Impaired trafficking and activation of tumor necrosis factor-alpha-converting enzyme in cell mutants defective in protein ectodomain shedding.

Protein ectodomain shedding is a specialized type of regulated proteolysis that releases the extracellular domain of transmembrane proteins. The metalloprotease disintegrin tumor necrosis factor-alpha-converting enzyme (TACE) has been convincingly shown to play a central role in ectodomain shedding, but despite its broad interest, very little is known about the mechanisms that regulate its activity. An analysis of the biosynthesis of TACE in mutant cell lines that have a gross defect in ectodomain shedding (M1 and M2) shows a defective removal of the prodomain that keeps TACE in an inactive form. Using LoVo, a cell line that lacks of active furin, and alpha1-Antitrypsin Portland, a protein inhibitor of proprotein convertases, we show that TACE is normally processed by furin and other proprotein convertases. The defect in M1 and M2 cells is due to a blockade of the exit of TACE from the endoplasmic reticulum. The processing of other zinc-dependent metalloproteases, previously suggested to participate in activated ectodomain shedding is normal in the mutant cells, indicating that the component mutated is highly specific for TACE. In summary, the characterization of shedding-defective somatic cell mutants unveils the existence of a specific mechanism that directs the proteolytic activation of TACE through the control of its exit from the ER.

Protein ectodomain shedding is a specialized type of proteolysis that releases the extracellular domain of cell surface proteins, leaving the transmembrane-cytoplasmic region bound to the plasma membrane. Ectodomain shedding can regulate the function of transmembrane growth factors, turn growth factor receptors into agonist or antagonist of the ligands they bind, or modulate cell-cell and cell-extracellular matrix interactions (1). Furthermore, the shedding of the ␤-amyloid precursor protein (also known as ␣-secretase cleavage) prevents the formation of the ␤-amyloid peptide, the main component of the senile plaques found in brains of patients with Alzheimer's disease (2).
Experiments with specific inhibitors indicate that practically all of the shedding events analyzed to date are mediated by zinc-dependent metalloproteases that belong the metzincin family (3). In addition, overexpression of certain metzincins or their dominant-negative forms point to the participation of these metalloproteases in ectodomain shedding (for example see Refs. 4 and 5). A definitive confirmation of the role of a member of the metalloprotease disintegrins (also known as ADAMs or MDCs), a subfamily of the metzincins, in ectodomain shedding came with the development of mice genetically deficient in the tumor necrosis factor-␣-converting enzyme (TACE 1 /ADAM17) (6). The characterization of cell lines derived from TACE knock-out mice (TACEϪ/Ϫ cells) confirmed the involvement of this metalloprotease disintegrin in the shedding of a surprisingly large group of cell surface proteins (reviewed in Ref. 1).
Metzincins are synthesized as zymogens containing a prodomain with an odd cysteine, which coordinates the zinc present in the catalytic site inhibiting the metalloprotease activity (7). Certain metzincins contain a stretch of basic amino acids between the prodomain and the metalloprotease domain and are activated in the secretory pathway by proteolytic removal of the former. It has been frequently assumed that serine proteases of the proprotein convertase family that are highly spe-□ S The on-line version of this article (available at http://www.jbc.org) contains cific for basic amino acid-based motifs and, typically, act in the trans-Golgi network (8) cleave the prodomain of these metzincins. The proprotein convertase family is composed by seven members: four of them (furin, PACE4, PC6, and PC7) are expressed in a broad range of tissues, and two of them (PC1 and PC2) are expressed in neuroendocrine tissues, whereas the expression of PC4 is restricted in spermatogenic cells (Ref. 9 and references therein).
The number of zinc-dependent metalloproteases involved and the mechanisms that regulate ectodomain shedding remain largely undetermined. To address these issues, several years ago we isolated two independent mutant cell lines defective in the shedding of the transmembrane growth factor pro-TGF-␣ (M1 and M2 cells) (10,11). Characterization of these mutant cells showed that they are also defective in the shedding of a variety of proteins such as the transmembrane proheparin-binding epidermal growth factor-like growth factor (proHB-EGF) (12), ␤APP (10), the cell adhesion molecule Lselectin, the ␣ subunit of the receptor for interleukin-6 (11), and the cytokines proTNF-␣ and fractalkine (13,14). The shedding of these molecules is also defective in TACEϪ/Ϫ fibroblasts (6,12,13,(15)(16)(17)(18). Nonetheless, somatic cell fusions between mutant cells and TACEϪ/Ϫ fibroblasts recover the wild type phenotype, suggesting that the mutation affects a component different from TACE but necessary for its activity and perhaps that of other metalloproteases (13). Therefore, the characterization of these cell lines may reveal novel aspects of the regulation of ectodomain shedding.
