Characterization of Growth Factor-induced Serine Phosphorylation of Tumor Necrosis Factor-α Converting Enzyme and of an Alternatively Translated Polypeptide*

Tumor necrosis factor-α converting enzyme (TACE) is a prototype member of the adamalysin family of transmembrane metalloproteases that effects ectodomain cleavage and release of many transmembrane proteins, including transforming growth factor-α. Growth factors that act through tyrosine kinase receptors, as well as other stimuli, induce shedding through activation of the Erk mitogen-activated protein (MAP) kinase pathway without the need of new protein synthesis. How MAP kinase regulates shedding by TACE is not known. We now report that the cytoplasmic domain of TACE is phosphorylated in response to growth factor stimulation. We also identified a naturally expressed smaller polypeptide corresponding to most of the cytoplasmic domain of TACE. This protein, which we named SPRACT, is derived through alternative translation of the TACE-coding sequence and is, similarly to TACE, phosphorylated in response to growth factor and phorbol 12-myristate 13-acetate stimulation. Phosphoamino acid analysis revealed that growth factor-induced phosphorylation of TACE occurs only on serine and not on threonine or tyrosine. Tryptic mapping experiments coupled with site-directed mutagenesis identified Ser819 as the major target of growth factor-induced phosphorylation, whereas Ser791undergoes dephosphorylation in response to growth factor stimulation. The phosphorylation of Ser819, but not the dephosphorylation of Ser791, depends on activation of the Erk MAP kinase pathway. Increased SPRACT expression or mutation of the TACE cytoplasmic domain to inactivate growth factor-induced phosphorylation did not detectably affect growth factor-induced shedding of transmembrane transforming growth factor-α by TACE. The roles of SPRACT and the cytoplasmic phosphorylation of TACE remain to be defined.

Various cell surface proteins with diverse functions undergo regulated proteolytic cleavage of their extracellular domains, a process that results in ectodomain shedding (for review see Refs. [1][2][3][4]. The subsequent release of the polypeptide from the cell can have profound consequences not only at the cellular physiological level but also at a more tissue-wide or even systemic level. For example, transmembrane L-selectin is an adhesion protein for leukocytes, and ectodomain cleavage of Lselectin may therefore down-regulate inflammation by limiting neutrophil accumulation and lymphocyte activation (5)(6)(7)(8). Because shedding plays such an important role in the regulation of cellular functions, its dysregulation can contribute to disease conditions. Thus, ectodomain cleavage of transmembrane tumor necrosis factor-␣ (TNF-␣) 1 and consequent release of soluble TNF-␣ is thought to contribute to cachexia and arthritis, which cannot be induced by transmembrane TNF-␣ (9,10). In addition, families carrying a non-cleavable TNF-␣ receptor are presented with severe autoimmune reactions (11). However, despite its importance, the molecular mechanisms underlying ectodomain processes are poorly understood.
Ectodomain shedding also regulates the activities and roles of transmembrane growth factors such as transforming growth factor-␣ (TGF-␣) and its family members. These growth factors are expressed at the cell surface as transmembrane proteins, yet can also be released as soluble growth factors as a result of their ectodomain shedding (12)(13)(14)(15)(16)(17). Both the soluble and the transmembrane forms of these growth factors can activate the receptors, i.e. the EGF/TGF-␣ receptor or related transmembrane tyrosine kinases (13, 18 -21). Thus, transmembrane TGF-␣ or related growth factors only exert autocrine activity and stimulate receptors on adjacent cells and are most likely unable to induce receptor internalization. In contrast, release of soluble ligand following ectodomain shedding allows the cell to induce biological effects, e.g. TGF-␣-mediated proliferation, on non-adjacent cells and allows for ligand-induced internalization of the receptors. In addition to these two modes of ligand signaling depending on ectodomain shedding of the transmembrane growth factor, their receptors can also be subject to ectodomain cleavage, thereby down-regulating ligandinduced receptor activation (22)(23)(24).
Diverse stimuli are known to activate ectodomain shedding (25)(26)(27)(28)(29). These inducers are often artificial compounds that are thought to correlate with natural signaling pathways. For example, the phorbol ester PMA, which activates protein kinase C, is known as a potent activator of ectodomain shedding of various transmembrane proteins (27,30,31). In contrast, little is as yet known about natural inducers and signaling pathways that activate ectodomain shedding under physiological conditions. Recent studies (32)(33)(34)(35)(36) have uncovered the roles of mitogen-activated protein (MAP) kinase signaling pathways in the activation of ectodomain shedding. Thus, growth factors and activated tyrosine kinase receptors rapidly induce ectodomain shedding of TGF-␣ and other transmembrane proteins through induction of the Erk MAP kinase signaling pathway without the need for new protein synthesis (32). PMA-induced ectodomain shedding of TGF-␣ or its family member heparin binding-epidermal growth factor-like growth factor (HB-EGF) and TNF-␣ is also mediated through activation of the Erk signaling pathway (32,33,35). In addition, activation of the p38 MAP kinase pathway, often a result of inflammatory mediators or physiological stress, also leads to ectodomain shedding (32,37). Thus, inhibition of both Erk and/or p38 MAP kinase signaling pathways strongly suppresses ectodomain shedding of diverse transmembrane proteins in response to various stimuli (32,37). Inhibitor studies and functional experiments have revealed that the cleavage of TGF-␣ and various other transmembrane proteins is mediated by cell surface metalloprotease(s) (30, 38 -40). Functional inactivation through gene targeting has subsequently implicated a role for the transmembrane protease TACE in ectodomain shedding of not only TNF-␣ but also L-selectin and TGF-␣ in vivo (31). Accordingly, cells deficient in functional TACE expression display impaired release of soluble TGF-␣ as well as the TGF-␣-related amphiregulin, HB-EGF, and neuregulins (17,41). Finally, TACE has also been implicated in shedding of various other proteins, e.g. ErbB4, TNF receptors, ␤-amyloid precursor protein, and Notch, thus illustrating its role in diverse physiological contexts (31,(42)(43)(44).
TACE, a member of the adamalysin (ADAM) family of transmembrane metalloproteases and also known as ADAM 17, is a prototype "sheddase." Its characterization and studies on the mechanisms of activation may therefore provide models for how other adamalysins function. The protein sequence of TACE consists of an extracellular metalloprotease domain flanked upstream by a prodomain and downstream by a cysteine-rich disintegrin domain, a single transmembrane domain and a cytoplasmic domain (45,46). The prodomain serves as an inhibitor of the protease in its zymogen state and is removed at the late Golgi processing stage (47). Although it was originally thought to play a role in substrate recognition, the disintegrin domain has been shown to be required for shedding of an interleukin 1 receptor only among several TACE substrates tested (42).
The regulation of ectodomain shedding by TACE and the ability of MAP kinase signaling pathways to activate shedding without the need for new protein synthesis suggest that the cytoplasmic domain may act as a signal transducer that regulates shedding by the protease domain of TACE in response to intracellular activities. Interestingly, the 130-amino acid cytoplasmic domain of TACE, like several other ADAM members, has a potential Src homology 3 ligand protein-binding motif (45,46). This sequence or a proximal region of TACE interacts with mitotic arrest deficient 2, although the physiological relevance of this interaction is unknown (48). More recently, the PDZ domain-containing protein PTPH1 has been shown to interact with the carboxyl terminus of the cytoplasmic domain and to possibly down-regulate TACE function (49).
