arginine cluster mutants were used. The mutants were named accordingly. B, ubiquitylation of the K214R mutant was analyzed by transient transfection and HA immunoprecipitation as described for Fig. 1C. Ubiquitylated c-FLIP was detected by Western blotting as shown in the upper panel. The presence of proteins in the cell lysates before immunoprecipitation is shown in the lower panel. Hsc70 was used as a loading control. C, the ubiquitylation of the cluster mutants with and without hemin treatment (30 M for 16 h) was analyzed by transient transfection and HA immunoprecipitation as described above. Ubiquitylated c-FLIP was detected by Western blotting in the middle panels. The presence of proteins in cell lysates before immunoprecipitation is shown in the lower panels. M stands for mock-transfected cells. Hsc70 was used as a loading control. D, the ubiquitylation of the c-FLIP S K192R,K195R mutant was analyzed and controlled as in Fig. 1C. E, the ubiquitylation of the c-FLIP L K192R,K195R mutant was analyzed and controlled as for Fig. 1C (Fig. 5A). Surprisingly three of the cluster mutants (DED1, DED2A, and DED2B) had little or no effect on ubiquitylation of c-FLIP S (Fig. 5C). In contrast, in the TAIL mutant, carrying only three C-terminal lysine residues (192,195, and 214) mutated to arginines, the ubiquitin conjugation was efficiently, albeit not fully, inhibited (Fig. 5C). As the K214R mutation on its own did not affect ubiquitylation of c-FLIP S , it seemed likely that Lys 192 and Lys 195 would be the principal ubiquitin acceptors in c-FLIP S . Indeed the K192R,K195R double mutant displayed a clearly reduced level of ubiquitylation (Fig. 5D). As Lys 192 and Lys 195 are also present in c-FLIP L , we examined whether the same lysines are targets for ubiquitylation in c-FLIP L . However, the c-FLIP L K192R,K195R double mutant was ubiquitylated at a level similar to the wild type (Fig. 5E), indicating that the major ubiquitin target residues are distinct in c-FLIP S and c-FLIP L . These results provide direct evidence for differential regulation of ubiquitin conjugation in the two c-FLIP isoforms.
C Terminus of c-FLIP S Is Indispensable for Its Degradation-As ubiquitylation has recently been implicated in processes not related to proteasomal degradation (for a review, see Ref. 41), we examined whether the decreased ubiquitylation of the ⌬203-221 mutant would be manifested as an increased half-life of the mutant protein as compared with the wild type c-FLIP S . Due to a stabilizing effect of c-FLIP S overexpression, the expression levels of both constructs were titrated close to the endogenous level, and cells were treated with cycloheximide for the indicated time periods (Figs. 6, A and B). As expected, the tail deletion ⌬203-221 strongly stabilized c-FLIP S , demonstrating that the unique C terminus is indispensable for the ubiquitin-mediated degradation of the less stable c-FLIP isoform. It should be noted that in all experiments the stability of ⌬203-221 mutant was substantially increased when compared with wild type c-FLIP S even if the half-lives of exogenously expressed proteins varied between experiments (most likely due to differences in protein expression levels and transfection efficiency). We also observed that the stability of the ⌬203-221 mutant was elevated in the WM35 melanoma cell line as indicated by consistently higher expression levels of the mutant protein (data not shown).