In this report, we show that the processing of TACE is impaired in M1 and M2 cells. An analysis of cells defective in the activity of furin and experiments with ␣ 1 -Antitrypsin Portland (␣ 1 -PDX), an inhibitor specific for certain proprotein convertases, indicates that in normal cells TACE is processed by furin and other proprotein convertases. However, the mutant cells possess a normal repertoire of proprotein convertases, which are not likely a target of mutations. An analysis of the subcellular location of TACE in these mutant cells reveals a defective intracellular trafficking that leads to the accumulation of the metalloprotease in the early secretory pathway and explains the phenotype observed. Other metzincins processed by furin-like convertases and putatively involved in ectodomain shedding such as ADAM10, ADAM9, or MT1-MMP as well as ADAMTS1 are correctly processed in the mutant cells, indicating that the defect is highly specific for TACE. In summary, characterization of shedding defective somatic cell mutants reveals the existence of a mechanism that specifically mediates the intracellular trafficking and in turn the proteolytic activation of TACE by proprotein convertases.

MATERIALS AND METHODS
Antibodies and Reagents-The anti-TACE polyclonal antibodies and protein A-purified IgG directed against the prodomain or the cytoplasmic or metalloprotease domains of human TACE have been previously characterized (19). Anti-furin and anti-ADAM10 antibodies were generous gifts from Dr. Wolfgang Garten and Dr. Alain Israel, respectively. Monoclonal antibodies against human MT1-MMP were kindly provided by Dr. Alicia G. Arroyo, and antibodies directed against ADAM-TS1 as well as the cDNA encoding the human molecule were a gift from Dr. Luisa Iruela-Arispe. Anti-human ␣ 1 -antitrypsin developed in rabbit was purchased from Sigma. Goat anti-mouse horseradish peroxidaseconjugated antibody was from Amersham Biosciences (Little Chalfont, United Kingdom), and FITC-conjugated anti-mouse IgG was from BD Biosciences. Antibodies against the Golgi marker gm130 were from BD Biosciences, and antibodies anti-KDEL were from Stressgen Biotechnologies. Sulfo-NHS-LC-biotin, horseradish peroxidase-conjugated streptavidin, and the SuperSignal chemiluminescence kit were from Pierce. Cell culture reagents were from Invitrogen.
cDNAs, Cell Lines, Transfections, and Viral Infections-The cDNAs encoding furin, PC7, PC6B/Flag, and ␣ 1 -PDX were provided by Dr. Gary Thomas, and the cDNAs encoding PACE4 and ADAM10 were gifts from Dr. Joseph F. Sucic and Dr. Paul Glynn, respectively. The cDNA encoding Notch tagged at the C terminus with the Myc epitope and Myc-tagged TACE have been previously described (13). To facilitate detection, HA epitopes were introduced right after the propeptide cleavage site of PC7 (after Ser-142 of human PC7) and at the beginning of the cysteine-rich domain of human PACE4 (after Pro-697), respectively, using standard techniques. LoVo and CaCo cells were provided by Drs. Alain Israel and Senén Vilaró, respectively. Transient transfections were performed using the DEAE-dextran method. The efficiencies of transfection were routinely controlled and ranged between 30 and 40%. CHO cells permanently transfected with ␣ 1 -PDX were obtained by co-transfecting ␣ 1 -PDX subcloned into the pcDNA3.1 Z(ϩ) vector (Invitrogen) and the selectable plasmid pREP-4 (Invitrogen) at a DNA ratio of 10:1 using the calcium phosphate precipitate method as previously described (10). Transfectants were selected in 600 g/ml hygromycin and subcloned. The levels of expression of ␣ 1 -PDX were assessed by Western blot using anti-human ␣ 1 -antitrypsin antibodies.
