Tumor Necrosis Factor- Converting Enzyme (TACE) Regulates Epidermal Growth Factor Receptor Ligand Availability*

We previously implicated tumor necrosis factorconverting enzyme (TACE/ADAM17) in the processing of the integral membrane precursor to soluble transforming growth factor(TGF), pro-TGF. Here we examined TGFprocessing in a physiologically relevant cell model, primary keratinocytes, showing that cells lacking TACE activity shed dramatically less TGFas compared with wild-type cultures and that TGFcleavage was partially restored by infection of TACE-deficient cells with TACE-encoding adenovirus. Moreover, cotransfection of TACE-deficient fibroblasts with proTGFand TACE cDNAs increased shedding of mature TGFwith concomitant conversion of cell-associated pro-TGFto a processed form. Purified TACE accurately cleaved pro-TGFin vitro at the N-terminal site and also cleaved a soluble form of pro-TGFcontaining only the ectodomain at the C-terminal site. In vitro, TACE accurately cleaved peptides corresponding to cleavage sites of several epidermal growth factor (EGF) family members, and transfection of TACE into TACEdeficient cells increased the shedding of amphiregulin and heparin-binding EGF (HB-EGF) proteins. Consistent with the hypothesis that TACE regulates EGF receptor (EGFR) ligand availability in vivo, mice heterozygous for Tace and homozygous for an impaired EGFR allele (wa-2) were born with open eyes significantly more often than Tace / Egfr counterparts. Collectively, these data support a broad role for TACE in the regulated shedding of EGFR ligands.

The proteolytic processing of growth factors and cytokines is a key regulatory mechanism controlling receptor-mediated signaling. For example, proteases activate latent forms of transforming growth factor-␤ (1) and hepatocyte growth factor (2), regulate interactions between insulin-like growth factors and their binding proteins (3), and mediate the release of numerous soluble growth factors from their membrane-anchored precursors (4). In the case of membrane-anchored growth factors, the proteolytic processing has been proposed to regulate the avail-ability of active, soluble forms, switch receptor signaling from autocrine or juxtacrine modes to paracrine or endocrine mechanisms, and/or influence the nature (e.g. duration) of the signaling event. Growth factors and cytokines that are released from membrane-anchored forms include members of the EGF 1 superfamily, colony-stimulating factor-1, TNF-␣, and the Kit ligand (4).
The EGF superfamily includes two structurally related subfamilies: the EGF-like growth factors and the neuregulins (5). The EGF subfamily also includes TGF-␣, amphiregulin (AR), HB-EGF, betacellulin, and epiregulin as well as the recently described epigen (5,6). Soluble EGF family growth factors are all derived by proteolytic cleavage of the ectodomains of integral membrane precursors. However, the precursors display no significant homology outside of the 40 -50-amino acid EGF-like motif that is the shared bioactive and structural feature of this family. Proteolytic cleavage at the C terminus of the EGF-like sequence, required to release soluble forms, generally occurs within 15 amino acids of the transmembrane domain despite the absence of a consensus cleavage site. In contrast, cleavage at the N terminus of the EGF motif is variable. Many cell types release varying levels of larger bioactive forms of TGF-␣ in addition to the mature, 50-amino acid growth factor due to cleavage at the C-terminal site only (7,8). On the other hand, the EGF-like domains of HB-EGF and AR are typically released with N-terminal extensions rich in basic residues that confer heparin binding ability (9,10). In both cases, multiple cleavage sites that are N-terminal to the basic domains have been observed (11,12).
The proteolytic processing of EGF family members shares several characteristics with the general phenomenon of ectodomain shedding of cell surface proteins. It is a regulated event that can be rapidly induced upon exposure of cells to phorbol esters, calcium ionophores, serum factors, and phosphatase inhibitors (13)(14)(15), and it is sensitive to inhibitors of metalloproteinases (16 -19). However, until recently, the identity of the processing enzymes remained obscure. A major advance was the identification of a novel protease, tumor necrosis factor-␣ converting enzyme (TACE/ADAM17), responsible for converting membrane-anchored TNF-␣ to its soluble form (20,21). TACE is a member of a family of integral membrane proteins termed ADAMs (A Disintegrin and Metalloproteinase) that contain multiple conserved domains, including a catalytic domain with homology to the adamalysin family of metzincin proteases, an extracellular disintegrin sequence, and a cytoplasmic domain with putative signaling motifs (reviewed in Ref. 22).