Structural Analysis of the c-FLIP S C Terminus-To better understand the contribution of the unique C-terminal tail to the turnover and the isoform-specific function of the lysine residues Lys 192 and Lys 195 , we analyzed the structural features of the C-terminal segment (amino acids 176 -221) of c-FLIP S containing the critical determinants for the specific regulation of c-FLIP S . Amino acids 176 -202 of c-FLIP S are identical in sequence with the c-FLIP L isoform, and residues Tyr 182 -Asp 196 are strongly predicted by five independent secondary structure prediction computer programs to have an ␣-helical secondary structure (Fig. 7). To search for possible sites of interaction between the ␣-helix and the DED2, we constructed a molecular model of the DED2 based on the known DED structures of PEA-15 (Protein Data Bank code 1n3k) and the FADD death effector domain (Protein Data Bank code 1a1w). The loop region Ala 178 -Arg 183 , located between the predicted ␣-helix and the DED2, is very flexible in the NMR structure for PEA-15 (28). Thus, the conformation of this loop does not define the proper positioning of the predicted ␣-helix relative to the DED2. According to the model presented in Fig. 8A, both lysines, Lys 192 and Lys 195 , reside along the same, polar side of the predicted amphipathic ␣-helix; the opposite face of the helix is hydrophobic (Fig. 8A, gray spheres), strongly suggesting that the hydrophobic face of the helix must interact with hydrophobic regions on the DED2 or with some other structure. In the model structure, Lys 192 and Lys 195 are accessible to solvent and could accept polyubiquitin chains equally well.
To further elucidate the functional motifs in c-FLIP S that could be involved in regulating isoform-specific turnover, we analyzed the hydrophobic/hydrophilic characteristics of the surface of the modeled structure of c-FLIP S . There are two lipophilic clusters along the C-terminal end of FLIP S (Fig. 8B). Lipophilic surface I, including residues Tyr 182 , Leu 186 , Ile 190 , and Leu 194 , forms the hydrophobic face of the ␣-helix described above. A second prominent hydrophobic surface, lipophilic surface II, is formed by residues Phe 201 , Met 203 , Tyr 207 , and Cys 210 ; Met 203 , Tyr 207 , and Cys 210 are unique to c-FLIP S . The surface of the DED2 has two hydrophobic grooves (Fig. 8C). sponding to the lipophilic surface II are not hydrophobic but are charged (Arg 207 and Glu 210 ), making it unlikely that c-FLIP L would adopt a structure similar to that predicted for c-FLIP S . The molecular modeling together with our experimental data strongly suggests that the structural differences in the region composed of amino acids 176 -221 determines the differential ubiquitin-mediated degradation of the c-FLIP isoforms.
The ␣-Helical Region of c-FLIP S Is Required for Its Efficient Ubiquitylation and Fast Turnover-As the molecular model suggested that the ␣-helix and its proper positioning may de- , which directly follows the C-terminal ␣-helix. In C two lipophilic grooves (shown as "Lipophilic Groove I" and "Lipophilic Groove II") are visible (dotted lines) on the surface of the DED2. The two lipophilic grooves are exposed to solvent and could accommodate the two lipophilic surfaces on the C-terminal tail, thus positioning Lys 192 and Lys 195 on the solvent-exposed surface of c-FLIP S . termine the c-FLIP S ubiquitylation, we generated c-FLIP S mutants in which the ␣-helix was either deleted (⌬182-196) or three of the critical hydrophobic residues were mutated (3xmut, Y182S,L186R,I190N) to disrupt lipophilic surface I. Ubiquitylation assays clearly demonstrated that the ␣-helix is required for c-FLIP S ubiquitylation as both mutants displayed significantly reduced ubiquitin conjugation (Fig. 9A). Moreover when the stability of the mutants was investigated, they showed increased stability when compared with wild type c-FLIP S (Fig. 9, B and C). In conclusion, as suggested by molecular modeling, the ␣-helical region of c-FLIP S is critical for the efficient ubiquitylation and for its fast turnover. DISCUSSION c-FLIP S and c-FLIP L operate at a focal point in the death receptor pathway by acting as specific gatekeepers of death receptor signaling (for a review, see Ref. 11). Although the importance of gene expression in regulating apoptotic signal transduction (both intrinsic and extrinsic) has been emphasized in numerous studies, post-translational modifications, especially phosphorylation and ubiquitylation, have recently surfaced as important regulators of proteins in the death receptor pathway (for reviews, see Refs. 42 and 43). Ubiquitinmediated degradation of the c-FLIP isoforms has been implicated under different circumstances, including p53 activation, peroxisome proliferator-activated receptor ␥ ligand administration, and adenoviral infection (26,27,44). However, the mechanisms underlying the ubiquitylation of c-FLIP S and c-FLIP L and possibilities for isoform-specific regulation have not been characterized. Our analysis showed that, although both c-FLIP isoforms are actively degraded via the ubiquitin-proteasomal pathway, their regulation is isoform-specific. Moreover the mutational analyses provide a putative molecular mechanism for the preferred degradation of c-FLIP S . The DISC analysis showed that although the C-terminal tail region of c-FLIP S is needed for its degradation it is not required for the antiapoptotic function. This result implies that the sequence elements regulating the antiapoptotic function and the stability of c-FLIP S are strictly separated.