LoVo cells were infected with retrovirus as follows: 293T cells were plated at 2 ϫ 106 cells/60 mm dish. Five minutes prior to transfection, 25 M chloroquine was added to each plate. The transfection solution was DNA (2.5 g of pMDG-VSV, 5 g of pNGUL-MLV-gag-pol, 3 g of the retroviral vector pLZR-IRES-GFP bovine-containing furin, 61 l of 2,5 M CaCl 2 , and 500 l of double-distilled water. The medium was replaced with 3 ml of fresh virus-collecting medium 24 h later, and the retrovirus were recovered 1 day thereafter. Cells were then infected with the viral supernatants containing 6 g/ml Polybrene. Western Blotting-Approximately 10 6 parental HeLa, CHO, M1, M2, Caco, or LoVo cells were lysed in PBS containing 1% Nonidet P-40, 10 mM 1,10-phenanthroline, 5 mM EDTA, and 20 M BB-94 (lysis buffer), and insoluble material was removed by centrifugation. The resulting supernatants were incubated with 25 l of concanavalin A (ConA)-Sepharose (Amersham Biosciences) for 1 h at 4°C to concentrate glycoproteins. ConA-Sepharose beads were washed twice with lysis buffer, and the final beads were treated with or without endoglycosidase H (Endo H) or N-glycosidase F as detailed below, resuspended in sample buffer, and boiled. Samples were then electrophoresed in 7% polyacrylamide gels under reducing conditions and subjected to Western blotting analysis using 1/500 of the anti-prodomain, anti-cytoplasmic domain antibodies directed against TACE, anti-human ␣ 1 -antitrypsin or anti-ADAM10, 1/1000 of anti-furin or anti-MT1-MMP, or 1/2.500 of anti-ADAM-TS1 as indicated. Signals were detected with SuperSignal West Dura Extended Duration substrate (Pierce). As indicated, films were digitized and quantified using MacBAS software (Fuji film).
Metabolic Labeling and Immunoprecipitation-To analyze TACE processing, ϳ2.10 6 CHO or M2 cells were metabolically labeled with 1 mCi/ml [ 35 S]Translabel (Biolink 2000, Arlington Heights, IL) for 1 h in methionine-and cysteine-free medium at 37°C and chased for variable periods of time in complete medium. Cells were then lysed in lysis buffer as described above, and the cell lysates were immunoprecipitated with 2 l of antibodies directed against the cytoplasmic tail of TACE. Immune complexes were collected with protein A-Sepharose, washed three times with washing buffer (PBS containing 0.1% Triton X-100 and 0.1% SDS) and diluted 1:50 with lysis buffer and incubated with ConA-Sepharose to remove non-glycosylated background proteins as previously described (19). The beads were washed, resuspended in loading buffer, and analyzed on 7% SDS-polyacrylamide gels under reducing conditions. Gels were exposed to x-ray films or quantified using a BAS-1800 instrument and software (Fuji film).
To analyze the processing of Myc/Notch by furin, CHO or M2 cells were transfected with Myc/Notch in the presence or absence of furin. Transfected cells were metabolically labeled as above and chased for 1 h in complete medium. Cells were then lysed in lysis buffer, and the cell lysates were immunoprecipitated with 1/100 of anti-Myc antibodies. Immune complexes were collected with protein G-Sepharose, washed, and analyzed on 6% SDS-polyacrylamide gels under reducing conditions. Gels were analyzed as above.
Deglycosylation Studies-For Endo H (Roche Applied Science) and N-glycosidase F (Roche Applied Science) treatments, washed ConAbinding proteins were resuspended in 40 l of a buffer containing 0.15 M sodium citrate, pH 5.5, 1 mM phenylmethylsulfonyl fluoride, and 0.25% SDS or in buffer containing 20 mM sodium phosphate, pH 8.3, 10 mM EDTA, and 1% Nonidet P-40, respectively. ConA beads were then boiled for 2 min, allowed to cool down and treated with 2 milliunits of Endo H or 1 unit of recombinant N-glycosidase F for 12-16 h at 37°C, resuspended in SDS sample buffer, and analyzed by Western blot as described above.
