Role of the Juxtamembrane Domains of the Transforming Growth Factor-α Precursor and the β-Amyloid Precursor Protein in Regulated Ectodomain Shedding*

Although regulated ectodomain shedding is a well known process that affects a large group of transmembrane molecules, it is not clear how the shedding system selects its substrates. Here we investigate the structural requirements for the regulated shedding of two substrates of the general shedding system, the transforming growth factor-α precursor, pro-TGF-α, and the β-amyloid precursor protein, β-APP. The ability of different regions of pro-TGF-α or β-APP to confer susceptibility to the shedding system was tested using as a reporter a transmembrane molecule that is not a substrate of this shedding system. For this purpose we chose the TGF-β accessory receptor, betaglycan, since genetic and biochemical evidence showed that betaglycan is not a substrate of the shedding system. We determined that replacement of the 14 extracellular amino acids adjacent to the transmembrane region of betaglycan with the corresponding regions of TGF-α or β-APP rendered betaglycan susceptible to ectodomain shedding. These domain swap constructs were cleaved in response to protein kinase C stimulation, and cleavage was prevented by the metalloprotease inhibitor TAPI, both effects being characteristic of the general shedding system. Domain swap constructs containing the transmembrane and/or the cytoplasmic domains of pro-TGF-α did not undergo regulated ectodomain cleavage. We conclude that despite a lack of sequence similarity, the extracellular regions of pro-TGF-α and β-APP immediately preceding their transmembrane domains are key determinants of ectodomain shedding.

A large, functionally and structurally heterogeneous group of transmembrane proteins can undergo cleavage and release of their extracellular domain into the medium. This proteolytic process is often referred to as "ectodomain shedding" and affects a large set of otherwise unrelated cell surface molecules such as growth factors, growth factor receptors, ectoenzymes, and cell adhesion molecules. In most cases described to date, ectodomain shedding is a regulated process that can be activated by protein kinase C (PKC) 1 activation and other mecha-nisms (1,2). Although the components of this shedding system have not been identified, recent reports have provided insights into the general characteristics of the shedding machinery. Mutant CHO cell lines selected for lack of regulated shedding of the membrane-anchored growth factor pro-TGF-␣ are also defective in the cleavage of unrelated molecules such as the cell adhesion molecule L-selectin, the ␣-receptor for interleukin-6, the ␤-amyloid precursor protein (␤-APP), and a variety of endogenous CHO cell surface proteins (3,4). Hydroxamic acidbased compounds developed as inhibitors of the shedding of the membrane-anchored growth factor TNF-␣ by a putative metalloprotease (5)(6)(7) can also block shedding of TNF-␣ receptors, pro-TGF-␣, ␤-APP, L-selectin, interleukin-6 receptor, and Fas ligand with similar potency in all cases (3, 8 -10). A few differences have been described in the shedding process of these diverse proteins. Among these, the cytoplasmic domains of pro-TGF-␣ and the 80-kDa TNF-␣ receptor are required for the shedding of these two molecules (11,12), whereas Kit ligand, interleukin-6 receptor, and ␤-APP, devoid of their cytoplasmic domains, are shed to the same extent as the wild type molecules (13)(14)(15). Taken together, these findings suggest the existence of a common shedding machinery acting on most transmembrane molecules whose ectodomain can be released into the medium via regulated proteolytic cleavage. Although small deletions in the membrane proximal segment abolish shedding (15)(16)(17)(18)(19), mutational analysis of residues around the cleavage site of ␤-APP (20), the 60-kDa TNF-␣ receptor (18), interleukin-6 receptor (15), L-selectin (19), and TNF-␣ (16) has shown a lack of strict sequence specificity for cleavage. Rather, cleavage occurs at a site located at a fixed distance from the membrane (13, 16, 18 -20). These observations raise questions about the role of the cleaved amino acid sequence as a determinant of cleavage specificity.
