Evaluation of the Contribution of Different ADAMs to Tumor Necrosis Factor α (TNFα) Shedding and of the Function of the TNFα Ectodomain in Ensuring Selective Stimulated Shedding by the TNFα Convertase (TACE/ADAM17)*

Tumor necrosis factor-α (TNFα), a potent pro-inflammatory cytokine, is released from cells by proteolytic cleavage of a membrane-anchored precursor. The TNF-α converting enzyme (TACE; a disintegrin and metalloprotease17; ADAM17) is known to have a key role in the ectodomain shedding of TNFα in several cell types. However, because purified ADAMs 9, 10, and 19 can also cleave a peptide corresponding to the TNFα cleavage site in vitro, these enzymes are considered to be candidate TNFα sheddases as well. In this study we used cells lacking ADAMs 9, 10, 17 (TACE), or 19 to address the relative contribution of these ADAMs to TNFα shedding in cell-based assays. Our results corroborate that ADAM17, but not ADAM9, -10, or -19, is critical for phorbol ester- and pervanadate-stimulated release of TNFα in mouse embryonic fibroblasts. However, overexpression of ADAM19 increased the constitutive release of TNFα, whereas overexpression of ADAM9 or ADAM10 did not. This suggests that ADAM19 may contribute to TNFα shedding, especially in cells or tissues where it is highly expressed. Furthermore, we used mutagenesis of TNFα to explore which domains are important for its stimulated processing by ADAM17. We found that the cleavage site of TNFα is necessary and sufficient for cleavage by ADAM17. In addition, the ectodomain of TNFα makes an unexpected contribution to the selective cleavage of TNFα by ADAM17: it prevents one or more other enzymes from cleaving TNFα following PMA stimulation. Thus, selective stimulated processing of TNFα by ADAM17 in cells depends on the presence of an appropriate cleavage site as well as the inhibitory role of the TNF ectodomain toward other enzymes that can process this site.

TNF␣ 1 is a pro-inflammatory cytokine that has a critical role in autoimmune disorders such as rheumatoid arthritis and Crohn's disease (1,2). TNF␣ is synthesized as a trimeric type II membrane-anchored precursor referred to as pro-TNF␣ (3). Upon cleavage in the juxtamembrane domain, the mature form of TNF␣ is released from the cell and can enter the blood stream (4,5). This proteolytic release of TNF␣ from the membrane is referred to as "protein ectodomain shedding" (6,7). Protein ectodomain shedding also affects the function of a variety of other structurally and functionally diverse molecules on the cell surface, including cytokines and growth factors, their receptors, adhesion proteins, and other molecules, such as the amyloid precursor protein, Notch and Delta (6 -9).
Because of the critical role of TNF␣ in rheumatoid arthritis, considerable efforts have been made to identify the TNF␣ convertase. ADAM17 (a disintegrin and metalloprotease 17, also referred to as TNF␣ converting enzyme or TACE) is considered to be an important, if not the major, sheddase for TNF␣ (10,11). ADAM17 was initially purified based on its ability to process a peptide, which mimics the physiological cleavage site of TNF␣, in exactly the same position that is used by the TNF␣ converting activity in cells (see Table I and Refs. 10 and 11). A targeted deletion of ADAM17 in mice revealed a critical role in TNF␣ shedding in T cells (11). ADAM17 has also been shown to be the major sheddase of several other proteins, including transforming growth factor ␣ (12, 13), heparin-binding epidermal growth factor-like growth factor (13)(14)(15), fractalkine (16), p75 neurotrophin receptor (17), and MUC1 (18). Besides ADAM17, additional candidate TNF␣ convertases have emerged from biochemical studies. ADAM10, which is most closely related to ADAM17, also can cleave a TNF␣ peptide at the physiological cleavage site in vitro (see Table I and Ref. 19). Furthermore, the TNF␣ cleavage site peptide can be processed by recombinant soluble forms of ADAMs 9 and 19 in vitro, although the cleavage sites for these ADAMs do not match the physiologically relevant site (see Table I and Refs. 20 and 21). In addition to these ADAMs, several other enzymes have been implicated in TNF␣ shedding. MMP7/matrilysin is critical for TNF␣ shedding in a mouse model for resorption of herniated discs (22), whereas the serine protease PR3 can release TNF␣ under conditions where it is highly expressed, such as in acute local inflammatory processes (23).
The ability of other ADAMs besides ADAM17 to cleave a TNF␣ peptide in vitro raises the question of whether these ADAMs may also contribute to TNF␣ shedding in vivo. To address this question, we evaluated TNF␣ shedding in cells isolated from wild-type, adam9Ϫ/Ϫ, adam10Ϫ/Ϫ, adam17Ϫ/Ϫ, and adam19Ϫ/Ϫ mice. In addition, we overexpressed different ADAMs to test which one(s) is(are) capable of processing pro-TNF␣ in cell-based assays. Our results confirmed that ADAM17 is the major stimulated sheddase of TNF␣, at least in mouse embryonic cells. In addition, we found that ADAM19 is also able to cleave TNF␣ and may therefore contribute to TNF␣ shedding in cells. Finally, we addressed the requirements for cleavage of TNF␣ by ADAM17 through testing how mutations in TNF␣ affect its shedding by ADAM17. These studies revealed that an inhibitory function of the TNF␣ module toward other enzymes contributes to the selective role of ADAM17 in stimulated cleavage of TNF␣.