The potential importance of TACE in EGF family shedding was demonstrated through a germ line mutation in mice that eliminated the zinc-binding domain (Tace ⌬Zn ), thereby inactivating the protease (23). Tace ⌬Zn/⌬Zn mice exhibited wavy hair and skin and eye defects identical to those observed in mice lacking transforming growth factor-␣ (TGF-␣) (24,25), and immortalized fibroblasts derived from Tace ⌬Zn/⌬Zn mice shed reduced levels of TGF-␣ into the medium (23). Moreover, TACE-deficient mice died perinatally, exhibiting widespread epithelial defects reminiscent of EGFR-deficient mice (26 -28). These observations suggested that TACE might regulate the availability of EGF family ligands (23,29).
In this report, we provide several lines of evidence derived from in vitro, cell-based, and whole animal studies that establish TACE as a major pro-TGF-␣ convertase. We also provide evidence consistent with a role for TACE in mediating the release of other EGF family members, especially AR and HB-EGF. Finally, we show that limiting the gene dosage of TACE exacerbates the phenotype of a weak EGFR allele (waved-2) in vivo. Collectively, these results indicate a broad role for TACE in regulating EGFR activity.

EXPERIMENTAL PROCEDURES
Materials-Monoclonal anti-FLAG M2 was obtained from Sigma, anti-HA was obtained from Covance (Berkeley, CA), and anti-Myc 9E10 was obtained from Sigma or Covance. Horseradish peroxidase-conjugated secondary antibodies were from Roche Molecular Biochemicals. Oligonucleotide primers were synthesized by the University of North Carolina Lineberger Comprehensive Cancer Center Nucleic Acid Core Facility.
Animals-All animals were housed at the University of North Carolina at Chapel Hill Department of Laboratory Animal Medicine. The Tace ⌬Zn mutation was carried on a C57BL/6 ϫ 129 hybrid background (23). Mice harboring the Egfr wa-2 mutation (B6EiC3H-a/A-Egfr wa-2/wa-2 Wnt3a vt ) and the corresponding hybrid control strain (B6EiC3H) were obtained from The Jackson Laboratory (Bar Harbor, ME). Tace ⌬Zn/⌬Zn and Egfr wa-2/wa-2 mice were initially identified phenotypically (open eye at birth and wavy hair, respectively) and subsequently verified by PCR amplification of genomic tail DNAs. Littermate controls were used for all experiments.
Primary Keratinocyte Preparation-Newborn mice were killed by halothane inhalation. Primary epidermal keratinocytes were prepared as described elsewhere (30) and cultured in Keratinocyte SFM (Invitrogen) with 10% dialyzed fetal bovine serum (FBS). For conditioning, cells were grown to near confluence, washed twice with 1ϫ phosphatebuffered saline, and shifted to 1% FBS medium containing 100 ng/ml EGF (added to improve recovery of soluble TGF-␣). Media were collected after 48 h and concentrated using Sep-Pak C-18 reverse phase cartridges (Waters, Milford, MA) as described previously (31). To collect cell lysates, cells were washed with 1ϫ phosphate-buffered saline and lysed on ice in 1% Triton X-100, 50 mM Tris, pH 7.4, 150 mM NaCl with 10 g/ml leupeptin, 20 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Protein concentration was determined using the Pierce BCA assay kit (Pierce), and TGF-␣ levels in media and lysates were analyzed by radioimmunoassay (RIA) (32).
Adenovirus Production and Infection-Recombinant adenoviruses were produced as described (33) using pAdTrackCMV and pAdEasy-1 adenoviral vectors kindly provided by Dr. Bert Vogelstein (The Johns Hopkins University). Ad TACE virus was created by subcloning a Cterminal Myc-tagged TACE cDNA into pAdTrackCMV. Primary epidermal keratinocytes from Tace ⌬Zn/⌬Zn mice were grown to 70% confluence and infected with control virus or Ad TACE virus at a multiplicity of infection of 5-10. Conditioned media and cell lysates were harvested after 24 h and analyzed for TGF-␣ content by RIA as described above.
Epitope Tagging-All cDNAs used in this study were expressed from pcDNA3 (Invitrogen). Epitope-tagged pro-TGF-␣ is described elsewhere (34). Pro-TGFecto was created from the full-length epitope-tagged construct by removing sequences between the TGF-␣ juxtamembrane domain and the C-terminal FLAG tag using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Mouse AR cDNA was generated by reverse transcription-PCR. Murine HB-EGF cDNA was kindly provided by Michael Klagsbrun (Harvard Medical School). Carl Blobel (Memorial Sloan-Kettering Cancer Center) kindly provided modified forms containing C-terminal Myc epitopes. We subsequently inserted HA tags near the N terminus of the EGF-like domain but C-terminal to the heparin-binding domain of each growth factor precursor using the QuikChange site-directed mutagenesis kit. A Myc tag was introduced at the C terminus of murine TACE using a PCR strategy. The resulting product was subcloned into the TACE cDNA, and the presence of the epitope was confirmed by DNA sequencing. The primers used for mutagenesis were: for pro-TGFecto, 5Ј-GGCTGCCAGCCAGA-AGAAGCAAGATTACAAGGACGACGATGACAAGG-3Ј; for AR, 5Ј-GG-CAGAAGGAATAAGAAGAAAAAGTACCCATACGACGTCCCAGACT-ACGCTAATCCATGCACTGCC-3Ј; for HB-EGF, 5Ј-GGGGTTAGGGAA-GAGATACCCATACGACGTCCCAGACTACGCTGACCCATGCCTCA-GG-3Ј; and for TACE: 5Ј-CCGAGATCTAGCAACAAGGTGTGTGGC-3Ј (forward) and 5Ј-GCTCTCGAGCTAAAGGTCCTCCTCGGAGATCAGC-TTCTGCTCGCACTCTGTCTCTTTGCT-3Ј (reverse).