Reverse Transcription-PCR and Sequencing of TACE from Wild Type and M2 Cells-Double-stranded cDNAs from CHO cells, M1 and M2 were synthesized using the Amersham Biosciences kit and instructions of the manufacturer. To sequence TACE, different fragments were amplified using aliquots of the cDNA from parental CHO cells or M2 cells as template and the oligonucleotides shown in Supplementary  Table I. The PCR fragments were purified from agarose gels and sequenced using BigDye TM Terminator Cycle Sequencing Ready Reaction DNA sequencing kit from Amersham Biosciences.
Reverse transcription-PCR analysis of parental CHO cells and M1 and M2 mutants was performed using the Amersham Biosciences kit and instructions of the manufacturer. Aliquots of cDNA were used as template for 20, 25, or 30 cycles of PCR, and the products of the reaction were separated on agarose gels and analyzed with a Fluor S device and software (Bio-Rad). The oligonucleotides used to amplify PACE4 and PC6 fragments have been previously described (20). The rest of oligonucleotides used were as follows: furin, 5Ј-GTG GTC TCC ATC CTG and GCC ATC ATC CTC AGG-3Ј; PC7, 5Ј-AAC GGC TTC AAT GAC TG-3Ј and 5Ј-ATC CTT GCT GCT GCT-3Ј; and actin, 5Ј-CCT GAC CGA GCG TGG CTA C-3Ј and 5Ј-GAA GCA TTT GCG GTG GAC G-3Ј.
Cell Surface Biotinylation-For biotinylation of cell surface TACE, ADAM10, and MT1-MMP, wild type or M2 cells were washed three times with PBS buffer at 4°C and incubated in the same buffer containing 1 mg/ml sulfo-NHS-LC-biotin (Pierce) for 30 min at 4°C. Excess biotinylating reagent was quenched by washing with 10 mM glycine in PBS three times. Cells were lysed as described above, and cell lysates were incubated with neutravidin-agarose (Pierce). Washed beads were electrophoresed and analyzed by Western blot with antibodies against the cytoplasmic domain of TACE, polyclonal antibodies against ADAM10, or monoclonal antibodies against MT1-MMP.
Confocal Microscopy-Cells grown in coverslips were fixed for 30 min in 2% paraformaldehyde in PBS for 20 min. Cells then were permeabilized with 0.1% saponin in PBS for 30 min. Permeabilized cells were incubated for 45 min in PBS with 1% bovine serum albumin containing 1/100 of IgG purified from polyclonal antibodies against the cytoplasmic tail of TACE, anti-KDEL monoclonal antibodies, or FITC-labeled antibodies against the Golgi marker gm130. After incubation with primary antibodies, cells were washed with PBS and incubated for 45 min with Texas Red-conjugated anti-rabbit antibodies, FITC-conjugated antirabbit antibodies, or Texas Red-conjugated anti-mouse antibodies at a 1/100 dilution. Coverslips were washed with PBS, mounted, and viewed on a Leica TCS 4D laser scanning confocal microscope (Leica Lasertechnik). As a control, M2 cells transiently transfected with Myc/TACE were immunostained with anti-Myc and the polyclonal antibody against the cytoplasmic tail of TACE. The extensive colocalization indicates that the staining observed with the polyclonal antibody against TACE is specific ( Supplementary Fig. 3A). To further assess its specificity, the antibody was preincubated with 100 g/ml TACE cytoplasmic tailglutathione S-transferase fusion protein produced using standard techniques. As expected, no signal was detected in CHO cells in the presence of competing protein ( Supplementary Fig. 3B) To specifically stain cell surface proHA/TGF-␣, cells were shifted to 4°C and incubated with 1/100 of anti-HA in PBS containing 1% bovine serum albumin for 1 h. Cells were fixed, permeabilized, and treated with antibodies against the cytoplasmic tail of TACE, FITC-conjugated anti-rabbit, or Texas Red-conjugated anti-mouse antibodies as above.

Processing of TACE in Shedding-defective Mutant Cells-M1
and M2 are CHO cell mutants independently isolated because of their lack of proTGF-␣ shedding (10,11). Previous results suggested that the component mutated in these cells and TACE, the metalloprotease-disintegrin responsible for the shedding of the growth factor (6), are different (13). To confirm this conclusion, we sequenced the cDNAs encoding TACE from parental CHO and M2 cells. Both sequences were found identical (data not shown). Thus, we corroborated that the component mutated in M2 cells is different from TACE.