MATERIALS AND METHODS
Reagents-All chemicals and reagents were purchased from Sigma unless otherwise indicated. Talon metal affinity resin was purchased from BD Biosciences, concanavalin A-Sepharose and Protein G-Sepharose were purchased from Amersham Biosciences. BB94 was kindly provided by J. D. Becherer (GlaxoSmithKline, Research Triangle Park, NC).
Expression Vectors-The pAP-TNF␣ and pAP-TRN (TRANCE) expression constructs have been described previously (21,24). In both constructs, an alkaline phosphatase moiety is attached to the C terminus of the full-length wild-type protein. To generate pAP-⌬TNF␣, a PCR fragment encoding the amino acids 1-87 of TNF␣ was cloned into pAPtag5 vector (Genehunter Corp.) between the NheI and BglII sites. The resulting mutant form of TNF␣ lacks the ectodomain but contains the juxtamembrane domain, including the cleavage site for ADAM17. The pAP-TNF-Ecto (TRN) construct encodes a chimeric molecule in which the ectodomain of TNF␣ is replaced with that of TRANCE/OPGL. It was generated from previously described chimeras (25) by subcloning into the pAPtag5 vector. pAP-TNF-Ins (TRN) was generated by cloning PCR fragments encoding the cytoplasmic domain of TNF␣ (amino acid residues 1-53), the membrane-proximal juxtamembrane domain of TRANCE (amino acid residues 72-115), and the juxtamembrane domain and ectodomain of TNF␣ (amino acid residues 54 -233) into pAPtag5. pAP-TNF-Ins (TRN) gives rise to a mutant form of TNF␣ with a portion of the TRANCE juxtamembrane domain inserted between the transmembrane domain and the juxtamembrane domain of TNF␣. The pAP-⌬TNF-Ins (TRN) was generated by inserting PCR fragments encoding the cytoplasmic domain of TNF␣ (amino acid residues 1-53), the membrane-proximal juxtamembrane domain of TRANCE (amino acid residues 72-115), followed by the juxtamembrane domain of TNF␣ (amino acid residues 54 -87) into pAPtag5. A diagram of the constructs used in this study is presented below in Fig. 3A.
Cell Culture, Transfection, and Ectodomain Shedding Assays-CHO cells were maintained in F-12 medium with 5% (v/v) fetal bovine serum, 2 mM L-glutamine, 100 IU/ml penicillin G, and 100 g/ml streptomycin. COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 5% (v/v) fetal bovine serum, 2 mM L-glutamine, 100 IU/ml penicillin G, and 100 g/ml streptomycin. Primary mouse embryonic fibroblasts (mEFs) lacking one or more ADAMs were isolated from corresponding adamϪ/Ϫ knockout mice as previously described (15,21,26,27). mEFs and immortalized adam10Ϫ/Ϫ and adam10ϩ/Ϫ cells (27) were grown in Dulbecco's modified Eagle's medium with 10% (v/v) fetal bovine serum. Cells seeded in six-well tissue culture plates (Falcon) were transfected with the appropriate expression plasmids using LipofectAMINE2000 (Invitrogen). The transfection solution was removed after 5 h, and cells were allowed to recover in complete medium overnight. To measure shedding of the introduced AP-fusion proteins under basal conditions, cells were washed once with PBS, and then cultured in Opti-Mem for 1 h. The Opti-Mem medium was collected and replaced with fresh Opti-Mem medium with 25 ng/ml phorbol 12-myristate 13-acetate (PMA), or 100 M pervanadate (PV), or 1 M of the hydroxamic acid-type metalloprotease inhibitor batimastat (BB94) at the indicated concentration for another hour to assess shedding under stimulated or inhibited conditions. The change in ectodomain shedding upon addition of each activator or inhibitor of shedding was analyzed as described previously (15,24). Finally, bafilomycin A, a specific inhibitor of vacuolar type Hϩ ATPase that prevents acidification of the lysosome, was added to cells to assess what role lysosomal protein degradation has in generating C-terminal stubs in cells expressing TNF␣ or ⌬TNF␣. After CHO cells were transfected with either TNF␣ or ⌬TNF␣ and had recovered overnight, they were incubated with or without 10 g/ml bafilomycin A1 for 1 h. Then the cells were lysed in cell lysis buffer, and the extracts were subjected to Western blot analysis as described below. All experiments were repeated at least three times with very similar results.