We demonstrate that proteasomal degradation can be used to rapidly and specifically regulate the stoichiometry of the c-FLIP isoforms. This newly characterized isoform-specific regulatory feature is well in accordance with the diverging roles of c-FLIP L and c-FLIP S . The c-FLIP isoforms have been perceived primarily as caspase-8 inhibitors with a direct relationship between c-FLIP protein levels and death receptor sensitivity (for a review, see Ref. 11). However, there are a number of reports demonstrating that c-FLIP L and c-FLIP S have different molecular modes of action (15,45). Initially both c-FLIP isoforms were believed to simply compete with procaspase-8 binding to the DISC, thereby inhibiting caspase-8 activation. Recently several reports have indicated that rather than preventing procaspase-8 recruitment to the DISC, c-FLIP has isoform-specific effects on the cleavage-dependent activation of procaspase-8 (13,14). For example, whereas c-FLIP S completely prevents the proximity-induced cleavage of procaspase-8, c-FLIP L induces the first cleavage of caspase-8 to the p43/41 form but prevents the further cleavage of caspase-8 (15,45). A differential regulation may also be required for the diverging roles of c-FLIP isoforms in conveying signals from antiapoptotic signaling pathways; for example, in response to CD3 stimulation, NF-B is preferentially activated by c-FLIP L , and mitogen-activated protein kinase/extracellular signal-regulated kinase is preferentially activated by c-FLIP S (46). The region critical for the activation of NF-B signaling resides in the caspase-8-cleaved p43 fragment of c-FLIP L . This fragment specifically interacts with TRAF2 and induces NF-B activation (16). Given the striking functional differences between the two c-FLIP isoforms, the regulation of their expression will be an important determinant for cells to adjust their apoptotic threshold and responsiveness to extracellular signals.
Our discovery of the shorter half-life of c-FLIP S compared with that of c-FLIP L exposes a versatile regulatory feature of this protein pair that cells can use to regulate their death receptor signals. In the cell models that we used, c-FLIP S was the more actively regulated isoform. However, given suitable preconditions, this type of regulation could be used to reverse the ratio between the c-FLIP isoforms. Accordingly in SHEP neuroblastoma cells, the half-life of c-FLIP L has been shown to be shorter than that of c-FLIP S (47), suggesting that the halflife of c-FLIP isoforms can also be regulated in a cell typespecific manner. As c-FLIP S was less stable in our cell models, we wanted to unravel the molecular determinants that are responsible for triggering the ubiquitylation-mediated degra- dation of c-FLIP S . In many cases, ubiquitylation is primed by phosphorylation of the substrate protein. We tested this possibility because c-FLIP has been reported to be phosphorylated (48,49). Although c-FLIP S showed serine-directed phosphorylation, it was not associated with the inducible ubiquitylation of c-FLIP S (data not shown). Therefore, the triggering mechanisms for ubiquitylation remain to be determined.