Phosphorylation plays a critical role in intracellular signaling. It has been reported that PMA, an artificial shedding activator, can induce TACE phosphorylation (45,50). Interestingly, MDC9, which shares sequence similarity with TACE in its cytoplasmic domain, is phosphorylated by protein kinase C␦ (51). In one study this phosphorylation was shown to regulate the induction of HB-EGF shedding by PMA (51), whereas other findings indicate that PMA may activate HB-EGF shedding through the Erk MAP kinase signaling pathway instead (33,35). We now show that growth factor or PMA stimulation leads to enhanced phosphorylation of TACE through activation of the Erk MAP kinase signaling pathway. We also identified an alternative translation product of TACE that we name SPRACT. SPRACT corresponds to most of the cytoplasmic domain of TACE and similarly to TACE undergoes regulated phosphorylation. We show that TACE is phosphorylated on serines upon growth factor or serum stimulation. We identified the major site of phosphorylation in response to growth factor stimulation, as well as another serine, which shows decreased phosphorylation upon stimulation. The function of these phosphorylation sites and the role of SPRACT remain to be characterized.
Four polyclonal anti-TACE antibodies were used throughout the course of this study. A rabbit antiserum from QED Bioscience, Inc. (San Diego, CA), and a goat antiserum from Santa Cruz Biotechnology (Santa Cruz, CA) both recognize a carboxyl-terminal sequence. The antibody from QED Bioscience was found preferable for analyzing endogenous TACE because of a higher affinity to protein A-agarose beads in immunoprecipitation and a cleaner background in Western blotting as compared with the antibody from Santa Cruz Biotechnology. The blocking peptide for the antisera as well as normal rabbit IgG was purchased from Santa Cruz Biotechnology. A rabbit polyclonal antibody raised against the recombinant TACE cytoplasmic domain fused to glutathione S-transferase, and a polyclonal antiserum against the extracellular domain of TACE were generous gifts from Dr. Blobel (Sloan-Kettering Institute, New York) (52,53).
The expression vector for the wild type TACE was constructed by PCR amplification of the human TACE cDNA cloned in a bacteriophage vector that was provided by Dr. Roy Black (Immunex, Seattle, WA). The PCR-amplified cDNA was inserted into the pRK5 vector, which uses the human cytomegalovirus promoter to drive expression of an inserted gene in mammalian cells. Deletion and single or multiple amino acid substitution mutants of the TACE coding sequence were constructed using PCR-based approaches. The sequences of wild type and mutant TACE coding sequences in all expression plasmids were confirmed by automated DNA sequencing performed at the University of California, San Francisco, Biomolecular Resource Center. Information on the plasmid design and construction will be provided upon request. Expression plasmids for constitutively activated MEK1 (⌬N-S218E-S222D) (54) and constitutively activated Erk2 (MCMV5-Erk2-MEK1 LA) (55) were provided by Dr. Nathalie G. Ahn (University of Colorado) and Dr. Melanie H. Cobb (University of Texas Southwestern Medical Center), respectively.
Cell Lines, Culture Conditions, and Transfection-The HeLa S3 cell line and HEK293 cell line were maintained in Dulbecco's modified minimal essential medium (DMEM) supplemented with 7% fetal bovine serum. CHO cells were cultured in DMEM supplemented with 2 mM proline and 10% fetal bovine serum. Transfection of CHO and HEK293 cells was achieved using the LipofectAMINE reagent (Invitrogen). The EC2 cell line that lacks the expression of catalytically active TACE as a result of targeted gene inactivation, i.e. the tace ⌬Zn/⌬Zn mutation, was kindly provided by Dr. Roy Black (Immunex). EC2 cells were maintained in DMEM containing 10% fetal bovine serum and were transfected using LipofectAMINE and the Plus reagent (Invitrogen). K562, THP-1, and U937 cell lines were obtained from the University of California San Francisco Tissue Culture Facility and cultured in RPMI medium supplement with 10% FBS.
In Vivo 35 S and 32 P Labeling of TACE-CHO cells, grown in 6-well plates, were transfected with expression vectors for wild type or mutated TACE and cultured in serum-free medium overnight (32). Metabolic labeling of TACE proteins with [ 35 S]cysteine/methionine was performed as described previously (32). To detect phosphorylation of TACE in vivo, HeLa S3 cells at 90% confluency in 6-or 10-cm dishes or CHO cells in 6-well plates transiently transfected with a TACE expression vector were subjected to overnight serum starvation before 32 P labeling, washed with phosphate-free DMEM, and cultured with the DMEM supplemented with 10% (v/v) [ 32 P]orthophosphate (PerkinElmer Life Sciences or Amersham Biosciences) for 2 h. To observe induced TACE phosphorylation, FGF (final concentration, 10 ng/ml), EGF (final concentration, 10 ng/ml), dialyzed fetal bovine serum (FBS) (final concentration, 20%) or PMA (final concentration, 20 nM), with or without the MEK inhibitor U0126 (final concentration, 10 M), was added into the labeling medium. When FBS was used, an equal volume of phosphatefree DMEM was added to control wells. The cells were cultured for an additional 15 min, and labeling was terminated by placing the culture plates on ice and washing with ice-cold phosphate-buffered saline.
Immunoprecipitations-Cells were lysed in phosphate-buffered saline containing 1% Nonidet P-40, 2 mM sodium orthovanadate, 1 mM NaF, 1 M 4-(2-aminoethyl)benzenesulfonyl fluoride, 5 mM EDTA, 10 mM phenanthroline as well as "Complete" protease inhibitor mixture (Roche Applied Science). Immunoprecipitation of TACE in the cell extract with antibodies from rabbit and protein A-conjugated agarose beads was carried out by rotating on a platform at 4°C for 2 h, followed by three washes with the above lysis buffer. Protein G beads were used for immunoprecipitation using an anti-TACE antibody from goat. In selected experiments analyzing the nature of SPRACT, 10 M TAPI-1 was included in the steps of cell lysate preparation and immunoprecipitation. The immunoprecipitated proteins, including TACE and SPRACT, were separated by SDS-polyacrylamide discontinuous gradient gel electrophoresis using a Hoefer SE 400 vertical electrophoresis apparatus. The 16-cm gel consisted of 9 cm of 13.5% gel at the bottom, 3-4.5 cm of 7.5% gel in middle, and regular 4.5% stacking gel on the top. Electrophoresis was done at 50 V for about 18 h, using a prestained broad range molecular standard (Bio-Rad). In most experiments, the 52-kDa ovalbumin entered the 13.5% gel, and the 85-96-kDa bovine serum albumin stayed in the 7.5% gel. Occasionally, the bovine serum albumin ran into the 7.5% gel. For 35 S-or 32 P-labeled samples, the gel was dried, and radiolabeled TACE and SPRACT were visualized by autoradiography. The phosphorylation levels of these two proteins in selected autoradiograms were quantitated by the NIH Image 1.61 software using a Scion 1.62a image acquiring system (Scion Corp., Frederick, MD).