Western Blot and Immunoprecipitations-To prepare samples for Western blot analysis, transfected cells were washed with PBS, then lysed in 500 l of cell lysis buffer (PBS, pH 7.4, with 1% (v/v) Triton X-100 and 5 mM 1,10-phenanthroline) per well. Lysates were cleared by centrifugation at 13,000 ϫ g for 30 min then incubated with concanavalin A-Sepharose for 2 h at 4°C. After washing twice with cell lysis buffer and once with PBS, the beads were incubated with 2ϫ sample loading buffer at 95°C for 5 min. The samples were separated by 8% SDS-PAGE, transferred to nitrocellulose, incubated with antibodies against ADAM9 (28), ADAM10 (CHEMICON International), or ADAM19 (21), and developed using a chemiluminescence detection system as described previously (28). For immunoprecipitations, cell lysates prepared as described above were cleared by centrifugation at 13,000 ϫ g for 30 min and then incubated with anti-FLAG M2 monoclonal antibody (Sigma) overnight, followed by Protein G-Sepharose Fast Flow beads for 1 h at 4°C. After washing twice with cell lysis buffer and once with PBS, the beads were incubated with 2ϫ sample loading buffer containing 10 mM dithiothreitol at 95°C for 5 min. The samples were then separated by 15% SDS-PAGE, transferred to nitrocellulose, and subjected to Western analysis with anti-FLAG M5 antibodies as described previously (28).

TNF␣ Shedding in Cells Lacking
Candidate ADAM Sheddases-To address whether ADAMs 9, 10, or 19 make a significant or at least detectable contribution to the ectodomain shedding of TNF␣ in mouse embryonic fibroblasts, we compared TNF␣ shedding in cells derived from wild-type mice and mice lacking ADAMs 9, 10, 17, or 19 (12,26,27,29). As shown in Fig. 1A, TNF␣ was shed constitutively in wild-type primary mouse embryonic fibroblasts mEFs, and its release could be stimulated by addition of either 25 ng/ml PMA or 100 M pervanadate (PV). When constitutive and stimulated TNF␣ shedding was assessed in adam9Ϫ/Ϫ, adam10Ϫ/Ϫ, and adam19Ϫ/Ϫ cells, no difference compared with wild-type mEF was observed (Fig. 1A). Thus ADAMs 9, 10, and 19 are dispensable for the constitutive or stimulated shedding of TNF␣ in mouse embryonic cells, even though all three ADAMs are expressed in these cells (15,26,27,29). On the other hand, PMA-stimulated TNF␣ shedding was abolished, and PV-stimulated shedding was strongly reduced in adam17Ϫ/Ϫ mEF (Fig. 1B). Both PMA-and PV-stimulated shedding could be rescued by reintroduction of ADAM17 cDNA into adam17Ϫ/Ϫ mEFs (Fig. 1B). These findings confirm that ADAM17 has a critical role in PMA-and PV-stimulated shedding of TNF␣ in mEF cells. Interestingly, constitutive TNF␣ shedding was still observed in adam17Ϫ/Ϫ mEFs, although it was slightly increased after reintroduction of wild-type ADAM17 cDNA (Fig.  1B). This result suggests that ADAM17 as well as one or more other minor TNF␣ sheddases contribute to constitutive shedding of TNF␣ Effects of Overexpression of ADAMs 9, 10, 17, and 19 on TNF␣ Shedding-Although ADAMs 9, 10, and 19 are not essential for constitutive or stimulated shedding of TNF␣ in mEF cells, one or more of these ADAMs could conceivably contribute to shedding of TNF␣, especially in cells or tissues where they are highly expressed. To address this possibility, we tested whether overexpression of ADAMs 9, 10, 17, or 19 enhances constitutive or stimulated TNF␣ shedding in either CHO cells, COS-7 cells, or in adam10Ϫ/Ϫ or adam17Ϫ/Ϫ embryonic cells. When TNF␣ was co-expressed with wild-type ADAM9 in CHO or COS-7 cells, no change in its constitutive or stimulated shedding was observed in comparison to control experiments with the inactive mutant form ADAM9EϾA (Fig. 2, A and B, COS-7 cell data not shown). When ADAM19 was co-expressed with TNF␣ in COS-7 or CHO cells, this led to an increased constitutive shedding of TNF␣, whereas co-expression with the catalytically inactive ADAM19EϾA did not (Fig. 2, A and B). Separate Western blot experiments confirmed that the wildtype and EϾA mutant forms of ADAM9 as well as of ADAM19 were expressed at comparable levels in these experiments (data not shown). Interestingly, constitutive and stimulated shedding of TNF␣ was not noticeably increased when ADAM17 was co-expressed in CHO or COS-7 cells (Fig. 2, A and B).