To further characterize the mutant cell lines, we analyzed the species of TACE expressed by them using specific antibodies. In agreement with previous results (19), antibodies directed against the cytoplasmic domain of TACE stain two proteins in HeLa and CHO cells (Fig. 1, A and B), whereas antibodies against the prodomain of TACE only stain the slowest migrating protein (Fig. 1A). These bands were identified as full-length and processed TACE devoid of prodomain, respectively (Fig. 1A). In M1 and M2 cells, the only apparent TACE species is the full-length form, suggesting that these cells have a defect in the processing of TACE or, alternatively, that the processed form of TACE is specifically and rapidly degraded in the mutant cells. To distinguish between these possibilities, we performed pulse-chase experiments. Because M1 and M2 cells belong to the same complementation group (11) and show a similar absence of processed TACE, pulse-chase and most subsequent experiments were performed only with M2 cells. The full-length form of TACE is long-lived in M2 cells (Fig. 1C, t1 ⁄2 ϳ20 h) compared with that in wild type cells (Fig. 1C, t1 ⁄2 Ͻ 8 h, and data not shown). In these cells, full-length TACE is processed to the form that lacks the prodomain undetectable in M2 cells (Fig. 1C). These results strongly suggest that M2 cells are unable to process the prodomain of TACE. Unexpectedly, at long periods of chase (24 h) the electrophoretic migration of processed TACE in wild type cells is slightly slower than that at the 12-h time point (Fig. 1C, wild type (WT), compare lanes 12 and 24 h). This slight difference in migration cannot be appreciated in HeLa, CaCo, or LoVo cells (data not shown) and provides an explanation for the migration of the processed form of TACE from CHO cells as a broad band in Western blots (for example see Fig. 1B).
TACE Is Processed by Proprotein Convertases-Because the prodomain is connected with the metalloprotease domain by a typical furin cleavage site, it has been widely assumed that TACE is activated by furin-like proprotein convertase(s) in the secretory pathway (17,19,21). To test this assumption, we analyzed TACE species in cells devoid of furin activity and in cells overexpressing the ␣ 1 -PDX, a protein-based inhibitor of proprotein convertases highly specific for furin and PC6B (22).
The human colon carcinoma LoVo cell line has been used to characterize the processing of numerous proteins, because it expresses inactive forms of furin (23). As judged by Western blot analysis using anti-TACE antibodies, LoVo cells express the full-length as well as low levels of the processed form (compared with those expressed by CaCo cells (Fig. 2A), indicating that convertases different from furin can process the prodomain of TACE (Fig. 2A). The electrophoretic migration of full-length TACE from LoVo cells is different from that observed using a variety of cell lines such as CaCo, HeLa, or CHO cells ( Fig. 2A and data not shown). Because proteins are differentially glycosylated in different cell lines and glycosylation frequently alters the electrophoretic migration, we de-glycosylated samples from the different cell lines and analyzed the electrophoretic migration of TACE in the resulting preparations. Treatment of samples with N-glycosidase F confirmed that the differences in the apparent molecular weight of TACE from LoVo and Caco cells were because of differential glycosylation. To determine the possible ability of furin to process TACE, we infected LoVo cells with retrovirus that included a vector coding for furin and analyzed the processing of the metalloprotease. As shown in Fig. 2B, overexpression of furin increased the levels of processed TACE, indicating that the lack of this convertase accounts for the low levels of TACE processing observed in these cells (Fig. 2A, compare the levels of processed TACE in LoVo and CaCo cells) To confirm these results, we showed that ␣ 1 -PDX inhibits the processing of TACE, although even in cells with a high level of overexpression of this inhibitor, some residual processing can be detected (Fig. 2C). Collectively, these results indicate that TACE is processed by furin in normal cells. In the absence of furin, alternative convertases can remove the prodomain of TACE, albeit less efficiently.