Phosphoamino Acid Analysis-For phosphoamino acid analysis of overexpressed TACE, CHO cells grown on 10-cm culture plates were transfected with the TACE expression plasmid. 32 P labeling and immunoprecipitation of TACE were performed as described above. The precipitated TACE was resolved with 7.5% SDS-PAGE using a Bio-Rad mini-protein II electrophoresis apparatus and was transferred onto PVDF membrane. The radioactive band as visualized by autoradiography was cut out and incubated for 45 min at 120°C in the presence of 6.0 N HCl. The resulting samples were dried and redissolved in a solution containing unlabeled phosphorylated serine, threonine, and tyrosine standards. Separation of the amino acids by thin layer electrophoresis and staining of the amino acid standards with ninhydrin were done as described previously (56). Visualization of TACE-derived 32 Pphosphorylated amino acid(s) was accomplished by autoradiography.
To perform phosphoamino acid analysis on endogenous TACE and SPRACT, HeLa S3 cells grown on 15-cm dishes were labeled with [ 32 P]orthophosphate. TACE and SPRACT were immunoprecipitated and resolved by discontinuous gel electrophoresis as described above. Autoradiography was carried out with dried gel, and radioactive TACE bands were cut out and treated with 6 N HCl at 100°C for 90 min. The TACE-derived amino acids in the solution were dried, reconstituted in a solution containing unlabeled phosphorylated serine, threonine, and tyrosine, and separated by thin layer chromatography (57). Location of the amino acid standards was visualized by ninhydrin staining, and TACE-derived 32 P-phosphorylated amino acids were visualized following autoradiography (57).
CNBr Mapping-32 P-Labeled TACE and SPRACT were immunoprecipitated from transiently transfected 10-cm plates of CHO cells, resolved by discontinuous SDS-PAGE as described above, and transferred onto nitrocellulose membranes. The radioactive bands were cut out and incubated with 5 M CNBr in 70% formic acid (56). The resulting peptides were resolved by 18% SDS-PAGE, and their phosphorylation was revealed by autoradiography.
Tryptic Peptide Mapping-32 P-Labeled TACE was prepared as for phosphoamino acid analysis, and phosphorylated TACE was visualized by direct autoradiography of the wet gels. Gel bands containing radiolabeled TACE and SPRACT were excised, cut into 6 -8 pieces, and subjected to two 45-min washes at 37°C with a solution containing equal volumes of acetonitrile and 20 mM ammonium carbonate, followed by overnight in-gel digestion with modified trypsin. The resulting peptides were separated by thin layer electrophoresis followed by thin layer chromatography (56) and visualized by autoradiography.
Western Blotting-For detection of TACE and SPRACT from transfected CHO cell lines, cells grown on 6-well plates, cells were dissolved by adding SDS-PAGE sample buffer (500 l/well), removed by scraping, and sonicated for 10 s to shear the genomic DNA. Samples of 20 l were heated to 100°C for 5 min and subjected to gel electrophoresis. Proteins were transferred onto PVDF membrane with 0.2-m pore size (Bio-Rad). We found that SPRACT is relatively poorly retained by PVDF membranes, especially by those with 0.45-m pore size, and that the retention was sensitive to the amounts of the detergent Tween 20 in the washing buffer and the duration of washes. The final conditions we adopted for the detection of TACE and SPRACT are as follows. The PVDF membrane was blocked with 5% bovine serum albumin for 2 h, reacted with polyclonal antibodies to the cytoplasmic domain of TACE or its carboxyl-terminal sequence at the dilution of 1:1000 for 2 h, washed three times each for 10 min with Tris-buffered saline containing 0.05% Tween 20 (TBST), and then reacted to horseradish peroxidase-conjugated goat anti-rabbit IgG (Sigma) or mouse anti-goat IgG (Santa Cruz Biotechnology) for 1 h. After three washes as described above, visualization of TACE and SPRACT was achieved by chemiluminescence with the ECL kit (Amersham Biosciences).
Endogenous TACE and SPRACT could also be detected in a similar manner as described above using larger amounts of cell extracts and the more sensitive ECL Plus kit for visualization. However, the increased total cell extracts distorted the gel leading to poor resolution of the protein bands. We found that in HeLa S3 and HEK293 cells this problem could be circumvented by modifying the sample preparation. Confluent cells grown on 6-well plates were extracted with the lysis buffer used for immunoprecipitation (200 l/well). The lysates were mixed with 0.2 volume of the 5ϫ concentrated sample buffer, and 50 l of the mixtures were subjected to gel electrophoresis. The remaining steps of the Western blotting to detect the endogenous proteins were as described for transfected CHO cells, except for the use of the ECL Plus kit (Amersham Biosciences).
Evaluation of TGF-␣ Ectodomain Shedding-TGF-␣ ectodomain shedding was assessed using a pulse-chase assay with transiently transfected CHO cells, as described (32,58). The same assay was also adapted for the use of EC2 cells that lack the catalytically active TACE.

TACE Is Phosphorylated in Response to Growth Factor and
PMA-Growth factor stimulation induces ectodomain shedding of transmembrane proteins, including TGF-␣, TNF-␣, and Lselectin, through activation of the Erk MAP kinase signaling pathway (32). Because the transmembrane metalloprotease TACE has been implicated in the cleavage of these transmembrane proteins, we assessed the phosphorylation state of TACE. In extracts prepared from [ 32 P]orthophosphate-labeled HeLa S3 cells, two radioactive TACE bands were detected, a 128-and a 100-kDa band (Fig. 1A). Both TACE bands were also apparent in parallel Western blots of these lysates (Fig. 1B). According to previous studies (47), the 100-kDa band is the mature TACE that is derived from the 128-kDa glycosylated TACE band after the removal of the 196-amino acid prodomain (Fig. 1A).
Treatment of serum-starved cells with FGF, EGF, fetal bovine serum, or PMA for 15 min significantly increased the TACE phosphorylation level (Fig. 1A). Western blot analysis showed that the stimuli did not show a significant effect on the level of the total TACE protein (Fig. 1B). Although some nonspecific bands also underwent increased phosphorylation (Fig.  1A), they were also present in the IgG control immunoprecipitations and were not detected in Western blots using the TACE antiserum ( Fig. 1B). Furthermore, they were not competed out in immunoprecipitations in the presence of excess peptide immunogen (Fig. 1C). These results indicate that both the glycosylated pro-TACE and mature TACE are phosphorylated and that EGF, FGF, serum, and PMA, which are known to activate ectodomain shedding, rapidly induce phosphorylation of the TACE cytoplasmic domain.