Because the contribution of ADAM19 and other ADAMs to stimulated shedding may be obscured by the activity of ADAM17 in COS-7 and CHO cells, we also performed co-expression experiments of TNF␣ with different ADAMs in adam17Ϫ/Ϫ primary mEF and in immortalized adam17Ϫ/Ϫ mEF cells (referred to as E2 cells). Overexpression of ADAM19 in adam17Ϫ/Ϫ cells resulted in increased unstimulated shedding of TNF␣, yet no additional increase was achieved following stimulation with PMA (Fig. 2C). This suggests that the catalytic activity of ADAM19 is not sensitive to PMA stimulation, which is consistent with a previous study in which ADAM19dependent cleavage of TRANCE/OPGL was also not stimulated by PMA (29). As had been seen in CHO and COS-7 cells, overexpression of ADAM9 and ADAM9EϾA also did not affect constitutive or stimulated shedding of TNF␣ in adam17Ϫ/Ϫ cells (Fig. 2D), even though ADAM9 and ADAM9EϾA were expressed at similarly high levels (data not shown). In addition, we did not observe any noticeable increase in constitutive or stimulated shedding of TNF␣ when we overexpressed ADAM10 in adam17Ϫ/Ϫ cells (data not shown), which is consistent with the results of a previous study by Reddy et al. (30).
To further explore a potential role of ADAM10 in TNF␣ shedding, we repeated a similar experiment in adam10ϩ/Ϫ and adam10Ϫ/Ϫ cells. Shedding of the EGF-receptor ligand betacellulin was used as a positive control for the catalytic activity of ADAM10 (15). The upper panel in Fig. 2E shows soluble forms of alkaline-phosphatase-tagged betacellulin in the supernatant of transiently transfected adam10ϩ/Ϫ cells. The faster migrating form (marked by an arrow) is generated by ADAM10-dependent shedding: it is not detectable in the supernatant of adam10Ϫ/Ϫ cells, but is recovered when adam10Ϫ/Ϫ cells are rescued by co-transfection with wildtype ADAM10. The slower migrating form of betacellulin is presumably generated by an activity that is resistant to metalloprotease inhibitors (15). In a parallel experiment, co-expression of ADAM10 with TNF␣ in adam10Ϫ/Ϫ cells did not increase constitutive shedding of TNF␣ compared with adam10Ϫ/Ϫ cells expressing only TNF␣ (Fig. 2E, lower panel).
Evaluation of the Effect of Mutations in TNF␣ on Its Processing by ADAM17-Because the studies described above further support the notion that ADAM17 is the major stimulated TNF␣ sheddase, we set out to learn more about the domains in TNF␣ that are important for its cleavage by ADAM17. For this purpose, we introduced mutations into the ectodomain of TNF␣ and generated chimera between TNF␣ and TRANCE, a TNF family member that is not a substrate for ADAM17 (25) (Fig.  3A). In all cases, an alkaline phosphatase moiety was fused to the ectodomain of these type II transmembrane proteins to monitor their shedding. The shedding profile of the chimera lacking the TNF␣ ectodomain (⌬TNF␣) in CHO cells was similar to that of TNF␣ in that it could be stimulated by addition of PMA or pervanadate (Fig. 3B). In adam17Ϫ/Ϫ cells, the PMA-stimulated shedding of full-length TNF␣ was abrogated, but could be restored by reintroduction of wild-type ADAM17 ( Fig. 3C; see also Fig. 1B). Unexpectedly, when ⌬TNF␣ was introduced into adam17Ϫ/Ϫ cells, its shedding was stimulated by PMA even in the absence of ADAM17 (Fig. 3C). Rescue of adam17Ϫ/Ϫ cells by reintroduction of ADAM17 further increased the level of PMA-dependent as well as PV-dependent shedding of ⌬TNF␣ ( Fig. 3D and data not shown). Taken together, these results suggest that stimulated shedding of ⌬TNF␣ can be performed by ADAM17 and by an additional enzyme that is distinct from ADAM17 (or by several additional enzymes), whereas stimulated shedding of full-length TNF␣ depends almost completely on ADAM17 in mouse embryonic cells. Stimulated and constitutive shedding of ⌬TNF␣ in adam17Ϫ/Ϫ mEF cells is sensitive to treatment with the hydroxamate-based metalloprotease inhibitor batimastat (BB94, Fig. 3E), suggesting the involvement of one or more metalloproteases that are distinct from ADAM17. To learn more about how the ectodomain of TNF␣ may prevent other PMA-and PV-stimulatable enzymes from processing the TNF␣ cleavage site in cells, we evaluated the shedding of a mutant in which the ectodomain of TNF␣ was replaced by that of the TNF family member TRANCE (TNF-TRN). Shedding of this chimera could be stimulated by PV in adam17Ϫ/Ϫ cells (Fig. 3F), suggesting that the ectodomain of a different TNF family member is not sufficient to prevent stimulated processing of the TNF␣ juxtamembrane domain in adam17Ϫ/Ϫ cells. In addition, PV-stimulated shedding of TNF-TRN in adam17Ϫ/Ϫ cells could be further enhanced by co-expression of ADAM17 (Fig. 3F).
