Evidence for a Role of a Tumor Necrosis Factor-α (TNF-α)-converting Enzyme-like Protease in Shedding of TRANCE, a TNF Family Member Involved in Osteoclastogenesis and Dendritic Cell Survival*

Tumor necrosis factor (TNF)-related activation-induced cytokine (TRANCE), a member of the TNF family, is a dendritic cell survival factor and is essential for osteoclastogenesis and osteoclast activation. In this report we demonstrate (i) that TRANCE, like TNF-α, is made as a membrane-anchored precursor, which is released from the plasma membrane by a metalloprotease; (ii) that soluble TRANCE has potent dendritic cell survival and osteoclastogenic activity; (iii) that the metalloprotease-disintegrin TNF-α convertase (TACE) can cleave immunoprecipitated TRANCE in vitro in a fashion that mimics the cleavage observed in tissue culture cells; and (iv) that in vitro cleavage of a TRANCE ectodomain/CD8 fusion protein and of a peptide corresponding to the TRANCE cleavage site by TACE occurs at the same site that is used when TRANCE is shed from cells into the supernatant. We propose that the TRANCE ectodomain is released from cells by TACE or a related metalloprotease-disintegrin, and that this release is an important component of the function of TRANCE in bone and immune homeostasis.

Tumor necrosis factor (TNF) 1 -related activation-induced cytokine (TRANCE), a recently identified member of the TNF family, is a dendritic cell survival factor that also has a role in bone homeostasis (1)(2)(3)(4). Like TNF-␣, TRANCE is a type II integral membrane glycoprotein of ϳ45 kDa with a long extracellular stalk region followed by a receptor-binding core domain (5). TRANCE expression in osteoblasts and stromal cells can be induced with vitamin D3, prostaglandin E 2 , interleukin-1, or glucocorticoids (4). In turn, TRANCE is known to induce differentiation and activation of osteoclasts. This suggests that TRANCE provides an important link between the action of hormones and physiological cytokines and bone resorption. TRANCE is also expressed on activated T cells (5), where it induces dendritic cell survival, thereby enhancing T cell priming (1,2). Therefore TRANCE may also regulate antigen presentation during an immune response. TRANCE mediates its effects through the membrane-anchored TRANCE receptor (TRANCE-R, also referred to as RANK (receptor activator of nuclear factor-B)), which results in activation of c-Jun Nterminal kinase and nuclear factor-B (2,5). Finally TRANCE is known to bind to a soluble receptor, termed osteoprotegerin or osteoclast inhibitory factor (OPG/OCIF), which is a member of the TNF-␣ receptor family (6). OPG/OCIF presumably functions as a decoy receptor, since systemic overexpression or injection of OPG/OCIF causes osteopetrosis in mice (7), whereas OPG/OCIF deficiency results in osteoporosis (8).
Similar to TNF-␣, which is thought to be released from the plasma membrane by the metalloprotease-disintegrin TNF-␣ convertase (TACE) (9 -11), TRANCE may also be shed by TACE or a related metalloprotease. Metalloproteases have been implicated in the shedding or release of several different cell surface proteins from the plasma membrane. These proteins include various cytokines, cytokine receptors, adhesion proteins, and other proteins such as the ␤-amyloid precursor protein (12). Shed membrane proteins may have very different properties compared with the membrane-anchored forms. Sol-* This work was supported in part by a grant from Glaxo-Wellcome (to C. P. B.), by National Institutes of Health NIAID Grant AI44264 (to Y. C.), and by National Science Foundation Grant DBI-942013 (to P. T.). 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 1 The abbreviations used are: TNF, tumor necrosis factor; TRANCE, TNF-related activation-induced cytokine; TRANCE-R, TRANCE receptor; TACE, TNF-␣ convertase; OPG/OCIF, osteoprotegerin/osteoclast inhibitory factor; ␤APP, ␤-amyloid precursor protein; PBS, phosphatebuffered saline; GST, glutathione S-transferase; mAb, monoclonal antibody; DC, dendritic cell; PAGE, polyacrylamide gel electrophoresis; STI, soybean trypsin inhibitor; TPCK, N-tosyl-L-phenylalanine chloromethyl ketone; TPA, 12-O-tetradecanoylphorbol-13-acetate. ubilized Fas-ligand (FasL), for example, has been shown to be a much weaker inducer of apoptosis than its transmembrane form, suggesting that shedding down-regulates the activity of FasL (13). In contrast, soluble TNF-␣ is a potent pro-inflammatory cytokine, and release of TNF-␣ into the circulatory system contributes to its systemic effects and to pathologic conditions such as septic shock (14). In this study, we provide evidence that TACE or a related metalloprotease-disintegrin is a likely candidate for the proteolytic release of the TRANCE ectodomain. Furthermore, we show that soluble TRANCE promotes dendritic cell survival and osteoclast differentiation in tissue culture.