Growth factor-or PMA-induced phosphorylation of TACE was also examined using CHO cells, transfected to express the cloned human TACE cDNA. CHO cells express a very low level of TACE, but transfection of TACE readily allowed detection of TACE by immunoprecipitation of 35 S-labeled ( Fig. 2A) or 32 Plabeled (Fig. 2B) proteins. Besides the 126-kDa precursor form of TACE and the 100-kDa mature TACE, we also detected a 112-kDa form ( Fig. 2A). Pulse-chase experiments suggested that the intermediate 112-kDa band is likely to be the unglycosylated TACE precursor (data not shown). Previous studies (47) suggest that TACE matures rather inefficiently. This may explain why the major TACE form detected in transfected CHO cells is the glycosylated full-length TACE precursor. FGF treatment for 15 min significantly increased the TACE phosphorylation level (Fig. 2B), without an effect on the level of TACE protein ( Fig. 2A). In these experiments, the phosphorylation was restricted to the larger, glycosylated TACE form (Fig. 2B). Fetal bovine serum and PMA, which also induce ectodomain shedding, also stimulated the level of TACE phosphorylation after 15 min (Fig. 2C). These results indicate that TACE, encoded by the characterized cDNA, is phosphorylated in response to FGF, serum, or PMA as observed with endogenous TACE.
SPRACT, a Polypeptide That Corresponds to the TACE Cytoplasmic Domain-Immunoprecipitation and Western blot analyses of HeLa S3 cell lysates using an antibody against the carboxyl-terminal sequence of TACE detected not only the two TACE forms mentioned above but also a 20-kDa protein, which we named SPRACT for "small protein reactive with antibody against the cytoplasmic domain of TACE" (Fig. 1). SPRACT was also detected in 293 human embryonic kidney cells and the U937, THP-1, and K562 hematopoietic cell lines by direct Western blotting or immunoprecipitation followed by Western blotting, using two additional antibodies against the cytoplasmic domain of TACE (data not shown). Additionally, SPRACT was expressed from the cloned TACE cDNA in transfected CHO cells (Fig. 2). Similarly to the TACE forms, SPRACT showed rapidly increased phosphorylation in response to EGF, FGF, serum, or PMA stimulation ( Fig. 1A and Fig. 2, B and C), whereas its protein levels remained constant (Figs. 1B and 2A). The detection of SPRACT by antibodies against the cytoplasmic domain of TACE in Western blot analyses rules out the possibility that SPRACT is a TACE-binding protein and suggests that SPRACT may correspond to a cytoplasmic segment of TACE. Consistent with this interpretation, SPRACT was not detected by an antibody that was raised against the extracellular domain of TACE (47) (data not shown). Furthermore, excess peptide to which the anti-TACE antiserum was raised competed out the signals of 32 P-labeled TACE and SPRACT in the extracts of PMA-stimulated HeLa S3 cells (Fig. 1C). It should be noted that SPRACT is poorly retained by the PVDF membrane. Up to 90% of SPRACT can be lost depending on the stringency of the wash condition (data not shown), and therefore the level of SPRACT, detected by Western blotting, may be underestimated.
SPRACT Is an Alternative Translation Product-It has been suggested that the active form, but not the full-length precursor form, of TACE can release its cytoplasmic domain through autolysis (47). We therefore included phenanthroline (10 M), which has been demonstrated to inhibit TACE autolysis (47) and yet also inhibits other metalloproteases, in the cell lysis buffer and throughout the subsequent immunoprecipitation and washes used to generate the results in Fig. 1. The detection of both the mature 100-kDa TACE and 20-kDa SPRACT under these conditions suggests that SPRACT is unlikely to be generated by autolysis during sample preparation. Previous studies (30,39) have shown that the metalloprotease inhibitor TAPI-1 inhibits the cleavage activity of TACE. Inclusion of TAPI-1 (10 M) and phenanthroline (10 mM) in the cell lysis buffer and throughout subsequent washes in immunoprecipitation did not decrease the level of SPRACT in such experiments (data not shown). Additionally, transfection with an expression vector encoding a catalytically inactive TACE mutant, due to replacement of Glu 406 with alanine, generated the same level of SPRACT as the expression vector for the wild type catalytically active TACE (Fig. 3A). Furthermore, inclusion of numerous other types of protease inhibitors in the cell A, HeLa S3 cells were serum-starved, labeled with [ 32 P]orthophosphate, and treated with the indicated stimuli. Cells were then lysed, and 32 P-labeled TACE was immunoprecipitated using an antibody recognizing the cytoplasmic domain of TACE obtained from QED Bioscience or normal rabbit IgG. The 128-kDa larger TACE form is the glycosylated TACE zymogen with its prodomain, whereas the 100-kDa band corresponds to the mature form without the prodomain. These TACE forms are marked as T in all figures. The 20-kDa protein is SPRACT, which is marked as S in all figures. B, stimulation with growth factors and PMA did not lead to significant changes of the level of TACE protein. HeLa S3 cells were serum-starved, treated with indicated stimuli, and lysed. Western blotting analysis was performed with the same antibody or control rabbit IgG as in A. C, the blocking peptide specifically interferes with the detection of phosphorylated TACE and SPRACT. 32 P labeling of HeLa S3 cells and immunoprecipitation of TACE were performed as described in A. For peptide competition, the estimated molar ratio of peptide to antibody is 1000 to 1, based on the average molecular mass of IgG as 180 kDa. Note only the TACE and SPRACT bands but not the nonspecific bands were blocked by the peptide in the cell lysate prepared from PMA-stimulated cells.
lysis and during processing of samples, as well as cell lysis using boiling lysis buffer, did not decrease the SPRACT level either (data not shown), further suggesting that SPRACT is unlikely to be derived through autolysis or proteolysis by other proteases during sample preparation.
During the construction of TACE expression plasmids, we unintentionally obtained an expression vector, now named pTshift, which contained a cytosine deletion, in the codon for Ser 681 located in the transmembrane domain, thus causing a frameshift and an immediate premature translation termination (Fig. 3B). The resulting truncated protein lacks the carboxyl-terminal portion of the transmembrane domain and the entire cytoplasmic domain and was detected by immunoprecipitation using an antibody raised against the extracellular domain of TACE (Fig. 3C). Nevertheless, the antibody against the TACE cytoplasmic domain demonstrated that SPRACT was also expressed from pT-shift, albeit with a somewhat larger molecular weight due to an added carboxyl-terminal FLAG epitope (Fig. 3D). We therefore conclude that translation of SPRACT did not depend on the synthesis of full-size TACE. In contrast, SPRACT was not expressed from a plasmid that encodes a cytoplasmically truncated TACE (the ⌬CD mutant) (Fig. 3, B and D). These data suggest that SPRACT is translated from the sequence encoding the cytoplasmic domain of TACE and argue further against the possibility that SPRACT is derived through proteolytic cleavage of TACE. In keeping with this, SPRACT is expressed by a plasmid that encodes the cytoplasmic domain of TACE without its extracellular and transmembrane domains, i.e. the ⌬ED/TM mutant (Fig. 3, B  and D).
The cytoplasmic domain of TACE has two juxtamembrane methionines (Met 715 and Met 719 , Fig. 3B). The corresponding ATG triplets are both located within a sequence, which, based on the consensus rules of Kozak (59), should provide efficient translational initiation. To test if these ATGs could serve as initiation sites for the translation of SPRACT, we constructed expression plasmids for full-size TACE with either or both ATG codons mutated to GCG codons for alanine (Fig. 3B). Whereas SPRACT is still expressed from TACE expression plasmids lacking either ATG, it is no longer produced when both juxtamembrane ATGs are mutated to encode alanines (Fig. 3D). Collectively, these data indicate that SPRACT is an alternative translation product that can be initiated at either juxtamembrane ATG codon in the sequence encoding TACE cytoplasmic domain.