Transfection and Western Blot Analysis-Wild-type and Tace ⌬Zn/⌬Zn fibroblasts (clones EC-4 and EC-2 (35)) were grown in Dulbecco's modified Eagle's medium-F12, 1% FBS and COS-1 cells in Dulbecco's modified Eagle's medium-H, 10% FBS. Cells were transfected in serum-free media with the indicated expression vectors using LipofectAMINE (Invitrogen) and, after 24 h, returned to complete media containing 100 ng/ml EGF. Conditioned media and lysates were collected 24 h later and treated as described above. Equivalent amounts of protein (50 -100 g) were separated by SDS-PAGE and transferred to Immobilon polyvinylidene difluoride (Millipore; Bedford, MA), and membranes were blocked in Tris-buffered saline, 0.1% Tween 20, 5% nonfat milk. Membranes were incubated with 1:1000 dilutions of primary antibody followed by 1:10,000 dilutions of peroxidase-conjugated secondary antibodies. Resulting bands were visualized using the Pierce SuperSignal West Pico chemiluminescent detection system and autoradiography.
In Vitro TACE Reactions-For N-terminal cleavage, lysate from a rodent cell line stably overexpressing epitope-tagged TGF-␣ (34) was incubated with ϳ200 units of partially purified native TACE (through the Mono-Q step in Ref. 20) from Chinese hamster ovary cells or with buffer only (1 unit of TACE cleaves 0.42 fmol/min of TNF peptide at 37°C, pH 7.5). Resulting products were identified by SDS-PAGE/Western blot as described above. For sequencing of TACE cleavage products, lysates from the same cells were pooled and enriched for pro-TGF-␣ using anti-FLAG M2 affinity resin (Sigma) in the presence of 0.1% Nonidet P-40 according to the manufacturer's instructions. Pro-TGF-␣ was eluted with FLAG peptide, and eluates were pooled and concentrated using Centricon-10 concentrators (Millipore). Concentrated pro-TGF-␣ was incubated with ϳ2 g of the recombinant human TACE extracellular domain (36). The products were separated by SDS-PAGE, transferred to Sequi-blot polyvinylidene difluoride (Bio-Rad), and stained with Coomassie Blue according to the manufacturer's instructions. The identity of Coomassie bands was confirmed by a parallel Western blot of a small portion of the reaction. The appropriate bands were excised from the membrane for N-terminal sequencing by Edman degradation (37) using an Applied Biosystems 494Ht sequencer.
For C-terminal cleavage, medium from COS-1 cells transfected with the pro-TGFecto construct was combined and concentrated via Sep-Pak C-18 cartridge as described above. Sep-Pak eluates were lyophilized, resuspended in 10 mM Tris, pH 8.0, and concentrated in Centricon-10 concentrators. Aliquots of the resulting solution were incubated with ϳ1 g of recombinant human TACE extracellular domain (36). For determination of the C-terminal cleavage site, reactions were stopped by addition of EDTA to a final concentration of 10 mM and immunoprecipitated using anti-FLAG M2 affinity gel (Sigma). The anti-FLAG beads were spotted directly onto a target, and the molecular weights of the peptides were determined at the University of North Carolina Proteomics Core Facility by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI/TOF-MS) using a Bruker Reflex III mass spectrometer (Bruker Instruments Co, Billerica, MA). 2 Peaks resulting from the TACE reaction were compared with those from a negative control (beads only) and a synthetic peptide, VVAASQKKQDYKDDDDKVV, matching the sequence of the expected product. To determine the sequence of the 1811-m/z product, beads were prepared as above and analyzed by MALDI-MS/MS on an API QSTAR-Pulsar mass spectrometer (Applied Biosystems, Foster City, CA). 3 Peptide Cleavage Assays-Synthetic peptides of the indicated sequences, based on the published sequences of precursor and mature forms of the human growth factors (10, 12, 19, 23, 38 -42), were incubated with the indicated concentrations of recombinant human TACE extracellular domain (36) for 4 h at 37°C. The resulting products were analyzed by LC/MS as described previously (23).