To investigate these questions, we determined whether any portion of pro-TGF-␣ or ␤-APP might confer the ability to be shed to a protein that normally is not subject to the pro-TGF-␣ and ␤-APP shedding system. As the test protein we chose the TGF-␤ accessory receptor, betaglycan. Betaglycan is a membrane-anchored proteoglycan that binds the growth factor TGF-␤ and facilitates its interaction with signaling receptors (21). Although betaglycan can be shed, we found in the present studies that this is a very slow process that has none of the characteristics of the shedding process described above for pro-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  TGF-␣, ␤-APP, and other proteins. Taking advantage of this fact, we generated chimeras between betaglycan and domains of TGF-␣ and ␤-APP and determined which domains can confer susceptibility to the shedding system. Here we report that short juxtamembrane sequences of TGF-␣ and ␤-APP endow betaglycan with the ability to be cleaved by the regulated shedding system.
Cells and Transfections-Wild type and CHO cell lines defective in pro-HA/TGF-␣ shedding have been described elsewhere (4). Wild type and mutant CHO cells were stably transfected with the pCEP-4 plasmid (Invitrogen) containing the cDNA encoding the different molecules used in this work using the calcium phosphate precipitate method. Transfectans were selected in 600 g/ml hygromicin (Calbiochem) and subcloned. For transient transfection of L17 cells (22), the various cDNA constructs were subcloned into pCMV5 vector and transfected using the DEAE dextran method as described (22).
Flow Cytometry Analysis-Cells expressing various constructs were washed with Dulbecco's modified Eagle's medium for 1 h at 37°C and then treated with or without 50 M TAPI-2 for 5 min and treated with or without 1 M PMA, 1 M calcium ionophore A23187, or 10% fetal bovine serum and/or TAPI as indicated for additional periods of time. Cells were then incubated for 45 min at 4°C with 10 g/ml anti-HA monoclonal antibody (12CA5, Babco) or 10 g/ml of anti-myc monoclonal antibody 9E10 (23) in phosphate-buffered saline (PBS) containing 5% bovine serum albumin and stained for 30 min at 4°C with fluorescein isothiocyanate-conjugated anti-mouse IgG (Becton Dickinson) in PBS containing 5% bovine serum albumin. Flow cytometry was done on a FACscan instrument and software (Becton Dickinson).
Metabolic Labeling and Immunoprecipitation-Approximately 2.10 6 exponentially growing CHO cells expressing various molecules were labeled for different periods of time with 250 or 1,000 Ci/ml [ 35 S]cysteine and 250 or 1,000 Ci/ml Tran 35 S-label (NEN Life Science Products) in methionine-and cysteine-free medium at 37°C. The label was chased in complete medium for various periods of time in the presence or absence of 50 M TAPI-2 and/or 1 M PMA, 1 M calcium ionophore A23187, or 10% fetal bovine serum as indicated. Cells were then washed with cold PBS and lysed in PBS containing 1% Nonidet P-40 and 5 mM EDTA (lysis buffer). Aliquots from the media and from cell lysates were immunoprecipitated with anti-HA (Babco) or anti-Myc monoclonal antibodies. Immune complexes were collected by incubation of cell lysates and media samples with protein A-Sepharose (for anti-HA antibody) or protein G-Sepharose (for anti-Myc antibody) for 45 min at 4°C, washed three times with PBS containing 0.1% Triton X-100 and 0.1% SDS, and analyzed by SDS-PAGE. For quantification of secreted material, specific bands corresponding to betaglycan or pro-TGF-␣ were excised from the gel, counted in a Beckman scintillation counter, and the amount of secreted material expressed as percentage relative to anti-HA or anti-myc immunoprecipitable counts at the end of the chase from cell lysates.