Proprotein Convertases in Wild Type and M2 Cells-Because furin and probably other proprotein convertases participate in the removal of the prodomain of TACE, one or more of these processing enzymes could be inactive in the mutant cells. To investigate this possibility, we first determined the set of proprotein convertases expressed by wild type and mutant cells by performing semi-quantitative PCR with oligonucleotides specific for the four convertases widely expressed (24). The result of this analysis showed no major differences in the levels of furin, PACE4, and PC7 in wild type and mutant cells (Fig. 3A). Although it has been previously found that PC6B is expressed in CHO cells (24), we were unable to detect this convertase using two different sets of primers (data not shown).
We next analyzed whether the convertases expressed by M2 cells are the target of the mutation that prevents the processing of TACE by analyzing the effect of the overexpression of wild type convertases on the mutant phenotype. None of the convertases tested had any detectable effect on the processing of TACE in wild type or M2 cells (Fig. 3B), indicating that convertases are not rate-limiting in the processing of TACE in wild type cells and that they are not the defective component in M2 cells. To rule out the possible interference of the epitope tags with the proteolytic activity of PC7, PC6B and PACE4, parallel experiments were performed with non-tagged versions of the convertases and the results were indistinguishable from those shown in Fig. 3B (data not shown). As a positive control, we showed that overexpression of furin in wild type and M2 cells leads to an enhanced processing of Notch, a previously described substrate of this convertase (25), as judged by the quantification of full-length Notch and one of the products of the processing reaction, the C-terminal fragment (Notch TM) ( Fig. 3C and see also supplementary Fig. 1). It should be noted that in contrast to the processing of TACE, the processing of overexpressed Notch is limited by the amount of furin because it can be enhanced by furin overexpression. These results confirm that proprotein convertases are not likely the mutated component in the shedding-defective mutant cell lines.
Subcellular Location of TACE in M2 Cells-Proprotein convertases are synthesized as zymogens containing a prodomain. Upon exit from the ER, the prodomain is autocatalytically removed and the convertases reach the trans-Golgi system as active endoproteases (24). To test whether TACE progresses correctly through the secretory pathway in M2 cells and is, therefore, accessible to proprotein convertases, we analyzed the subcellular location of endogenous TACE in mutant cells and compared it with that in control wild type cells.
In agreement with previous results and results shown here (Figs. 2 and 4) (19), full-length TACE is sensitive to Endo H treatment in wild type cells, indicating that this form resides largely in the early secretory pathway and does not reach the medial Golgi. Processed TACE is resistant to Endo H (Fig. 4A) and can be specifically detected by cell surface biotinylation (Fig. 4B), indicating that this form has transversed the medial Golgi and reaches the plasma membrane. Consistently, immunofluorescence analysis on wild type cells with anti-TACE antibodies show intracellular staining that partially colocalizes with anti-KDEL and anti-gm130 (that bind to ER and cis-Golgi markers, respectively) (Fig. 4, C and D) and cell surface staining that partially colocalizes with cell surface proHA/TGF-␣ (Fig. 4E). In M2 cells, full-length TACE is entirely sensitive to Endo H treatment (Fig. 4A), indicating that it remains in the early secretory pathway. In agreement with this conclusion, TACE is not displayed at the cell surface (Fig. 4B). Furthermore, although the intracellular immunostaining pattern of TACE in M2 cells is similar to that in wild type cells, perhaps with a lower level of colocalization with anti-gm130 (Fig. 4D), no cell surface immunostaining can be detected in M2 cells (Fig. 4E). Collectively, these results indicate that the trafficking of TACE is impaired in M2 cells and that the metalloprotease remains largely in the ER and does not reach the trans-Golgi where the processing of TACE by proprotein convertases takes place.
The Processing of Different Metalloproteases Is Normal in Shedding-defective Mutants-M1 and M2 cells show a gross defect in the metalloprotease-dependent shedding of plasma membrane proteins (10,11); however, since the defect in these cell lines has remained elusive, it has not been possible to determine whether the affected component controls one, few, or many metalloproteases. To investigate this point, we analyzed the transport to the cell surface and processing of several metalloproteases, some of them putatively involved in ectodomain shedding, in M2 cells. Specific antibodies identified endogenous ADAM10 and MT1-MMP in CHO cells (Fig. 5). The processing of these metalloproteases is normal in M2 cells as judged by Western blot (Fig. 5). Moreover, biotinylation experiments showed that the processed forms of these metalloproteases is readily detected on the cell surface of M2 cells and accumulates to levels similar to those observed in wild type cells (Fig. 5), further supporting a normal trafficking of these proteases in M2 cells.