Correlation of TACE and SPRACT Phosphorylation with the Erk MAP Kinase Signaling Pathway-We have shown previously that activation of the Erk MAP kinase pathway mediates growth factor-and PMA-induced ectodomain shedding of TGF-␣, TNF-␣, and L-selectin. Accordingly, ectodomain shedding in response to these inducers is inhibited by U0126, an inhibitor of MEK1/2 that consequently prevents activation of Erk MAP kinase (32). Our data now show that U0126 also inhibits growth factor-induced phosphorylation of TACE and SPRACT in HeLa S3 cells to a level that is similar to the basal phosphorylation level in the absence of stimulation (Fig. 4A). A similar inhibition of phosphorylation by U0126 was also apparent in CHO cells transfected to express TACE (data not shown). Thus, growth factor-induced phosphorylation of TACE and SPRACT is mediated through activation of the Erk MAP kinase pathway. U0126 also inhibited PMA-induced phosphorylation of TACE, but only minimally decreased the phosphorylation of SPRACT (Fig. 4A). This suggests that PMA-induced phosphorylation of SPRACT, and possibly TACE, utilizes an additional kinase pathway(s) different from the MEK/Erk MAP kinase signaling pathway.
The effect of U0126 on growth factor-induced phosphorylation also led us to evaluate the effects of constitutively activated forms of MEK1 and ERK2. For this, we cotransfected the TACE expression plasmid together with an expression plasmid for a constitutively active form of MEK1 or Erk2 into CHO cells. As shown in Fig. 4B, overexpression of either enzyme resulted in increased phosphorylation of TACE and SPRACT.
Erk is a proline-directed Ser/Thr kinase that can directly phosphorylate Thr-Pro and Ser-Pro with increased preference when a proline is located at the Ϫ2 position (60,61). Interestingly, Thr 735 -Pro 736 is found as part of a Pro-Gln-Thr 735 -Pro sequence in the cytoplasmic domain of TACE, raising the possibility that TACE might be a direct target for Erk MAP kinase. To test this, we transfected CHO cells with an expression vector for wild type TACE or a mutant in which the Thr 735 was substituted by alanine, and we assessed the phosphorylation of TACE and SPRACT by gel electrophoresis and subsequent quantitation (Fig. 4C). Replacement of Thr 735 by alanine did not abolish the growth factor-and PMA-induced phosphorylation of TACE and SPRACT. This induction of phosphorylation of the T735A mutant TACE was blocked by the MEK inhibitor U0126 (Fig. 4C), similarly to wild type TACE. This suggests that the phosphorylation of TACE and SPRACT does not result from a direct phosphorylation of TACE by Erk MAP kinase but is more likely mediated by another protein kinase downstream from the Erk MAP kinase. This result does not exclude the FIG. 3. SPRACT is generated by alternative translation. A, SPRACT is expressed from wild type (WT) TACE or the catalytically inactive E406A mutant TACE (catϪ). CHO cells were transfected with the TACE plasmids or control vector pRK5 and directly lysed in SDS-PAGE sample buffer and subjected to Western blot analysis using the same antibody as in Fig. 1. B, schematic presentation of the mutations introduced in the coding sequence of TACE cDNA. The solid box represents the transmembrane domain coding sequence. The proposed internal translation initiation codons for Met in the cytoplasmic domain were mutated to GCG to encode Ala. The deleted cytosine in the codon for the transmembrane Ser 681 in pT-shift is underscored in the wild type sequence. The premature stop codon as a result of the deletion and consequent frameshift in pT-shift is shown. The dotted area in pT-shift represents a carboxyl-terminal FLAG epitope tag. C, detection of truncated TACE proteins produced by pT-shift and ⌬CD. CHO cells transfected with expression vectors for the mutated TACE forms or wild type TACE or the control pRK5 plasmid were metabolically labeled with [ 35 S]cysteine/methionine, and the TACE forms were immunoprecipitated with an antibody against the extracellular domain of TACE (47) and visualized by autoradiography. These truncated TACE forms are denoted as t. Although three forms of wild type TACE were seen, only one prominent band for the truncated TACE forms was detected. D, analyses of TACE and SPRACT expression by vectors shown in B. CHO cells transfected with the mutated expression plasmids or the pRK5 vector were lysed and subjected to Western blotting as described in A. Note that the carboxyl-terminal FLAG epitope tag caused a slower migration of SPRACT produced from pT-shift.

FIG. 4. Shedding activators induce phosphorylation of TACE and SPRACT through the Erk MAP kinase signaling pathway.
A, the MEK inhibitor U0126 inhibits the phosphorylation of endogenous TACE and SPRACT in HeLa S3 cells in response to EGF, serum (FBS), or PMA. The samples from cells treated with medium only were marked as M. HeLa S3 cells were labeled with [ 32 P]orthophosphate. Indicated shedding activators were added either alone or together with U0126 at 15 min before termination of the in vivo labeling. 32 P-Phosphorylated TACE and SPRACT were immunoprecipitated as described in the legend to Fig. 1. B, stimulation of TACE and SPRACT phosphorylation by expression of activated MEK or Erk MAP kinase. Duplicate CHO cells were transfected with TACE alone or together with an expression vector for constitutively activated MEK1 (caMEK1) or Erk2 (caErk2) and labeled in vivo with [ 35 S]cysteine/methionine or [ 32 P]orthophosphate. Radioactive TACE and SPRACT were detected by immunoprecipitation using an antibody against the TACE cytoplasmic domain (52,53). The amounts of 35 S-labeled TACE detected in the samples were first quantitated by PhosphorImager analysis and then used to correct the amounts of 32 P-labeled cell extracts for immunoprecipitation. This normalization was required since cotransfection of the activated kinases increased the TACE expression (data not shown). C, U0126 inhibits the induction of the phosphorylation in TACE and SPRACT encoded by expression vectors for wild type TACE and the T735A mutant. In vivo 32 P labeling and detection of phosphorylated proteins were performed as A. The top panel shows the autoradiogram; the bottom panel shows relative intensities of the 32 P-labeled TACE and SPRACT bands as quantitated by densitometry. The intensities of the bands were normalized against the intensity of the TACE or SPRACT bands under non-stimulated conditions. possibility that Erk MAP kinase has the ability to directly phosphorylate this or another sequence in TACE.
Phosphorylation of Serines in the TACE Cytoplasmic Domain-The cytoplasmic domain of TACE contains 16 serines, 7 threonines, including Thr 735 , mentioned above, and 1 tyrosine residue (Fig. 5A). Phosphoamino acid analysis of hydrolyzed protein showed that the endogenous forms of TACE isolated from EGF-treated HeLa S3 cells were phosphorylated on serine but not on threonine or tyrosine (Fig. 5B). In addition, only phosphorylated serine was detected in phosphoamino acid analyses of TACE in FBS-stimulated, transfected CHO cells (Fig. 5C). The absence of threonine phosphorylation further supports our conclusion that growth factor-induced phosphorylation does not result from direct phosphorylation of Thr 735 by Erk1/2 MAP kinase.