MATERIALS AND METHODS
cDNA Constructs and Reagents-A FLAG-tagged full-length murine TRANCE expression vector (pFLAG-TRANCE) has been described (5). hCD8-TRANCE was expressed in baculovirus and purified on an ␣-hCD8-Sepharose column as described (1). Mouse TRANCE-R fused to human IgG1 (TR-Fc) was cloned into the vector PVL1392 and expressed in baculovirus. Purification was performed by binding to a protein A-Sepharose column, eluting with glycine (0.2 M, pH 2.9), and dialyzing against PBS. M2 anti-FLAG mAb was purchased from Sigma. A cDNA fragment encoding for the human TACE cytotail (corresponding to amino acids 695-824) was cloned in frame to the coding region of GST in the pGEX-4T-1 vector (Amersham Pharmacia Biotech). The GST-TACE cytotail fusion protein was expressed and purified from BL21 bacteria and used as an immunogen to raise rabbit polyclonal antisera as described previously (15).
Transient Transfections-COS-7 cells were transiently transfected with pcDNA3 vector or pFLAG-TRANCE vector with LipofectAMINE (Life Technologies, Inc.) following the manufacturer's suggestions. Human 293T cells were transiently transfected with similar constructs using a standard calcium phosphate method.
Pulse-Chase Analysis-COS-7 cells transiently transfected with either pcDNA3 vector or pFLAG-TRANCE were subjected to pulse-chase with 200 Ci of 35 S-labeled methionine and cysteine (EXPRESS, NEN Life Sciences) as described previously (16). After chasing in Opti-MEM (Life Technologies, Inc.) for different amounts of time, supernatant and lysate samples were spun at 13,000 rpm in a Sorvall tabletop centrifuge for 15 min. Where indicated, the hydroxamate-based metalloproteaseinhibitor batimastat (BB-94) (17), the serine protease inhibitors leupeptin (Sigma), N-tosyl-L-phenylalanine chloromethyl ketone (TPCK, Sigma), soybean trypsin inhibitor (STI, Sigma), 4-(2-aminoethyl)benzenesulfonylfluoride hydrochloride (Pefabloc SC, Roche Molecular Biochemicals), the cysteine protease inhibitor L-trans-epoxysuccinylleucylamide-(4-guanidino)butane (E-64, Sigma), or the aspartate protease inhibitor pepstatin (Sigma) were included in the chase medium in the presence or absence of TPA (50 ng/ml). KMLS-8.3.5.1 cells were labeled overnight with 50 Ci of EXPRESS label and then stimulated for 3 h in the presence of ionomycin (500 ng/ml) and TPA (50 ng/ml). The cleared cell lysates were immunoprecipitated with the M2 anti-Flag mAb or TR-Fc (as indicated), and protein G beads, and the cleared supernatants were immunoprecipitated with the TR-Fc and protein A beads. The immunoprecipitated material was recovered by boiling in sample loading buffer, and separated by SDS-PAGE. Gels were fixed in 50% methanol, 10% acetic acid and incubated in 1 M salicylic acid for 15 min prior to drying and exposure to autoradiography film (Kodak XAR).
Western Blot Analysis-Approximately 500 g of cleared lysates from COS-7, 293T, and THP-1 cells were incubated with concanavalin A-Sepharose (Amersham Pharmacia Biotech). Bound glycoproteins were eluted with sample loading buffer and boiling at 95°C for 5 min. Western analysis was performed following SDS-PAGE as described previously (18).
In Vitro Cleavage of Full-length TRANCE, hCD8-TRANCE, and Nterminal Sequence Analysis-To generate full-length TRANCE, COS-7 cells transiently transfected with FLAG-TRANCE were pulse-labeled and then chased for 3 h as described above. FLAG-TRANCE was immunoprecipitated and washed three times in lysis buffer without protease inhibitor, followed by one wash in PBS. After incubation overnight at 37°C with recombinant TACE (10) (2.5 g/ml) or MMP-1 (19) (0.1 g/ml), the samples were eluted from the beads by boiling in sample loading buffer and subjected to SDS-PAGE. Proteolysis of hCD8-TRANCE was accomplished by incubating 3 g of purified hCD8-TRANCE with addition of 1 g of recombinant TACE in 30 l of PBS for 5 h at 37°C. The TRANCE ectodomain, which is shed into the culture supernatant of 293T cells transfected with FLAG-TRANCE, was precipitated with TR-Fc and protein A. For N-terminal sequence analysis, the samples were transferred to polyvinylidene difluoride membranes (ProBlott, Applied Biosystems) after electrophoresis, stained with Coomassie Blue R-250, destained in 50% methanol, 10% acetic acid, and rinsed with double-distilled H 2 O. The bands of interest were excised, and the N-terminal amino acid residues were analyzed by automated Edman degradation, using an Applied Biosystems 477A sequenator, with instrument and procedure optimized for femtomole level analysis as described (20).