Although ADAM9 is widely expressed (26), we were unable to detect it in CHO cells with polyclonal antibodies against the mouse protein; therefore, we analyzed the processing of transiently transfected mouse ADAM9 and found it indistinguishable in wild type and M2 cells (Fig. 5). Furthermore, the processing of transfected ADAM-TS1 to the previously described p87-and p65-soluble fragments (27) was also similar in both cell types (Fig. 5). Because it has been suggested that the metalloproteases analyzed are processed by furin-like convertases (27)(28)(29)(30), these results confirm that convertase activity is normal in mutant cells. Collectively, the evidences shown here unveil the existence of a highly specific mechanism defective in M1 and M2 cells that mediates the intracellular trafficking and therefore the processing of TACE. Ultimately, this mechanism controls the shedding of many proteins through the control of the activity of TACE. DISCUSSION The phenotype of TACE knock-out mice and that of the embryonic fibroblasts generated from these mice strongly sug- FIG. 3. Proprotein convertases in wild type and mutant CHO cells. A, CHO, M1, and M2 cells were subjected to analysis by reverse transcription-PCR using specific oligonucleotides as described under "Materials and Methods." The products of the reaction were analyzed in agarose gels. B, wild type (WT) and M2 cells were transiently transfected with different proprotein convertases as described under "Materials and Methods." Cell lysates from transfected cells were analyzed by Western blot using antibodies against the cytoplasmic domain of TACE, anti-furin polyclonal antisera, or monoclonal antibodies against the HA or FLAG epitopes as indicated. A nonspecific band observed in certain preparations is labeled with an asterisk. C, wild type and M2 cells transfected with Myc/Notch in the presence or absence of furin were metabolically labeled, chased for 1 h, and lysed, and the cell lysates were immunoprecipitated with anti-Myc antibodies. Immunoprecipitates were analyzed on SDS gels, and full-length and the Cterminal fragment (Notch TM), which arises after furin cleavage, were quantified. Results are expressed as the percentage of Notch TM relative to Total Notch (see also supplementary data) and are averages Ϯ S.D. of three independent experiments.
gest that TACE is necessary for the shedding of a subset of cell surface proteins with disparate structures and functions (6,12,13,(15)(16)(17)(18). Given this promiscuity, it seems reasonable to conjecture that the activity of TACE is under tight control to prevent unwanted proteolysis. However, little is known regarding the cellular mechanisms that control the activity of TACE. Furthermore, although it has been long known that ectodomain shedding is a regulated process (for example see Ref. 31), it is not known how the activity of TACE is up-regulated upon activation of protein kinase C or mitogen-activated protein kinases, two well characterized mediators of the activation of ectodomain shedding (for a recent review see Ref. 32) Several years ago, we undertook a genetic approach to characterize the shedding machinery that acts on proTGF-␣. First, we isolated chemically induced somatic cell mutants (M1 and M2 cells) defective in the activated shedding of proTGF-␣, and subsequently, we found that these mutant cell lines have a gross defect in the shedding of a variety of proteins, including many substrates whose shedding is also defective in TACEϪ/Ϫ cells (10 -14). However, the wild type phenotype of somatic cell fusions between M2 and TACEϪ/Ϫ cells indicated that the defect in M2 cells affects a component other than TACE (13). TACE transfected into M2 cells does not rescue the wild type phenotype, indicating that the component defective in these mutant cells is different from TACE (13). Furthermore, no mutations can be detected in TACE from M2 cells (this report). Therefore, the defect in M2 cells does not affect TACE; thus, the characterization of this defect provides an opportunity to study mechanisms that control the activity of TACE.