The ADAM17 cleavage site in TNF␣ is in close proximity to the transmembrane domain (20 amino acid residues from the membrane). To test whether the cleavage site of TNF␣ must be in this membrane-proximal position to be processed by ADAM17, we inserted the juxtamembrane domain of TRANCE (amino acid residues 72-115) between the transmembrane domain and the cleavage site of both TNF␣ and ⌬TNF␣ to create TNF-Ins(TRN) and ⌬TNF-Ins(TRN) (Fig. 4A). Because the juxtamembrane domain of TRANCE contains two cysteine residues that most likely form a disulfide bond, this insertion presumably places a peptide loop between the cleavage site of TNF␣ and its transmembrane domain. In CHO and COS-7 cells, both mutants behaved similarly to wild-type TNF␣ in that their shedding could be stimulated by PMA and PV (Fig.  4A). The released ectodomains of TNF-Ins(TRN) and ⌬TNF- FIG. 2. Evidence for a role of ADAM19, but not ADAMs 9 or 10 in shedding of TNF␣. A, effect of overexpressing different wild-type ADAMs as well as ADAMs carrying an inactivating EϾA mutation in their catalytic site on constitutive and PMA-stimulated TNF␣ shedding in CHO cells. No evident difference in the PMA-stimulated shedding of TNF␣ was observed when ADAMs 9, 9EϾA, 10, 17, 19, and 19 EϾA were co-expressed. However, in the presence of co-expressed ADAM19, constitutive TNF␣ shedding was slightly, but detectably increased. This type of an increase in TNF␣ shedding was never seen when any of the other wild-type or mutant ADAMs analyzed here were co-expressed. B, upper panel: corroboration of the increased constitutive shedding of TNF␣ in the presence of overexpressed ADAM19 through a side-by-side comparison of constitutive shedding in the presence of co-expressed ADAM9, ADAM9EϾA, ADAM10, ADAM17, ADAM17EϾA, ADAM19, ADAM19EϾA, or pcDNA3 as control. Constitutive shedding of TNF␣ was increased by overexpression of ADAM19, but not by overexpression of ADAM19 EϾA, or of other ADAMs or their catalytically inactive mutants. Lower panel: co-expression of ADAM19 or ADAM19EϾA with TNF␣ in COS-7 cells further substantiates the result obtained in CHO cells in a different cell line. Overexpression of ADAM19 in COS-7 cells also increases the levels of a shed form of TNF␣, which migrates slightly faster than TNF␣ shed by ADAM17. C, co-expression of TNF␣ with ADAM19 in immortalized adam17Ϫ/Ϫ mEF cells (E2 cells) results in increased constitutive shedding of TNF␣ compared with controls in which TNF␣ was co-expressed with ADAM19 EϾA or with the empty vector (pcDNA3). The more pronounced effect of co-expression of ADAM19 on TNF␣ shedding in E2 cells compared with COS-7 cells and CHO cells is presumably due to the absence of ADAM17, which also participates in constitutive TNF␣ shedding (see Fig. 1B). D, TNF␣ was co-transfected with vector, ADAM9 or ADAM9EϾA in adam17Ϫ/Ϫ E2 cells. Similarly high expression levels of ADAM9 and ADAM9EϾA were confirmed separately by Western blot analysis (data not shown), and shedding of EGF by overexpressed ADAM9 in parallel experiments was used to verify that ADAM9 is active in cell based assays (data not shown). No difference in constitutive or stimulated shedding of TNF␣ was observed in cells expressing ADAM9 versus the inactive ADAM9 EϾA. E, shedding of betacellulin and TNF␣ in adam10ϩ/Ϫ cells and in adam10Ϫ/Ϫ cells with or without co-transfected ADAM10. Upper panel: experiments with betacellulin, a known substrate of ADAM10 (15), are included as positive control for ADAM10 activity. The shed form of betacellulin generated by ADAM10 (marked by arrow) is present in adam10ϩ/Ϫ cells, but not in adam10Ϫ/Ϫ cells. ADAM10-dependent shedding of betacellulin in adam10Ϫ/Ϫ cells can be rescued by cotransfection with ADAM10 (see also Ref. 15), confirming that ADAM10 is active in these experiments. The slower migrating shed form of betacellulin, which is the major form seen in adam10Ϫ/Ϫ cells, is generated by a separate activity that is not sensitive to BB94 (15). Lower panel: under identical conditions, no difference in constitutive or PMA-dependent shedding of TNF␣ was seen in adam10Ϫ/Ϫ cells in the presence or absence of reintroduced ADAM10.  