We next set out to define which serines are phosphorylated in response to growth factor stimulation. The cytoplasmic domain of TACE contains 4 methionines including the 2 methionines at positions 715 and 719, which we propose to serve as the first amino acids of SPRACT (Fig. 5A). Cyanogen bromide cleavage, which occurs following a methionine, is expected to cleave following the cytoplasmic methionines at positions 715 and 719, 759, and 772. This should result in cytoplasmic fragments of 4 (Leu 716 -Met 719 ), 40 (Asp 720 -Met 759 ), 13 (Asp 760 -Met 772 ), and 51 (Asp 773 -Cys 824 ) amino acids, in addition to a 77-amino acid segment (Asn 638 -Met 715 ) that spans the transmembrane segment.
CHO cells transfected with the wild type TACE expression plasmid were 32 P-labeled in vivo in the absence or presence of serum, and 32 P-labeled TACE and SPRACT were isolated by preparative gel electrophoresis. Both TACE and SPRACT were digested using CNBr, and the 32 P-labeled digestion products were separated by gel electrophoresis (Fig. 5D). One or possibly two 32 P-labeled fragments with an apparent molecular mass of 4 -6 kDa were identified. These 32 P-labeled fragments were the same following digestion of TACE or SPRACT. We did not detect a fragment that was compatible with the 77-amino acid segment upstream from Met 715 and would only be present as a TACE and not a SPRACT digestion product. These results strongly suggest that the phosphorylation occurs similarly in both TACE and SPRACT and that no phosphorylation of serines occurs upstream from Met 715 . Additionally, no fragment that could correspond to the 13-amino acid fragment was detected, even though this fragment was expected to be resolved on gel. Thus, the Asp 760 -Met 772 segment is unlikely to be phosphorylated. The 4-amino acid Leu 716 -Met 719 fragment is not expected to be resolved on the gel. Taken together, these results suggest that serine phosphorylation occurs in either or both the Asp 720 -Met 759 and Asp 773 -Cys 824 segments.
Identification of Phosphorylated Serines in the TACE Cytoplasmic Domain-To identify the phosphorylation sites, we first gel-purified in vivo 32 P-labeled, full size TACE, obtained from starved or FGF-or serum-treated, transfected cells, and we compared their phosphopeptide maps following trypsin digestion. Comparison of FGF-treated with untreated cells revealed several 32 P-labeled peptides, one of which increased and another one decreased in intensity following FGF treatment (Fig. 6A). Serum stimulation resulted in enhanced or decreased 32 P-labeled intensities of the same two peptides, whereas some increases in intensities of other minor peptides were noted as well (Fig. 6A). A similar analysis of the corresponding gel segment from mock-transfected cells yielded only one spot, i.e. the one at the start position, as marked in Fig. 6A (data not   FIG. 5. Phosphoamino acid analysis and CNBr mapping of TACE phosphorylation. A, amino acid sequence of the carboxylterminal segment of TACE, composing the transmembrane and cytoplasmic domain. The transmembrane domain is underlined, and the methionine residues whose carboxyl peptide bond is targeted by CNBr are numbered according to their positions in the full-length TACE precursor and shown in boldface. All cytoplasmic serine residues are shown in boldface. In addition, the Ser 819 phosphorylation site and the Ser 791 dephosphorylation site, identified later, and Thr 735 are marked. B, EGF induces only serine but not threonine or tyrosine phosphorylation in both the zymogen (left) and mature (right) forms of TACE. In vivo 32 P labeling and precipitation of endogenous TACE were performed as described in Fig. 2A. TACE forms purified by SDS-PAGE were subjected to HCl hydrolysis. The resulting 32 P-phosphoamino acids were resolved by thin layer chromatography using phosphoserine (PS), phosphothreonine (PT), and phosphotyrosine (PY) as markers. C, serum induces the phosphorylation of overexpressed TACE on serine residue(s) only. In vivo 32 P-labeled TACE was immunoprecipitated from serum-stimulated CHO cells transfected to express TACE, purified by SDS-PAGE, and subjected to HCl hydrolysis. The resulting 32 P-phosphoamino acids were resolved by thin layer electrophoresis using the phosphoserine, phosphothreonine, and phosphotyrosine markers. D, CNBr mapping of in vivo 32 P-phosphorylated TACE and SPRACT. In vivo 32 P-phosphorylated TACE was immunoprecipitated from serumstimulated, transfected CHO cells, purified by SDS-PAGE, and subjected to CNBr hydrolysis. The resulting 32 P-peptides were resolved by SDS-PAGE and visualized by autoradiography. Goat anti-TACE polyclonal antibody was used for C and D.
FIG. 6. Tryptic peptide analysis of TACE phosphorylation. A, FGF or serum stimulation resulted in phosphorylation and dephosphorylation of different TACE peptides. In vivo 32 P-phosphorylated TACE from transfected CHO cells was immunoprecipitated, purified by SDS-PAGE, and subjected to in gel digestion with trypsin. The resulting 32 P-phosphorylated peptides were resolved by thin layer electrophoresis followed by thin layer chromatography and visualized by autoradiography. The origins of the samples are marked by dashed circles. The solid arrow points to the peptide whose phosphorylation level was increased after stimulation with FGF or serum, and the open arrow points to the peptide whose phosphorylation level was decreased after stimulation. The dephosphorylated peptide is rather "mobile" in the electric field, and we noted that its apparent isoelectric point is significantly affected by the temperature of the buffer for thin layer electrophoresis. B, phosphorylation, but not the dephosphorylation, of TACE peptides is inhibited by U0126, a MEK1/2 inhibitor. Samples were processed as in A except that U0126 was used to block serum-induced MEK activation.
shown). Together with phosphoamino acid analysis data shown in Fig. 5, B and C, these results suggest that growth factors induce phosphorylation and dephosphorylation of serine residues.
Because the U0126 inhibitor of MEK1/2 inhibited growth factor-induced shedding (32) and TACE phosphorylation (Fig.  4, A and C), we next assessed its effect on the phosphopeptide distribution. As shown in Fig. 6B, U0126 inhibited the growth factor-induced phosphorylation of the peptide, which normally shows an enhanced phosphorylation level, and did not visibly affect the growth factor-induced dephosphorylation of the other peptides. We therefore conclude that activation of the Erk MAP kinase pathway enhances the phosphorylation of that peptide in response to growth factor stimulation.
We next determined the identity of the two major phosphorylated peptides, i.e. the one with increased and the one with decreased phosphorylation in response to growth factor stimulation. We initially attempted to identify the phosphorylation sites by radioactive microsequencing, but we were unable to generate sufficient in vivo phosphorylated TACE for this purpose. We therefore initiated extensive mutagenesis of the serines following Met 715 , first in groups of several serine conversions into alanines. The G1 mutation of TACE resulted in replacement of Ser 717 , Ser 718 , and Ser 723 by alanines, whereas the G2 mutation replaced Ser 747 with an alanine. These serines are located within the 4-amino acid Leu 716 -Met 719 and the Asp 720 -Met 759 CnBr digestion fragments, respectively. Phosphoamino acid analyses revealed that these mutations did not affect the ability of serum to enhance or decrease the phosphorylation level of the two peptides, as detected using wild type TACE (data not shown). Two other clustered mutations addressed the phosphorylation states of the serines in the carboxyl-terminal Asp 773 -Cys 824 segment. The G3 mutation replaced Ser 785 , Ser 786 , and Ser 791 , whereas the G4 mutation replaced Ser 803 , Ser 808 , and Ser 819 by alanines. As shown in Fig. 7A the G3 mutation abolished 32 P labeling of the peptide, which normally shows decreased phosphorylation in response to FGF or serum stimulation, and did not affect growth factor-induced phosphorylation of the other peptide. In contrast, the G4 mutation abolished the phosphorylation of the peptide with increased phosphorylation in response to growth factor stimulation (Fig. 7B).