Statistical Analysis-Statistical analysis was performed with assistance from the UNC Lineberger Comprehensive Cancer Center's Biostatistics Shared Resource Group.

RESULTS
Efficient TGF-␣ Shedding Requires TACE-We showed previously that polyclonal Ras/Myc-transformed Tace ⌬Zn/⌬Zn embryonic fibroblasts shed about 20-fold less TGF-␣ into the medium as compared with wild-type counterparts (23). Here we compared TGF-␣ shedding by Tace ϩ/ϩ and Tace ⌬Zn/⌬Zn primary epidermal keratinocytes. RIA analysis showed that newborn Tace ⌬Zn/⌬Zn keratinocytes released 12-fold less TGF-␣ as compared with Tace ϩ/ϩ counterparts (Fig. 1). Northern blots revealed equivalent TGF-␣ mRNA levels in the two cell populations (not shown) excluding differences in TGF-␣ expression as an explanation for the difference in shedding. Both genotypes also exhibited similar amounts of keratin 5 (not shown), a marker of basal keratinocytes, indicating that differences in shedding were not accounted for by alterations in Tace ⌬Zn/⌬Zn keratinocyte differentiation. Consistent with a previous report that induced shedding of transfected TGF-␣ requires metalloproteinase activity (16), both TACE-dependent as well as the residual release of endogenous TGF-␣ in Tace ⌬Zn/⌬Zn keratinocytes were sensitive to the hydroxamic acid compound, IC-3 (20). Shedding by Tace ϩ/ϩ cells was inhibited 10-fold in the presence of IC-3, whereas release from Tace ⌬Zn/⌬Zn keratinocytes was inhibited Ͼ4-fold (Fig. 1).
To determine whether reintroduction of TACE restores pro-TGF-␣ processing, we infected Tace ⌬Zn/⌬Zn keratinocytes with either control adenovirus (Ad) (33) or virus-expressing mouse TACE cDNA (Ad TACE ). Staining for virus-encoded green fluorescent protein confirmed comparable infection by the two vectors. Ad TACE infection of Tace ⌬Zn/⌬Zn cells increased TGF-␣ shedding about 4-fold as compared with cells infected with Ad ( Fig. 1). The failure to fully restore TGF-␣ shedding may be due to incomplete infection of the cell population. It is unlikely that the incomplete restoration of TGF-␣ shedding reflects a dominant negative effect of the Tace ⌬Zn allele; the mutant TACE protein encoded by the Tace ⌬Zn allele does not function as a dominant negative since TNF and L-selectin were released at wild-type levels from Tace ϩ/⌬Zn cells (23).
Pro-TGF-␣ Is Converted to Mature Growth Factor upon Reintroduction of TACE-Besides mature, 50-amino acid TGF-␣, many cell types also release larger forms of the growth factor due to cleavage at the C-terminal site only (7,8). To determine whether reintroduction of TACE into Tace ⌬Zn/⌬Zn cells reconstituted complete processing of the TGF-␣ precursor, we transiently transfected these cells with pro-TGF-␣ containing HA and FLAG epitopes ( Fig. 2A). We previously used this construct to establish a pro-TGF-␣ processing scheme in rat liver epithelial cells (RLEC) in which an immature glycoprotein precursor of 22-25 kDa was converted to a mature, cell surface glycoprotein of 36 kDa; the latter was then cleaved to produce soluble TGF-␣ and a stable, cell-associated 16-kDa species corresponding to the residual tail of the precursor (34). Others have described similar pro-TGF-␣ processing schemes (43)(44)(45).
For these experiments, we utilized a clonal Ras/Myc-immortalized fibroblast cell line, EC-2 (35). EC-2 (Tace ⌬Zn/⌬Zn ) cells released about 20-fold less endogenous TGF-␣ into the medium as compared with corresponding Tace ϩ/ϩ cells (EC-4; Fig. 2B), confirming that TACE is required for TGF-␣ shedding by this cell type. Anti-FLAG recognized a prominent pro-TGF-␣ species of 25 kDa together with less abundant precursor proteins of 22 and 36 kDa in lysates of EC-2 cells transfected with TGF-␣ expression vector only (Fig. 2C). In contrast, anti-FLAG recognized a prominent 16-kDa protein in lysates of EC-2 cells transfected with both TGF-␣ and TACE expression vectors. Unlike the larger immunoreactive bands, the 16-kDa protein was not recognized by anti-HA (not shown); thus, it corresponds to the residual precursor tail resulting from cleavage of the ectodomain at the C-terminal site. Analysis of TGF-␣ species present in media supported this conclusion. In the absence of transfected TACE, two HA-reactive species of ϳ6 kDa were present, replaced by a more prominent band of intermediate size in the presence of TACE.