Fig. 1B). In contrast, PMA-stimulated shedding of ⌬TNF␣ occurs in adam17Ϫ/Ϫ mEF cells that have not been rescued with wild-type ADAM17. This result suggests that the ectodomain of TNF␣ can inhibit PMA-dependent processing of TNF␣ by an activity or activities that are distinct from ADAM17. D, the PMA-induced shedding of ⌬TNF␣ in adam17Ϫ/Ϫ cells can be further Ins(TRN) co-migrate with the shed ectodomains of TNF-AP or ⌬TNF-AP, respectively, suggesting that a similar or identical cleavage site is used in all cases (data not shown). Furthermore, the apparent mass of the membrane-anchored C-terminal stubs generated by ectodomain shedding of TNF-Ins(TRN) and ⌬TNF-Ins(TRN) was increased by about 5 kDa compared with the C-terminal stubs of wild-type TNF␣ or ⌬TNF␣ (asterisk in Fig. 4B). Because the predicted mass of the inserted juxtamembrane fragment of TRANCE is 5065 Da, this further supports the idea that both insertion mutants are cleaved at or near the TNF␣ cleavage site for ADAM17 (Table I). However, slightly lower levels of the C-terminal stubs were generated from TNF-Ins(TRN) (Fig. 4B, lane 3) compared with TNF␣ (Fig. 4B, lane 1), indicating that processing of the TNF cleavage site is less efficient when it is placed further away from the transmembrane domain. Finally, it should be noted that, under conditions where the shedding levels of TNF␣ and ⌬TNF␣ into the supernatant as well as the expression levels of the fulllength proteins were quite comparable (see Figs. 3B and 4B,  lanes 1 and 2), significantly higher levels of C-terminal membrane stubs were generated in cells expressing ⌬TNF␣ compared with cells expressing TNF␣ (Fig. 4B, lanes 1 and 2). A similar increase in production of C-terminal stubs was also seen in cells expressing ⌬TNF-Ins(TRN) versus TNF-Ins(TRN) (Fig. 4B, lanes 3 and 4). Treatment of cells expressing ⌬TNF␣ or TNF␣ with bafilomycin A1, which blocks acidification of the lysosome and thus lysosomal protein degradation, increased the levels of the TNF␣ C-terminal stubs but did not detectably alter the levels of the ⌬TNF␣ C-terminal stubs (Fig. 4C). This suggests that the TNF module somehow affects the pathway by which the C-terminal stub is degraded.
The shedding profile of TNF-Ins(TRN) was similar to that of TNF␣ (see above) in that its shedding was not stimulated by PMA or PV in adam17Ϫ/Ϫcells (Fig. 4D). However, PMA-and PV-stimulated shedding of TNF-Ins (TRN) in adam17Ϫ/Ϫ cells could be rescued by co-transfection with wt ADAM17 (Fig.  4D). This demonstrates that the cleavage site of TNF␣ does not have to be immediately adjacent to the transmembrane domain to be processed by ADAM17. The shedding profile of ⌬TNF-Ins(TRN), which lacks the TNF␣ ectodomain, resembled that of ⌬TNF␣ in that its shedding was stimulated by PMA and PV in adam17Ϫ/Ϫ cells (Fig. 4E, PV data not shown). Evidently, the TNF␣ ectodomain can prevent PMA-and PV-dependent processing of the TNF␣ cleavage site by enzymes other than ADAM17 even when the cleavage site is moved away from the plasma membrane. Finally, co-expression of ADAM17 with ⌬TNF-Ins(TRN) further enhanced its PMA-or PV-stimulated shedding in adam17Ϫ/Ϫ cells (Fig. 4E, PV data not shown).

DISCUSSION
Proteolytic release of TNF␣ from its membrane-anchored precursor is thought to play an important role in regulating the physiological and pathological functions of this pro-inflammatory cytokine (3,4,10,11,31,32). The TNF␣ convertase (ADAM17) has been shown to be critical for processing TNF␣ (10,11), yet in vitro studies have suggested that ADAMs 9, 10, and 19 may also participate in the release of TNF␣ from cells (19 -21). Here we used mouse embryonic fibroblasts from adam9Ϫ/Ϫ, adam10Ϫ/Ϫ, and adam19Ϫ/Ϫ mice to evaluate whether these three ADAMs might also participate in TNF␣ shedding in vivo. Even though ADAMs 9, 10, and 19 are expressed in mEFs (15), no detectable defect in constitutive or stimulated shedding of TNF␣ was found in adam9Ϫ/Ϫ, adam10Ϫ/Ϫ, or adam19Ϫ/Ϫ mEFs. These results demonstrate that ADAM17 is the major PMA-and pervanadatestimulated sheddase of TNF␣ in mEF cells.