Subsequent analyses using single amino acid mutations allowed the identification of the serines that undergo growth factor-induced phosphorylation or dephosphorylation. In the case of the G3 clustered mutation, single amino acid substitutions identified Ser 791 as the phosphorylated residue, which shows decreased phosphorylation in response to growth factor stimulation. Thus, Ser 791 to alanine mutation abolished phospholabeling of the corresponding peptide in the absence of growth factor stimulation ( Fig. 7C; data not shown). Complementary single amino acid mutations led to the identification of Ser 819 as the amino acid that shows growth factor-induced phosphorylation. Accordingly, replacement of Ser 819 by alanine abolished phospholabeling of the peptide that shows increased phosphorylation in response to growth factor stimulation (Fig.  7D). These results indicate that growth factor stimulation induces the phosphorylation of TACE at Ser 819 and dephosphorylation of Ser 791 . Accordingly, the S819A mutation of TACE showed a low phosphorylation level, when compared with wild type TACE, in the absence of growth factor stimulation, and largely abolished the growth factor-induced phosphorylation of TACE and SPRACT (Fig. 7E).
Together with the data in Fig. 5B, our results strongly suggest that activation of the Erk MAP kinase pathway mediates the enhanced Ser 819 phosphorylation in response to growth factor stimulation but had little effect on the dephosphorylation of Ser 791 .
TACE Phosphorylation and SPRACT Do Not Affect Growth Factor-induced TGF-␣ Ectodomain Cleavage-The identification of Ser 819 as the major growth factor-induced phosphorylation site led us to assess its role in growth factor-induced TGF-␣ shedding. We used the EC2 cell line, which had been derived from genetically modified mice carrying the ⌬Zn/⌬Zn TACE and lacked functional TACE expression (31,42). Cells, transfected with a TGF-␣ expression plasmid, did not show growth factor-induced TGF-␣ release, as measured using our previously established TGF-␣ ectodomain shedding assay. In contrast, wild type TACE expression conferred a basal level of TGF-␣ release, which was further enhanced in response to serum stimulation (Fig. 8A). Similarly, the S819A and the G4 mutants of TACE were also able to confer growth factor-induced TGF-␣ ectodomain shedding. Furthermore, a TACE mutant with its entire cytoplasmic domain, except for the proximal two amino acids deleted, was also able to confer serum-induced TGF-␣ cleavage (Fig. 8A). We therefore concluded that Ser 819 phosphorylation in response to growth factor stimulation or activation of the Erk MAP kinase pathway was not required for growth factor-induced ectodomain shedding by TACE. Furthermore, ectodomain shedding by TACE did not require its cytoplasmic domain.
To evaluate a possible regulatory role of SPRACT in ectodomain shedding, we assessed the effect of overexpressed SPRACT on TGF-␣ ectodomain shedding in transfected CHO cells. In these cells TGF-␣ ectodomain shedding depended on the endogenous proteases, because an expression plasmid of TACE was not cotransfected. As shown in Fig. 8B, the basal and serum-induced TGF-␣ ectodomain shedding was not affected by SPRACT overexpression.
We also assessed a possible role of SPRACT using transfections of EC2 cells. Cells were transfected with an expression vector for wild type TACE, thus generating full-size TACE and SPRACT or the M715A/M719A mutant of TACE, which only expresses TACE but not SPRACT (Fig. 8C, lanes 2 and 3). As is apparent from Fig. 8D, both versions of the TACE expression vectors conferred an equal ability to induce basal and growth factor-induced ectodomain shedding of TGF-␣. We also cotransfected an expression plasmid for the M715A/M719A mutant of TACE with one that only expresses SPRACT, i.e. the ⌬ED/TM plasmid (Fig. 3, B and C). As is apparent from the 4th lane in Fig. 8C, this cotransfection resulted in high expression of SPRACT and a lower level of the M715A/M719A mutant of TACE, when compared with the expression of TACE in the absence of overexpressed SPRACT. This lower level of M715A/ M719A mutant TACE expression in the latter cotransfection experiment may be an artifact, because expression of SPRACT did not affect endogenous TACE expression in HEK293 cells (data not shown). As shown in Fig. 8D, coexpression of SPRACT did not affect the basal and growth factor-induced levels of TGF-␣ ectodomain shedding, as observed with the M715A/M719A mutant of TACE or wild type TACE. Together, these data do not allow us to conclude that coexpression of SPRACT has an effect on ectodomain shedding by TACE. DISCUSSION We demonstrated the growth factor-induced phosphorylation of the TACE cytoplasmic domain on serine. We also identified an alternative translation product, named SPRACT, that corresponds to most of the TACE cytoplasmic domain. Similarly to TACE, SPRACT undergoes regulated phosphorylation. Our findings represent a first direct demonstration that a natural inducer of ectodomain shedding induces phosphorylation of an ADAM family sheddase. While this manuscript was in preparation, Diaz-Rodriguez et al. (50) reported that PMA induces phosphorylation of the TACE cytoplasmic domain and proposed that PMA-and growth factor-induced phosphorylation occurs through direct phosphorylation of Thr 735 in the TACE cytoplasmic domain by Erk MAP kinase.
Both the glycosylated pro-TACE as well as mature TACE with its prosegment removed showed enhanced phosphorylation in response to growth factor or PMA stimulation. In transfected cells that overexpressed TACE only the pro-TACE form was visibly phosphorylated, but this may have resulted from the high predominance of pro-TACE and its inefficient maturation (47). It should be noted that phosphorylation of mature TACE was not demonstrated upon PMA stimulation in previous studies (45,50). The level of TACE phosphorylation correlated with its shedding activity following induction by growth factor or PMA stimulation.
We have shown that enhanced phosphorylation of TACE upon growth factor stimulation required induction of the Erk MAP kinase signaling pathway. Accordingly, U0126, an inhibitor of MEK1/2 which prevents the activation of Erk MAP kinase, inhibits growth factor-induced phosphorylation of TACE and SPRACT to a level similar to that in the absence of stimulation. Our previous results demonstrated that activation of the Erk MAP kinase pathway mediates growth factor-induced shedding of TGF-␣, TNF-␣, and L-selectin. Therefore, activation of MEK1/2 and the downstream Erk MAP kinase is required for growth factor-induced phosphorylation of the TACE cytoplasmic domain and ectodomain cleavage (32). In contrast, inhibition of the MEK1/2 activity affected only minimally the PMA-induced phosphorylation of SPRACT and possibly TACE, suggesting the involvement of an additional kinase pathway different from the MEK/Erk MAP kinase signaling pathway in PMA-induced phosphorylation. In addition, constitutively active forms of either MEK1 or Erk2 induced enhanced TACE phosphorylation and activated ectodomain shedding.