TACE Cleaves Pro-TGF-␣ in Vitro-To assess the ability of TACE to directly cleave pro-TGF-␣, lysates of RLEC stably expressing HA/FLAG-tagged pro-TGF-␣ (34) were incubated in the absence or presence of native TACE. Consistent with previous observations (34), anti-HA recognized prominent pro-TGF-␣ proteins of about 25 and 36 kDa in Western blots of reaction mixtures incubated in the absence of TACE (Fig. 3A,  left). In the presence of TACE, a prominent new species of about 20 kDa was detected with corresponding decreases in larger pro-TGF-␣ species, especially the 25-kDa form. This 20-kDa product was readily detected after a 15-min incubation with exogenous TACE but was present at higher levels after 60 min. Its appearance was inhibited in the presence of IC-3. Minor amounts of the 20-kDa product were also detected when reactions were performed for 60 min in the absence of TACE, presumably due to endogenous TACE in RLEC. 4,5 Importantly, the novel, TACE-dependent 20-kDa product 4  Primary keratinocytes were prepared from Tace ϩ/ϩ or Tace ⌬Zn/⌬Zn newborn littermate pups and cultured to 90% confluence prior to shifting to conditioning media. After 48 h at 37°C, media and cell lysate were harvested and analyzed for TGF-␣ content using a specific RIA (32). Where indicated, the hydroxamic acid inhibitor, IC-3, was included in the conditioning media at 50 M. Other Tace ⌬Zn/⌬Zn cultures were infected with control adenovirus (Ad) or adenovirus encoding TACE (Ad TACE ) at a multiplicity of infection of 4 -10.
Results are expressed as the ratio of TGF-␣ in the medium compared with lysate and are normalized relative to the wild-type ratio. Statistical significance: for wild-type (WT) versus ⌬Zn, p ϭ 0.002 by the Wilcoxon rank sums test; for IC-3 treatment, n ϭ 2; for virus reconstitution, p ϭ 0.029 by the Wilcoxon rank sums test. was also recognized by anti-FLAG (Fig. 3A), indicating that TACE cleaved only the N-terminal processing site of pro-TGF-␣ in vitro and not the C-terminal site. Although we occasionally observed modest increases in the levels of 16-kDa tail following incubation with TACE (Fig. 3A, 15-min time point), anti-HA did not detect mature TGF-␣ in these reactions, which is consistent with limited or absent cleavage of the C-terminal site.
To confirm accurate cleavage by TACE, pro-TGF-␣ was partially purified by anti-FLAG affinity chromatography and incubated with recombinant human TACE ectodomain (36). The N-terminal sequence of the 20-kDa protein identified after SDS-PAGE was determined to be VVSHY, confirming cleavage at the site corresponding to the N terminus of mature TGF-␣ (46). As part of these analyses, we also determined N-terminal sequences of endogenous TGF-␣ proteins to confirm their identity. The sequence of the 16-kDa species was VVAAS, confirming its identity as the precursor tail derived by accurate cleavage at the C-terminal site of pro-TGF-␣. The N-terminal sequences of the FLAG-reactive 25-and 36-kDa pro-TGF-␣ species were identical: LEXST (the expected N residue at ϩ3 was not identified, presumably due to glycosylation), confirming these as forms of intact pro-TGF-␣ lacking only the signal peptide.
The failure of TACE to cleave the C-terminal site of pro-

FIG. 2. Reintroduction of TACE into Tace⌬Zn/⌬Zn fibroblasts promotes processing of pro-TGF-␣.
A, structure of pro-growth factor constructs used in this study. Locations of HA, FLAG, or Myc epitope tags are indicated. TM, transmembrane domain; CYT, cytoplasmic domain. Arrows designate cleavage sites for the release of mature growth factors. B, TGF-␣ release by Myc/Ras-transformed Tace ⌬Zn/⌬Zn and Tace ϩ/ϩ fibroblasts was compared as described in the legend for Fig. 1 (n ϭ 2). WT, wild-type. C, cotransfection of TACE and TGF-␣. Tace ⌬Zn/⌬Zn fibroblasts were transfected with epitope-tagged pro-TGF-␣ in combination with pcDNA3 or pcDNA3 containing the murine TACE cDNA (35). After 24 h, cells were shifted to conditioning media, and 24 h later, lysates and media were harvested for Western blot analysis of TGF-␣ species using the indicated antibodies. Media were concentrated using Sep-Pak C-18 cartridges prior to analysis. Apparent molecular sizes of TGF-␣ species as deduced from known standards are indicated.