The shorter half-life and preferred ubiquitylation of c-FLIP S are clearly determined by its unique C-terminal tail (amino acids 203-221), which does not act by providing a target lysine (Lys 214 ) but is instead required for correct positioning and stabilization of the ␣-helix that contains the principal target lysine residues, 192 and 195, on the surface of the DED2. Thus, the c-FLIP S -specific tail is the determinant that favors the ubiquitylation of Lys 192 and Lys 195 in c-FLIP S but not in c-FLIP L . It is, however, notable that despite the decreased ubiquitylation of c-FLIP S in the K192R,K195R mutant lacking the principal ubiquitin acceptor sites, this mutant did not display a significantly longer half-life when compared with wild type c-FLIP S (data not shown). It is likely that the residual ubiquitylation observed in the K192R,K195R mutant is enough to target the molecule to the proteasome. It has also been shown previously that ubiquitylation of some substrates may lack specificity so that replacement of one lysine with arginine leads to ubiquitylation of another lysine (for a review, see Ref. 50). On the other hand, the C-terminal deletion mutant showed a marked effect on both ubiquitylation and degradation, and this mutant appeared to adopt degradation kinetics that were close to that of c-FLIP L . It is interesting to speculate that c-FLIP L and c-FLIP S could be regulated by two distinct conjugation machineries, one shared and one specific for c-FLIP S , thereby yielding its faster turnover.
To elucidate the mechanisms for the preferred susceptibility of c-FLIP S to ubiquitylation, we combined our experimental data with molecular modeling. The molecular models imply that the overall structure of c-FLIP S is indeed regulated by the unique C-terminal tail, possibly leading to better availability of the ubiquitin target Lys 192 and Lys 195 . The C-terminal part of c-FLIP S includes two prominent lipophilic surfaces that can be accommodated along two hydrophobic grooves present on the surface of the DED2. Moreover the site-directed mutagenesis with subsequent analyses confirmed the critical role of the lipophilic surfaces in the ubiquitin-mediated degradation of c-FLIP S . Interestingly the residues in c-FLIP L corresponding to lipophilic surface II in c-FLIP S are charged rather than hydrophobic, and stabilizing interactions between this region and lipophilic groove II on the DED2 would not take place, possibly interfering with the positioning of Lys 192 and Lys 195 . Furthermore the structure of c-FLIP L is considerably longer at the C terminus, and other structural features (e.g. overall fold) may function to block access to the lysines and hence inhibit ubiquitylation. As the stability of c-FLIP S appears to be directly linked to the correct positioning and availability of the Cterminal ␣-helix as related to DED2, other DED-containing proteins may affect ubiquitylation of c-FLIP S by interacting with the DED2 in c-FLIP S . In this respect, it will also be of great interest to learn how the correct positioning of the c-FLIP S C terminus by its unique tail affects the interaction of c-FLIP S with its ubiquitin ligase(s).
Regarding c-FLIP regulation, there are several reports indicating that the amount of only one of the c-FLIP isoforms is specifically altered. As c-FLIP S has been specifically shown to be down-regulated in WM35 melanoma cells upon cisplatin treatment (39), we tested whether the faster turnover of c-FLIP S could contribute to this process. Indeed the half-life of c-FLIP S was clearly shorter than that of c-FLIP L , and cisplatin-induced down-regulation could be prevented by the proteasome inhibitor epoxomicin. Furthermore upon stimulation of the Tcell receptor, c-FLIP S is up-regulated in the activated T cells (51,52). Also in maturating dendritic cells, only c-FLIP L is up-regulated (53,54). The isoform-specific degradation described in this study is well suited to participate in these and other similar physiological processes. Moreover as c-FLIP isoforms have been shown to be up-regulated in various cancers, it is plausible that proteins involved in c-FLIP degradation may be mutated or down-regulated. Indeed it has been demonstrated that in B cell chronic lymphocytic leukemia cells as well as in Hodgkin disease-derived cells c-FLIP isoforms are significantly more stable than in other cell types (55,56). Therefore, inhibition of pathways that are involved in c-FLIP degradation may contribute to the apoptosis resistance of malignant cells, thereby affecting the possible outcome of cancer treatments. The identification of common and isoform-specific c-FLIP E3 ubiquitin ligases is highly warranted to further understand the regulation of c-FLIP degradation.