We demonstrated growth factor-induced phosphorylation of the TACE cytoplasmic domain on serine only but not on threonine or tyrosine. Extensive mutagenesis identified Ser 819 as the major target for phosphorylation in response to FGF and serum and Ser 791 as the major site for growth factor-induced FIG. 8. TACE phosphorylation and SPRACT expression do not detectably affect TGF-␣ shedding by TACE. A, wild type (WT), the S819A, G4, and the cytoplasmic domain truncated TACE mutants provide equal shedding of TGF-␣. EC2 cells that lack endogenous functional TACE were cotransfected with expression vectors for transmembrane TGF-␣ and the TACE mutants, serum-starved overnight, pulsed-labeled with [ 35 S]cysteine/methionine, and chased with cysteine and methionine with or without dialyzed FBS as a shedding inducer. Soluble TGF-␣ released into medium was immunoprecipitated and quantitated. B, increased SPRACT expression did not affect TGF-␣ shedding by CHO cells. The cells were transfected with a TGF-␣ expression plasmid alone (ϪSPRACT) or together with a plasmid for the ⌬ED/TM mutant (ϩSPRACT) (see Fig.  3 for schematic sequence presentation). Pulse-chase analyses were carried out as described in A. C, expression of TACE and SPRACT by wild type and mutated TACE forms. EC2 cells transfected with the indicated expression plasmids were metabolically labeled with [ 35 S]cysteine/methionine and subjected to immunoprecipitation with antibody against the TACE cytoplasmic domain. D, SPRACT expression did not affect TGF-␣ shedding in EC2 cells. Cells were transfected to express wild type TACE or M715A/M719A TACE that does not generate SPRACT, with or without the ⌬ED/TM mutant that only expresses SPRACT (see C). Sample preparation and quantitation of shedding were done as in A and B.
dephosphorylation. Accordingly, the S819A mutant has lost the ability to increase the phosphorylation level upon growth factor stimulation. Additionally, the phosphorylation of Ser 791 is markedly reduced upon the treatment. The dephosphorylation of Ser 791 is not inhibited by U0126, suggesting that it is independent of the Erk MAP kinase pathway. In contrast, growth factor-induced Ser 819 phosphorylation is inhibited by U0126 and thus depends on activation of the Erk MAP kinase pathway.
The two phosphorylation sites that we identified in the human TACE can also be found in the mouse, rat, and hamster TACE sequences. Sequence comparison of the cytoplasmic domains of other adamalysins did not reveal conservation of a sequence similar to the one flanking the phosphorylated Ser 791 in TACE. Interestingly, the carboxyl-terminal phosphorylated Ser 819 is near a motif that was identified to interact with PTPH1, which was suggested by overexpression to down-regulate TACE expression and TNF-␣ shedding (49). It is therefore possible that the phosphorylation of Ser 819 plays a role in the regulation of TACE processing. In addition, we noticed that the Ser-Lys dimer that composes Ser 819 exists with unusually high frequencies in the ADAM family. At least 13 of the 30 adamalysins with cytoplasmic domains ranging from 4 (ADAM 26) (62) to 196 amino acids (ADAM 19) (63) have at least one Ser-Lys sequence, with ADAM 30 containing 6 Ser-Lys repeats in its 85 amino acid cytoplasmic domain (64). Therefore, whether this sequence functions as a signaling motif is worth addressing.
While this manuscript was in preparation, Diaz-Rodriguez et al. (50) demonstrated that TACE shows increased phosphorylation on serine and threonine in response to PMA, and that Thr 735 , located within a favorable MAP kinase consensus site, can be directly phosphorylated by Erk MAP kinase in response to PMA. We did not find phosphorylation on threonine in response to growth factor stimulation, and mutation of Thr 735 did not affect growth factor-or PMA-induced phosphorylation of TACE and SPRACT. The basis for this discrepancy is unclear but may be related to the use of different cell types and the use of PMA at a 1 M concentration by Diaz-Rodriguez et al. (50), whereas we used 20 nM PMA as inducer.
Our use of TACE and protease inhibitors, as well as defined experimental conditions and the use of a catalytically inactive TACE, suggested that SPRACT was not derived from autolytic or proteolytic degradation of TACE either in the cell or during sample preparation. Instead, SPRACT appears to be expressed through translational initiation from proximal ATG codons in the TACE cytoplasmic domain. Consequently, mutation of the two proximal ATGs encoding Met 715 and Met 719 abolished the expression of SPRACT. Initiation at internal ATGs within an open reading frame has been observed previously (65)(66)(67)(68)in several cellular mRNAs and, more commonly, in viral RNAs. The expression of SPRACT in untransfected human cell lines as well as in transfected cells suggests that SPRACT is a physiological endogenous protein. The poor retention of SPRACT by PVDF membrane may explain why it was not recognized previously. Similarly to TACE, SPRACT underwent regulated phosphorylation in response to growth factor or PMA stimulation, and CNBr cleavage yielded the same phosphorylated peptide(s) as TACE. These data suggest that SPRACT may be phosphorylated and dephosphorylated on the same residues as TACE, and raise the possibility of a regulatory role for SPRACT in the function of TACE.
We have been unable to define a function for SPRACT and for the growth factor-induced phosphorylation of TACE and SPRACT. Mutation of Ser 819 into Ala or replacement of Ser 803 , Ser 808 , and Ser 819 by alanines (the G4 mutation) did not de-tectably affect the basal and growth factor-induced shedding of TGF-␣ by TACE. Additionally, and consistent with previous findings using PMA as inducer (42,69), cytoplasmic truncation still allowed for basal and growth factor-induced TACE activation. Furthermore, increased levels of SPRACT did not affect TACE activity, and mutation in TACE to abolish SPRACT expression did not affect growth factor-induced shedding either. Together, these data suggest that growth factor-induced changes in the phosphorylation state of the TACE cytoplasmic domain do not regulate TACE activation. Nevertheless, more quantitative assessments and analyses using different ectodomain cleavage substrates are needed to evaluate further the role of SPRACT and of growth factor-induced phosphorylation of TACE. In line with this cautious argument, it has been shown that the disintegrin domain of TACE is required for shedding of interleukin 1 type II receptor but not of other proteins tested (42,69), and TACE displays differential efficiencies in cleaving TGF-␣ at the distal site and the proximal site (17).
Our analyses are further complicated by the fact that the available TACE-negative cells have been immortalized with activated Ras and Myc (31,42), thereby presumably strongly deregulating endogenous signaling pathways that normally regulate TACE presentation and function. In addition, although these cells do not express enzymatically active TACE, the strategy to inactivate the TACE gene in these cells is expected to only disrupt the catalytically required structure within the TACE ectodomain and to allow endogenous expression of an inactive form of TACE, which nevertheless maintains the regulatory cytoplasmic sequences. Alternatively, SPRACT and the growth factor-induced changes in phosphorylation of the TACE cytoplasmic domain may not regulate activation of TACE per se but rather play a role in transport, processing, maturation, or presentation of TACE. Future research will be required to address the role of the cytoplasmic domain of TACE and its regulated phosphorylation in the elaboration of TACE function and the regulation of ectodomain shedding.