TGF-␣ could be due to interference by the hydrophobic transmembrane domain in vitro. To address this possibility, we created a soluble ectodomain construct ( Fig. 2A), pro-TGFecto. Pro-TGFecto was expressed in COS-1 cells, and conditioned media were harvested and incubated with recombinant human TACE ectodomain. In the absence of TACE, anti-FLAG and anti-HA both detected several 18 -25-kDa precursor forms (Fig.  3B). After a 60-min incubation with TACE, 6-and 8-kDa HAreactive products appeared with a concomitant decrease in pro-TGFecto species. The 8-kDa product was also FLAG-reactive, indicating that it resulted from N-terminal cleavage. After 4 h, pro-TGFecto was converted to a single product equal in size to fully processed 6-kDa TGF-␣ and containing the HA but not the FLAG epitope (Fig. 3B). The predicted 2.1-kDa C-terminal, FLAG-reactive product resulting from TACE cleavage at the C-terminal site was not detected on Western blots, presumably due to its small size.
To confirm that conversion to 6-kDa TGF-␣ resulted from cleavage of the C-terminal site in pro-TGFecto, we used mass spectrometry to characterize the C-terminal, FLAG-reactive product. MALDI/TOF-MS (Fig. 3C) revealed the singly protonated C-terminal product to be 2151 m/z, equal to a synthetic peptide matching the expected product. This confirms TACE cleavage at the physiological AV cleavage site (see Table I for sequence). A second product of 1811 m/z was also detected. MALDI-MS/MS 3 revealed that this product resulted from cleavage of the A-S peptide bond, 4 residues C-terminal to the AV site. TACE also cleaved the synthetic peptide at this site (not shown). However, since the N-terminal sequence (see above) of the endogenous cytoplasmic tail recovered from pro-TGF-␣-expressing RLEC was VVAAS, cleavage of the A-S site may be an in vitro artifact.
Peptide Cleavage Assays-Tace ⌬Zn/⌬Zn mice exhibit widespread epithelial defects reminiscent of those observed in EGFR Ϫ/Ϫ mice, suggesting that it might have a role in the shedding of multiple EGFR ligands (23,29). We therefore tested the ability of TACE to cleave peptides representing the published N-and C-terminal cleavage sites of multiple EGF family members (10, 12, 19, 23, 38 -42). Recombinant TACE cleaved peptides corresponding to the C-terminal sites of betacellulin, epiregulin, and HB-EGF as well as the N-terminal sites of AR and epiregulin (Table I). Notably, cleavage of all these peptides required 20-fold more TACE than cleavage of either TGF-␣ peptide. Interestingly, the C-terminal HB-EGF peptide was cleaved by TACE at two positions, both with nonpolar residues in the P 1 Ј position: a Pro-Val dipeptide corresponds to the site predicted from the sequence of mature HB-EGF released following PMA stimulation (42), whereas the Arg-Leu cleavage site was not reported previously. We did not observe cleavage of the Glu-Asn dipeptide shown previously to be a substrate for matrix metalloproteinase-3 (19). In contrast, TACE did not cleave peptides representing the C-terminal site for AR, the N-terminal site for betacellulin, or either of the mature EGF processing sites.
AR and HB-EGF Transfections-To extend the peptide assays, we tested the ability of transfected TACE to release the ectodomains of cotransfected epitope-tagged AR and HB-EGF in cell culture. For both, the HA tag was inserted in the mature growth factor domain, and a Myc epitope was inserted at the C terminus of the precursor (Fig. 2A). These cDNAs were cotrans- fected with TACE into Tace ⌬Zn/⌬Zn EC-2 cells, and the media and lysates were analyzed by Western blot. Anti-Myc detected multiple lysate bands for both AR and HB-EGF, consistent with complex patterns of AR and HB-EGF processing due to heterogeneous glycosylation and multiple cleavage sites in each case (11,12,17,47,48).
Cotransfection of TACE with pro-AR into EC-2 cells resulted in the selective loss of a 45-kDa Myc-reactive pro-AR from cell lysates (Fig. 4). Additionally, bands of ϳ30 -32 kDa, likely representing N-terminal processed forms (17), were slightly reduced in size. There was also a variable increase in the abundance of 20 -25-kDa species that were Myc-but not HAreactive (not shown), suggesting that they correspond to residual cleavage products (i.e. cytoplasmic tail). These alterations to lysate forms were accompanied by the appearance of anti-HA-reactive AR species of ϳ38 and 42 kDa in media in the presence of TACE. These sizes correspond to previously characterized soluble AR species (17,48).