Even though "loss of function" experiments did not uncover any evidence for a role of ADAMs 9, 10, or 19 in TNF␣ shedding in mEF cells, this does not rule out that one or more of these ADAMs might contribute to TNF␣ shedding, which might become more evident in cells or tissues where these ADAMs are highly expressed. However, to postulate that ADAMs 9, 10, or 19 might participate in TNF␣ shedding in vivo, it is important to provide evidence that they are capable of cleaving TNF␣ in a cell-based assay. In the case of ADAMs 9 and 10, we found no evidence for a contribution of these enzymes to constitutive or stimulated TNF␣ release in co-expression experiments with TNF␣. Taken together with the results of loss of function experiments with adam9Ϫ/Ϫ and adam10Ϫ/Ϫ cells, these findings argue against a role for ADAMs 9 and 10 in TNF␣ shedding in cells. Perhaps there are additional constraints on processing of the TNF␣ cleavage site in a more physiological context, i.e. when membrane-anchored TNF␣ is presented to a membrane-anchored ADAM in cells compared to an in vitro situation, where a soluble peptide is cleaved by a soluble recombinant ADAM. Nevertheless, these results can not completely rule out the possibility that ADAMs 9 and 10 may cleave TNF␣ in other cells or tissues, for example in the presence of putative cofactors that might not be expressed in mEF cells.
When ADAM 19 was co-expressed with TNF␣ in CHO cells, there was a clear increase in constitutive, but not in stimulated TNF␣ release. This is consistent with the results of a previous characterization of ADAM19, in which it was found to be a constitutively active enzyme that was not significantly stimulated by PMA in cell-based assays (29). Some soluble TNF␣ generated by ADAM19 migrates slightly faster than soluble TNF␣ released by ADAM17, indicating that ADAM19 cleaves TNF␣ at a different site than ADAM17. This finding can most likely be explained by the observation that ADAM17 and ADAM19 cleave the TNF␣ cleavage site peptide in different positions in vitro as well (see Table I). Taken together, these results demonstrate that ADAM19 can function as a constitutive TNF␣ sheddase in cells. ADAM19 could thus conceivably make more significant contributions to constitutive TNF␣ shedding in cells and tissues in which it is highly expressed, such as for example in the heart (29). Finally, overexpression of ADAM17 did not noticeably increase constitutive or stimulated shedding of TNF␣ in CHO cells or COS-7 cells. Evidently the endogenous levels of ADAM17 are not a rate-limiting factor in TNF␣ release. The lack of increase in constitutive shedding of enhanced by co-expressing wild-type ADAM17. This confirms that ADAM17 can process the cleavage site of TNF␣ in the absence of the TNF ectodomain. E, constitutive and stimulated shedding of ⌬TNF␣ from adam17Ϫ/Ϫ cells can be inhibited by 1 M of the hydroxamic acid-type metalloprotease inhibitor batimastat (BB94). F, shedding profile of TNF-TRN (see panel A), which is similar to that of ⌬TNF␣ in adam17Ϫ/Ϫ cells. TNF-TRN shedding can be stimulated by PV, and stimulated shedding is further increased by co-expression ADAM17 WT in adam17Ϫ/Ϫ cells. It should be noted that, because PMA-and PV-stimulated shedding of TNF␣ is almost completely abolished in adam17Ϫ/Ϫ cells, this study focuses on qualitative differences in the shedding profile of ⌬TNF␣versus TNF␣-constructs, i.e. whether or not there is a clearly detectable increase in shedding of these chimeric proteins after stimulation with PMA and PV. Nevertheless, semiquantitative information can be obtained by comparison of shedding levels for different samples within a given experiment. For example, the increase in stimulated shedding of constructs such as ⌬TNF␣ and TNF-TRN from adam17Ϫ/Ϫ cells after reintroduction of ADAM17 (see panels D and F, and Fig. 4E) was highly reproducible in any given experiment (n ϭ 4), even though there was some variability in the strength of the overall effect of PMA or PV stimulation between experiments.
TNF␣ when ADAM17 is overexpressed in CHO cells also suggests that the activity of ADAM17 is tightly regulated, even when it is overexpressed.