Construction and Expression of myc-tagged and Chimeric Betaglycan cDNAs-The pro-HA/TGF-␣ construct was described in (4). The myc/ betaglycan construct (myc/BG) was generated by inserting the myc epitope (EQKLISEEDL) six codons downstream of the putative signal sequence as described previously (21). HpaI and ClaI sites were introduced flanking the transmembrane domain of myc-tagged rat betaglycan (nucleotides 2399 and 2470 of wild type rat betaglycan, respectively) using a double-stranded oligonucleotide that replaced the original NcoI-AvrI fragment. The transmembrane and cytoplasmic domains of pro-TGF-␣ were amplified by polymerase chain reaction using as a template wild type rat pro-TGF-␣ and oligonucleotides containing HpaI and/or ClaI restriction sites as needed. The transmembrane, cytoplasmic, or transmembrane-cytoplasmic domains of pro-TGF-␣ were subcloned into a vector containing myc-tagged betaglycan with the HpaI and ClaI restriction sites using appropriate restriction enzymes. The resulting constructs were named myc/BG-T␣-BG, myc/BG-BG-T␣, and myc/BG-T␣-T␣, respectively (see Fig. 4). The juxtamembrane domains of pro-TGF-␣ or ␤-APP were introduced into myc/BG using two complementary oligonucleotides containing the coding sequence for amino acids 85-98 of rat pro-TGF-␣ or amino acids 592-606 of human ␤-APP 695 which were annealed and subcloned at the ends resulting from digesting a vector containing myc/BG (HpaI, ClaI) with HpaI and NcoI (located at nucleotide 2385 of the wild type betaglycan molecule).

Betaglycan Ectodomain Secretion Is a Slow Process
Unresponsive to PKC Activation-The ectodomain of betaglycan can be released to the medium by proteolytic cleavage at a site proximal to the transmembrane domain (21,24). However, this mechanism has been poorly characterized. To analyze the secretion of betaglycan, a vector encoding rat betaglycan tagged with a c-Myc epitope following the signal sequence was stably transfected into CHO cells. Immunoprecipitation of metabolically labeled myc/BG transfectants with anti-Myc antibody yielded two main products similar to those observed previously with wild type betaglycan. Based on previous characterization of the membrane-anchored forms (25), these products are identified as the heterogeneous 180 -250-kDa proteoglycan form and the 110-kDa betaglycan core protein devoid of glycosaminoglycan chains (Fig. 1A). Confirming previous results, a low level of myc/BG ectodomain was specifically immunoprecipitated from the conditioned media of these transfectants (see Fig. 1B).
Ectodomain shedding can be induced by activators of PKC, calcium ionophores, or serum factors, each acting, in part, through independent mechanisms (2). To find out whether the secretion of betaglycan can be activated by these agents, CHO cells expressing myc/BG were treated with the PKC activator PMA, the calcium ionophore A23187, or fetal bovine serum, and the levels of betaglycan in the cells and in media were analyzed. None of these agents was able to induce a decrease in the levels of cell surface myc/BG, as determined by immunoprecipitation of metabolically labeled cell lysates (Fig. 1A) and media (Fig. 1B)  immunofluorescence ( Fig. 2A). In contrast, these activators induced a marked decrease in the levels of cell surface pro-TGF-␣ (Figs. 1 and 2C), as reported previously (4). The proportion of betaglycan and pro-TGF-␣ released into the media by metabolically labeled transfectants was determined after a 7-h chase or 1-h chase, respectively. These different time periods were chosen because the half-life of betaglycan in CHO cells (approximately 4 h) is longer than that of pro-TGF-␣ (approximately 1.5 h). The amount of labeled betaglycan recovered from media after the chase was only 3.5-5% of the betaglycan labeled after the pulse (Fig. 1B). This level of soluble betaglycan was not increased significantly by cell treatment with the various agents. A slight increase in secretion in PMA-treated cells observed in the experiment of Fig. 2B was not reproducible. Under the same conditions, and in agreement with previous reports, PMA, A23187, and serum induced a nearly quantitative release of the TGF-␣ labeled during the pulse (Fig. 1B). These results indicate that betaglycan ectodomain release is a slow process in CHO cells and cannot be activated by agents that typically activate the shedding of several transmembrane molecules such as pro-TGF-␣.