Cotransfection of TACE and pro-HB-EGF in EC-2 cells reduced the intensity of cell-associated, higher molecular size, Myc-reactive HB-EGF species with variable increases in an ϳ25-kDa lysate band (Fig. 4). These changes were accompanied by significant increases in 22-, 24-, and 27-kDa HA-reactive media forms similar in size to previously characterized soluble HB-EGF (47). Additionally, a novel 41-kDa HA-reactive band appeared in the media of cells cotransfected with TACE; this may correspond to an extensively glycosylated form of the HB-EGF ectodomain.
We additionally tested whether treatment of transfected cells with 1 M PMA for 20 min before harvest would result in a further increase in the media levels of AR or HB-EGF. We observed a consistent increase in the media levels of ϳ38and 42-kDa AR species following PMA treatment but did not observe any increase in the media levels of HB-EGF species under the same conditions (data not shown). It may be that overexpressed TACE constitutively processes HB-EGF at maximal levels; alternatively, the EGF included in the conditioning media may stimulate processing to its highest possible level.
TACE Regulates EGFR Ligand Availability-We used a genetic approach to determine whether TACE regulates EGFR ligand availability in vivo. This approach was based on observations that mice deficient in EGFR or its ligands are frequently born with open eyes (26 -28, 49 -51), whereas Egfr wa-2/wa-2 mice bearing kinase-impaired EGFR only occasionally display this phenotype. To determine whether limiting the gene dosage of TACE exacerbates this phenotype, we bred the Tace ⌬Zn mutation to heterozygosity on the Egfr wa-2/wa-2 background (unlike Tace ⌬Zn/⌬Zn mice, Tace ϩ/⌬Zn animals are not born with open eyes). A similar approach has been used recently to demonstrate that Sos1 (52) and Shp2 (53) function in the same pathway as EGFR.
When Tace ϩ/⌬Zn mice were crossed with Egfr wa-2/wa-2 mice, only 3 of 25 homozygous Tace ϩ/ϩ Egfr wa-2/wa-2 mice were born with open eyes, and no doubly heterozygous Tace ϩ/⌬Zn Egfr ϩ/wa-2 mice displayed this phenotype (Fig. 5). In dramatic contrast, 11 of 24 mice with the Tace ϩ/⌬Zn Egfr wa-2/wa-2 genotype had open eyes at birth, a highly significant result (p ϭ 0.012). This genetic analysis supports a link between EGFR and TACE and is consistent with TACE functioning upstream to regulate the availability of multiple EGFR ligands.

DISCUSSION
Several lines of evidence unambiguously establish a role for TACE in TGF-␣ shedding. Although the Tace ⌬Zn/⌬Zn mutation is perinatal lethal, survivors are invariably born with open eyes and curly whiskers, and rare mice that survive beyond the first few days display misoriented hair follicles as well as wavy hair (23). These eye, skin, and hair phenotypes are the hallmark of TGF-␣-deficient mice (24,25). In addition, we showed previously that TACE cleaves peptide substrates corresponding to the two processing sites of pro-TGF-␣ in vitro and that transformed Tace ⌬Zn/⌬Zn fibroblasts are deficient in TGF-␣ shedding (23).
Here, we additionally show that the release of soluble TGF-␣ by primary epidermal keratinocytes derived from Tace ⌬Zn/⌬Zn mice is dramatically reduced as compared with shedding by corresponding TACE ϩ/ϩ cells. This deficiency is at least partly restored in both keratinocytes and fibroblasts upon introduction of active TACE. In addition, we demonstrate that native or recombinant TACE faithfully cleaves both the N-and C-terminal sites of a soluble form of pro-TGF-␣ in vitro. Cleavage at the N-terminal site by native TACE was observed with full-length pro-TGF-␣, whereas the C-terminal site was not cleaved in the intact precursor. We speculate that the absence of cleavage at this site in vitro is an artifact resulting from interference by the hydrophobic transmembrane domain of the precursor. To date, TACE has been implicated in the shedding of a large number of diverse cell surface molecules. However, apart from TNF-␣ (20,21), only pro-TGF-␣ has been shown to be directly cleaved by TACE at the correct sites. Collectively, these findings suggest that TACE functions directly as a major TGF-␣ convertase, although they do not exclude possible physiological roles for additional or alternative proteases. It is still possible that TACE functions in vivo as part of a cascade that regulates the processing of pro-TGF-␣.