Because of the current availability of polyclonal antibodies that allow the detection of endogenous TACE in CHO cells, we found that the processing of the metalloprotease is completely blocked in M1 and M2 cells. This result apparently contradicts a previous one, because a C-terminal tagged version of TACE was judged to be processed marginally but to the same extent in wild type and mutant cells (13). However, several lines of evidence strongly suggest that the fate of transfected TACE is not a good indicator of that of the endogenous protein and that the apparent "processing" previously observed with transfected TACE is artifactual. Pulse-chase experiments show that prodomain processing affects almost all of the detectable TACE labeled after the pulse (this report and Ref. 19), whereas it affects only to ϳ10% transfected TACE (13). In addition, the electrophoretic migration of the marginal amount of TACE processed in transient transfections is different from that of the FIG. 4. Subcellular localization of TACE in wild type and M2 cells. A, glycoproteins from CHO or M2 cells were treated with or without Endo H and analyzed by Western blot with polyclonal antibodies directed against the cytoplasmic domain of TACE. B, wild type (WT) or mutant cells were treated with sulfo-NHS-LC-biotin and lysed, and the cell lysates were immunoprecipitated with antibodies against the cytoplasmic domain of TACE. Biotinylated immunoprecipitates were analyzed by Western blot with Neutravidin peroxidase. C and D, wild type or mutant cells were permeabilized and incubated with protein A-purified rabbit IgG directed against the cytoplasmic tail of TACE and monoclonal anti-KDEL antibodies or FITC-labeled polyclonal antibodies against gm130. TACE immunostaining was visualized with FITClabeled or Texas Red-labeled anti-rabbit antibodies, and anti-KDEL was visualized with Texas Red-labeled anti-mouse antibodies. E, CHO cells or M2 cells expressing proHA/TGF-␣ were incubated with anti-HA monoclonal antibodies at 4°C, washed, permeabilized, and incubated with anti-TACE polyclonal antibodies, Texas Red-labeled anti-mouse antibodies, and FITC-labeled anti-rabbit antibodies. Note that the morphology of the cells changed during the incubation at 4°C. endogenous processed protein (data not shown). Currently, the basis for the artifactual processing of TACE in transient transfections is unknown. Presumably, they could be related to the saturation of a component needed for the processing of the metalloprotease. We are currently investigating this possibility.
An analysis of the processing of TACE in LoVo cells and in cells overexpressing ␣ 1 -PDX indicates that the processing of TACE is carried out by furin and other convertases. Because ␣ 1 -PDX is specific for furin and PC6B (22) and CHO cells express undetectable levels of the latter convertase, the clear inhibition of TACE processing in CHO overexpressing the serpin-based inhibitor cells strongly suggests that furin is the main convertase acting on TACE. However, two main evidences indicate that convertases are not the target of the mutations in M1 and M2 cells. Transfection of mutant cells with furin does not rescue the wild type phenotype, and the processing of several metalloproteases known to be substrates of furin (27)(28)(29)(30) is normal in mutant cells.
Although the components of the machinery responsible for the intracellular trafficking of TACE await to be identified, the results presented here show that the factor mutated in M1 and M2 cells is specific for TACE and is not necessary for the trafficking of other metzincins. This conclusion is particularly significant when considering the specificity of metalloprotease disintegrins. It has been proposed that ADAM10 and ADAM9 play a role in the shedding of ␤APP (4, 33) and that ADAM9 is involved in the shedding of proHB-EGF (34). However, the processing of these metalloprotease disintegrins is normal in M2 cells, indicating that at least in CHO cells, the activated ␣-secretase activity that sheds the ectodomain of ␤APP and that of proHB-EGF is identical to TACE. Although the phenotype of TACEϪ/Ϫ and ADAM9Ϫ/Ϫ cells obtained from knockout mice (that show defective and normal shedding of proHB-EGF and ␤APP, respectively) indicates that this is also true in mouse embryonic fibroblasts (12,13,15,35), it may well be possible that different metalloprotease disintegrins act on different cell lines. In this scenario, TACE would be the main endogenous shedding activity in CHO cells and mouse embryonic fibroblasts, whereas ADAM10 and ADAM9 could be responsible for the shedding of ␤APP in HEK 293 and in transiently transfected COS cells, respectively, and ADAM9 would shed proHB-EGF in Vero cells (34).
In summary, as expected, the characterization of mutant cell lines initially isolated for lack of proTGF-␣ shedding has allowed us to detect the existence of a mechanism that specifically controls the activity of TACE through the regulation of its intracellular trafficking. Obviously, the identification of the components that allows the export of TACE from the ER and therefore its proteolytic activation will be the next step in understanding this mechanism of control.