Betaglycan and TGF-␣ Are Secreted via Different Mecha-nisms-To determine whether the limited release of betaglycan is mediated by the pro-TGF-␣ shedding system or by a different process, we used the CHO mutant cell line M1, which is deficient in the shedding of pro-TGF-␣ and all proteins tested to date, including ␤-APP, interleukin-6 receptor, L-selectin, and various endogenous CHO cell surface molecules of unknown identity (4). M1 cells were stably transfected with myc/BG, metabolically labeled with [ 35 S]cysteine and [ 35 S]methionine, and then chased with or without PMA. Betaglycan shedding in M1 cells was tested and quantified as described above for parental CHO transfectants. The amount of soluble betaglycan secreted by M1 cells was comparable to the amount of betaglycan secreted by wild type CHO cells (Fig. 2B). In contrast, TGF-␣ immunoprecipitation showed a failure of M1 cells to shed pro-TGF-␣ (Fig. 2C), as shown previously (4). Additionally, we tested the effect of the hydroxamic acid-based inhibitor TAPI-2 on betaglycan shedding. TAPI-2 inhibits the shedding of TNF-␣ and most transmembrane molecules tested so far (4, 6 -10, 26). However, TAPI-2 had no effect on the secretion of betaglycan (Fig. 2B). Collectively, these results indicate that betaglycan ectodomain shedding in CHO cells is an inefficient process that involves a mechanism distinct from the shedding system that acts on pro-TGF-␣ and other transmembrane molecules.
Juxtamembrane Sequences as Determinants of Shedding-To identify regions of pro-TGF-␣ and ␤-APP which may confer susceptibility to shedding, we generated a panel of betaglycan chimeras containing various pro-TGF-␣ or ␤-APP sequences (Fig. 3). The cytoplasmic domain, the transmembrane domain, or both domains of betaglycan were replaced by the complementary domains of pro-TGF-␣. Additionally, we generated a betaglycan construct with a replacement of 14 juxtamembrane amino acids with the corresponding sequence from pro-TGF-␣ (Fig. 3). To survey the shedding of these chimeric molecules efficiently, we used a highly transfectable clone (L17) of the Mv1Lu mink lung epithelial cell line in transient transfection assays (22). Transiently transfected, metabolically labeled L17 cells were chased for different periods of time with or without PMA addition, and shedding was monitored by immunoprecipitation of metabolically labeled products. The different chimeric betaglycan molecules showed the same biosynthesis profiles as wild type betaglycan (Fig. 4  and data not shown). In the case of the constructs BG-T␣-BG, BG-BG-T␣, and BG-T␣-T␣, approximately 1% of total myc antibody-immunoprecipitable counts at the end of the pulse were released to the medium at the end of the chase, and the same result was obtained with the wild type betaglycan construct transfected under the same conditions (data not shown). The amount of released betaglycan ectodomain did not change in response to PMA. In contrast, the betaglycan construct containing the juxtamembrane domain of pro-TGF-␣ released significant amounts of betaglycan ectodomain into the media. Quantitation of the immunoprecipitated material indicated that 40% of the metabolically labeled betaglycan was released. Furthermore, the rate and extent of the release were increased in response to PMA (Fig. 4). These results indicate that the juxtamembrane domain of TGF-␣ is sufficient to confer susceptibility to the general shedding system.
To confirm these results in a better controlled system, we stably transfected the construct BG-T␣ juxt into CHO cells. Additionally, we generated a related construct containing the juxtamembrane 14 amino acids of ␤-APP (construct BG-␤-APP juxt) and stably transfected this construct into CHO cells. As shown in Fig. 5A, PMA treatment of CHO cells expressing myc/BG-T␣ juxt or myc/BG-␤-APP juxt induced the release of the ectodomain of the chimeric molecules as detected by immunoprecipitation via the c-myc epitope. The shedding of both chimeras was abolished completely by treatment with 50 M TAPI-2 (data not shown).