Previous studies found that the N-terminal processing site of pro-TGF-␣ was cleaved rapidly in Chinese hamster ovary cells, whereas the C-terminal site was cleaved much less efficiently unless ectodomain shedding was induced by phorbol esters, calcium ionophores, etc. (13)(14)(15)45). It is interesting, therefore, that TACE cleaved the N-terminal site for the pro-TGFecto substrate more efficiently than the C-terminal site. Specifically, two N-terminal cleavage products differing only at the C terminus were observed with short incubations; these were resolved to a fully processed product upon further incubation. Significantly, we never observed a product representing initial  (36) at 37°C for 4 h and products analyzed by LC/MS as previously described (23). Results of cleavage of TGF-␣ peptides (23) are included for comparison. Observation of cleavage and the site are indicated. Approximate concentration of TACE required for reaction is indicated. Note that cleavage of TGF-␣ peptides required approximately 10ϫ more TACE than cleavage of a TNF peptide (23). cleavage at the C-terminal site. These results are consistent with TACE cleaving both sites albeit with different efficiency. Since the sequences surrounding the two cleavage sites (AA2VV and LA2VV) are highly similar, other pro-TGF-␣ sequence or structural motifs likely contribute to substrate recognition. Several lines of evidence also indicate a role for TACE in processing additional EGF family members. These include in vitro peptide cleavage assays, cotransfection experiments, and in vivo genetic evidence. Peptides corresponding to the processing sites of several EGF family members, including AR and HB-EGF, were cleaved by recombinant TACE. Despite the failure of TACE to cleave the C-terminal peptide of AR in vitro, restoration of TACE expression in Tace ⌬Zn/⌬Zn cells resulted in a significant increase in the release of both AR and HB-EGF into the media. A role for TACE in pro-HB-EGF processing is further supported by the recent observation that Tace ⌬Zn/⌬Zn cells are deficient in the shedding of transfected human HB-EGF (55).
ADAM9 (MDC9/meltrin-␥), ADAM10, and ADAM12 have also recently been implicated in HB-EGF shedding (56 -58), suggesting that distinct enzymes mediate the shedding of a single substrate. In fact, recent studies, including one based on inhibitor data and membrane fractionation, 5 have pointed to the existence of one or more additional pro-TGF-␣ convertase activities (55). This may explain the residual, metalloproteinasedependent TGF-␣ release we observed from Tace ⌬Zn/⌬Zn primary keratinocytes and the limited TACE-independent processing of pro-TGF-␣ observed in transfection experiments. It seems likely that different proteases catalyze pro-growth factor processing in different cell types and in response to distinct stimuli, as has been suggested for other cell surface targets (59).
Significantly, phenotypic similarities between Tace ⌬Zn/⌬Zn mice and EGFR ligand knockouts suggest that soluble forms of the EGF family members are critical for normal development despite the apparent bioactivity of their membrane-anchored precursors (54, 60 -63). Consistent with a role for TACE in regulating EGFR ligand availability, we observed a dramatic increase in the frequency of the open-eye phenotype of Egfr wa-2/wa-2 mice with a reduction in TACE (Tace ϩ/⌬Zn ) gene dosage. This indicates that the TACE activity derived from a single wild-type allele did not mediate sufficient release of soluble ligand to overcome the impaired kinase activity of the waved-2 receptor. (We previously observed waved-2 receptor activation to wild-type levels in the presence of sufficient ligand (49).) This raises the intriguing possibility that TACE might be an important therapeutic target for treatment of hyperproliferative diseases, including cancer, that are characterized by excessive production of EGFR ligands. The therapeutic potential of TACE, as well as its demonstrated role in normal development, underscores the necessity of a thorough understanding of its role in the shedding of this important family of bioregulatory molecules.
Acknowledgments-We thank Nolan Yeung and Aileen Chang for technical assistance. We are also grateful for assistance from Christoph Borchers, Carol Parker, and the UNC Proteomics Core Facility with mass spectrometry and for advice from Marcia Moss on in vitro TACE experiments. We also thank Leslie Jackson and Kelly Troyer for helpful comments on the manuscript.

FIG. 4. Reintroduction of TACE results in increased AR and HB-EGF shedding.
Tace ⌬Zn/⌬Zn fibroblasts were transfected with epitopetagged pro-AR or pro-HB-EGF and either pcDNA3 or TACE as shown. Media were conditioned for 24 h, and lysates and media were analyzed as described in the legend for Fig. 2 except that HB-EGF-conditioned media were directly immunoprecipitated with anti-HA prior to Western blot. Indicated molecular sizes are derived from the migration of known standards. Asterisks indicate HB-EGF forms that increased in the media when TACE was present.

FIG. 5. TACE deficiency exacerbates the open-eye phenotype of
Egfr wa-2/wa-2 mice. Intercrossing of Tace ϩ/⌬Zn females and Efgr wa-2/wa-2 males produced pups of the indicated genotypes that were scored for open eye at birth. Note that the open-eye phenotype was restricted to Egfr wa-2/wa-2 animals and was significantly (p ϭ 0.012) more frequent in TACE heterozygotes.