What sequences in TNF␣ contribute to its selective cleavage by ADAM17? The structure/function analysis of TNF␣ revealed an unexpected role of the TNF␣ ectodomain in this process. Whereas stimulated shedding of TNF␣ is abrogated in adam17Ϫ/Ϫ cells, shedding of ⌬TNF␣, which lacks the TNF␣ ectodomain, can be enhanced by PMA and PV in these cells. Evidently, the ectodomain of TNF␣ prevents one or more PMAand PV-stimulated, BB-94-sensitive enzymes that are distinct from ADAM17 from cutting the TNF␣ cleavage site in cells. PMA-stimulated shedding of ⌬TNF␣ was not affected in adam9Ϫ/Ϫ, adam15Ϫ/Ϫ, or adam19Ϫ/Ϫ cells (data not , probed with an antibody against the N-terminal (cytoplasmic) FLAG tag. The full-length proteins and membrane-anchored stubs are marked by arrows. The asterisk marks the C-terminal membrane stubs generated from TNF␣-Ins(TRN) and ⌬TNF␣-Ins(TRN), which are about 5 kDa larger then the C-terminal membrane stubs generated from TNF␣ and ⌬TNF␣. C, CHO cells transfected with TNF␣ or ⌬TNF␣ were treated with 10 g/ml bafilomycin A1 (an inhibitor of lysosomal acidification that blocks protein degradation in the lysosome). Bafilomycin A1 treatment resulted in increased levels of the TNF␣ C-terminal membrane stubs but had no evident effect on the levels of ⌬TNF␣ C-terminal membrane stubs. D, the shedding profile of TNF-Ins(TRN) resembles that of TNF␣ in adam17Ϫ/Ϫ mEF cells (see Figs. 1B and 3B): stimulated shedding of TNF-Ins(TRN) by PMA and PV is abrogated in adam17Ϫ/Ϫ mEF cells but can be rescued by co-transfection with wild-type ADAM17. E, the shedding profile of ⌬TNF-Ins(TRN) resembles that of ⌬TNF␣ in adam17Ϫ/Ϫ mEF cells (see Fig. 3D) in that it can be stimulated by PMA. Co-transfection of ADAM17 further enhances the PMA dependent increase of ⌬TNF shedding. shown), arguing against a major contribution of these ADAMs to cleaving ⌬TNF␣. Rescue of adam17Ϫ/Ϫ cells by transfection with wild-type ADAM17 further enhanced the PMA-stimulated ⌬TNF␣ release compared with cells co-transfected with a control vector. This demonstrates that ADAM17 can recognize and process the membrane-anchored cleavage site of TNF␣ in cells even in the absence of the TNF␣ ectodomain. In this context it is interesting to note that PMA treatment also enhances the ADAM17-dependent processing of a soluble TNF␣ cleavage site peptide added to the culture medium in a cell-based assay (33). Finally, the presence or absence of the TNF module also appears to affect the levels of C-terminal membrane-anchored stubs that are generated through ectodomain shedding. Because the increased levels of C-terminal stubs from ⌬TNF constructs is not reflected in a corresponding increase in constitutive ectodomain shedding from cells, this indicates that the TNF␣ module may have a role in the intracellular transport and/or degradation of the C-terminal stubs. This notion is further supported by the finding that inhibition of lysosomal protein degradation increased the level of C-terminal membrane stubs generated from TNF␣, but not from ⌬TNF␣. Further studies will be necessary to provide a better understanding of the role of the TNF␣ module in the subcellular transport and processing of TNF␣ and in the degradation of the C-terminal stubs.
Replacement of the TNF␣ ectodomain with the ectodomain of the closely related TNF family member, TRANCE, demonstrated that a different TNF family ectodomain is not sufficient to prevent stimulated processing at the TNF␣ cleavage site by enzymes that are distinct from ADAM17. A previous study has shown that deletions of eight or more amino acid residues in the juxtamembrane domain of TNF␣ abrogate stimulated shedding, presumably by placing the ectodomain so close to the transmembrane domain that ADAM17 can no longer gain access to its cleavage site (34). Interestingly, the inhibitory effect of the TNF␣ ectodomain toward enzymes that are distinct from ADAM17 is also seen if the ectodomain is placed farther away from the plasma membrane, by inserting the juxtamembrane domain of TRANCE between the cleavage site of TNF␣ and its transmembrane domain. Thus the distance from the transmembrane domain to the TNF␣ module does not appear to be responsible for preventing access of the other enzyme(s) to the TNF␣ cleavage site. Perhaps the TNF␣ ectodomain blocks processing of the TNF␣ cleavage site by enzymes other than ADAM17, because it is too close to their cleavage site(s), as opposed to being too close to the transmembrane domain. Alternatively, the inhibition of these other enzymes by the TNF␣ module could depend on specific properties of the TNF␣ trimer compared with the TRANCE trimer. All known TNF family members self-assemble into non-covalently associated trimers, in which the three monomers oligomerize around an axis of 3-fold rotational symmetry (2). Because the trimeric structure of the TNF␣ ectodomain is predicted to be different from that of TRANCE, this could lead to a different orientation or presentation of the TNF␣ cleavage site when the TNF␣ ectodomain is replaced by the TRANCE ectodomain, or when it is removed, allowing other enzymes to gain access to this site.
In summary, loss of function studies combined with rescue experiments corroborated that ADAM17 has a key role as a TNF␣ convertase in mEF cells. In addition, gain of function studies showed that ADAM19 can contribute to TNF␣ release, suggesting that it might make a more prominent contribution to TNF␣ release in cells where it is highly expressed. On the other hand, we found no positive evidence for a role of ADAMs 9 and 10 in TNF␣ shedding. Using ectodomain deletion mutants and domain swap mutants between TNF␣ and TRANCE, we demonstrated that the ectodomain of TNF␣ contributes to the selective stimulated cleavage of TNF␣ by ADAM17 by preventing one or more other metalloproteases from processing the TNF␣ cleavage site. Taken together, this study provides new insights into the ability of different candidate TNF␣ convertases to shed TNF␣ and into the unexpected contribution of the TNF␣ ectodomain in ensuring selective stimulated cleavage of TNF␣ by ADAM17.