We next analyzed the kinetics of ectodomain shedding of these two chimeric molecules and compared it with that of TGF-␣. In agreement with previous results, the levels of cell surface pro-TGF-␣ decreased dramatically soon after treatment with PMA as determined by FACS analysis (Fig. 5B). The levels of cell surface pro-TGF-␣ were decreased 4-fold after a 5-min of treatment with PMA and became undetectable after 10 min of PMA addition. The ectodomain of the chimeric molecules BG-T␣ juxt and BG-␤-APP juxt were shed from CHO cells with approximately the same kinetics as pro-TGF-␣. Furthermore, upon addition of medium lacking PMA, the original levels of cell surface pro-TGF-␣ and BG-T␣ juxt were recovered with similar kinetics (t 1/2 ϳ 2.5 h) (Fig. 5C). These results demonstrate that the juxtamembrane domains of pro-TGF-␣ and ␤-APP are similarly effective at conferring susceptibility to cleavage by the regulated shedding system. DISCUSSION Genetic and biochemical evidence points to the existence of a general regulated ectodomain shedding system acting on a wide variety of cell surface proteins. Examination of surfacelabeled membrane proteins in CHO cells indicated that the transmembrane proteins whose ectodomain is proteolytically released into the medium constitute a diverse group and account for 2-5% of the total cell surface protein in CHO cells (3). The vast majority of these proteins are shed in response to PKC activation, and their release is prevented by a mutation in a common component (3). To date only two proteins, colonystimulating factor-1 (9) and betaglycan (this report), have been reported to be shed by an independent mechanism. In the present report, we show that, as in the case of colony-stimulating factor-1, the shedding of betaglycan is not stimulated by well known activators of the general shedding system and is not inhibited by the general shedding inhibitor TAPI. Furthermore, in the case of betaglycan, the rate and extent of release into the medium are very limited and not affected by mutations that disrupt the general shedding system.
Since betaglycan is in principle accessible to the general shedding system, we have used it as a reporter to identify TGF-␣ and ␤-APP sequences that would confer susceptibility to the general shedding system. Using a panel of betaglycan/ TGF-␣ chimeras, we demonstrate that the juxtamembrane 14 amino acids of pro-TGF-␣, which contain the natural pro-TGF-␣ cleavage site, are sufficient to confer susceptibility to the general shedding system. Furthermore, since the juxtamembrane domain of ␤-APP is as effective as that of pro- TGF-␣ at supporting betaglycan shedding, we conclude that the short juxtamembrane regions of pro-TGF-␣ and ␤-APP are determinants of ectodomain shedding. In contrast to this role of the juxtamembrane region, the transmembrane region and the cytoplasmic region of pro-TGF-␣, tested jointly or separately, failed to support betaglycan ectodomain cleavage.
It has been noted previously that the preferred cleavage site for ectodomain shedding is located at a certain distance from the membrane. Combined with a lack of evidence that the primary sequence of the cleaved region is important for cleavage these findings have led to the notion that shedding occurs by the action of an enzyme or set of enzymes which are sterically restricted to substrates adjacent to the membrane but are otherwise broad in their sequence specificity (15)(16)(17)(18)(19). The present observations clearly demonstrate that some feature of the 14 amino acid juxtamembrane sequences of pro-TGF-␣ and ␤-APP, not present in the corresponding region of betaglycan, is necessary for cleavage. No sequence similarities can be found between these juxtamembrane regions of pro-TGF-␣ and ␤-APP except for the presence of a cluster of hydrophobic amino acids at or following the cleavage sites (see Fig. 4). However, mutations in this hydrophobic cluster have limited effect on the shedding of pro-TGF-␣ (27), suggesting that this feature is not sufficient for recognition by the shedding system. It is therefore possible that the key determinant is in the as yet unknown secondary structure of this region. Alternatively, it is conceivable that this region in pro-TGF-␣, ␤-APP, and other shed proteins might be disordered, and a lack of secondary structure renders them susceptible to the shedding system.
Previous reports have emphasized the importance of the ␤-APP membrane proximal segment on the basal cleavage of this molecule (20,28), and recently it has been proposed that PKC activators enhance secretion of ␤-APP by enhancing budding of transport vesicles from the trans-Golgi network (29). Our observation that the juxtamembrane domain of ␤-APP introduced in betaglycan is sufficient to allow induced shedding of the chimeric molecule suggests that the proteolytic component of the ␤-APP shedding enzyme, often referred to as ␣-secretase, is activated via PKC. Stimulation of ␤-APP transport to the cell surface and activation of the ␣-secretase-mediated shedding are therefore two distinct and complementary mechanisms for activation of ␤-APP release by PKC.
In summary, our evidence suggests that a general shedding system regulated via PKC acts on diverse transmembrane proteins and selects their targets by recognition of a certain secondary structure (or a lack thereof) located in the juxtamembrane domain. Although the diversity of sequences cleaved by this system suggests the involvement of multiple proteases, it would appear from the present observations and previous results with hydroxamic acid inhibitors that these enzymes may be structurally and functionally related to each other.