Incubation of Substrate Peptides with TACE and MMP-1 and Determination of Cleavage Sites and Kinetics-Synpep (Dublin, CA) synthesized peptides corresponding to the 12 amino acid residues surrounding the reported cleavage sites for TNF-␣ and TRANCE. The peptide sequences and the reported cleavage sites are presented in Table I. TACE (600 nM) or the catalytic domain of MMP-1 (30 nM) was incubated with 50, 25, or 12.5 M TRANCE peptide substrate in 10 mM HEPES, pH 7.2 containing 0.05% bovine serum albumin (Sigma). For the TRANCE and TNF-␣ peptide substrates, reactions were timed to allow approximately 5-20% turnover of the substrate. Reactions were quenched using 1% heptafluorobutyric acid, and products were separated by high performance liquid chromatography reverse phase chromatography (C18 column, Vydac, Hesperia, CA) with absorbance monitored at 350 nm. Turnover was quantitated by integrating peak areas of the substrate and product. Liquid chromatography-mass spectroscopy was used to determine the masses of the product and therefore the cleavage site recognized by either TACE or MMP-1. Briefly, digestion mixtures were passed over a Hypersil C18 column and, after UV detection at 350 nm, the sample was routed into the ion spray source of a Sciex API-III triple quadrupole mass spectrometer. Specificity constants were calculated from initial velocities using the equation: k cat /K m ϭ (% turnover/100)/ ([E] ϫ time). Conditions of k cat /K m were verified by running the reactions at more than one substrate concentration. Enzyme concentrations were determined utilizing a potent hydroxamate-type inhibitor of TACE by the methods described in Ref. 21.
Dendritic Cell Survival and Osteoclastogenesis Assay-Mature murine bone-marrow derived dendritic cells (DCs) were isolated as described previously (22) and incubated with either medium alone, hCD8-TRANCE (1 g/ml), or supernatants (1:50 dilution) from 293T cells transfected with pFLAG-2 vector (Kodak) or pFLAG-TRANCE in 96well plates. TR-Fc (10 g/ml) was simultaneously added to certain samples as indicated. 293T cell supernatants were centrifuged (100,000 ϫ g, 1 h) and filtered through a 0.2-m cutoff filter membrane to remove cell debris and membranes. Amounts of released TRANCE into culture supernatants were estimated to be approximately 3 g/ml. This was determined by complete depletion of ecto-TRANCE with TR-Fc, and subsequently comparing amounts of precipitated ecto-TRANCE with bovine serum albumin standards on SDS-PAGE after Coomassie staining (data not shown). Cell viability was assessed by trypan blue staining 48 -72 h after treatment as described (1,5). Osteoclastogenesis assays were performed as described (7).

RESULTS AND DISCUSSION
To determine if TRANCE, like TNF-␣, is shed from the plasma membrane by a protease, pulse-chase analysis was performed in COS-7 cells transiently expressing full-length TRANCE with a cytoplasmic FLAG epitope tag (FLAG-TRANCE). Immediately after the pulse labeling, two closely co-migrating bands of 46 and 48 kDa could be immunoprecipitated from the extracts with an anti-FLAG mAb (Fig. 1, lane 2). After deglycosylation of an identical sample with peptide:Nglycosidase F, only a 35-kDa band was detected, corresponding to the predicted molecular weight of membrane-anchored TRANCE (data not shown). As the duration of the chases increased from 3 to 12 h (Fig. 1A, lanes 3-5), the relative amount of 46-and 48-kDa proteins immunoprecipitated with anti-FLAG mAb decreased. Simultaneously, a 26-kDa band, which could be precipitated from the culture supernatant using a soluble TRANCE-receptor-Fc fusion protein (TR-Fc), increased in intensity (Fig. 1A, lanes 7-9). N-terminal sequence analysis (see below) confirmed that the 26-kDa protein is a soluble form of TRANCE (referred to as ecto-TRANCE hereafter). After a 12-h chase period, an additional band of 24 kDa, which most likely represents a minor shed TRANCE product, could also be seen in the supernatant (Fig. 1A, lane 9). These observations indicate that TRANCE, like TNF-␣, can be released from the cell surface.
To confirm that TRANCE shedding also occurs in non-transfected cells, a similar experiment was performed using the T cell hybridoma KMLS-8.3.5.1 (5) from which TRANCE was originally cloned. Hybridoma cells were labeled overnight with [ 35 S]methionine/cysteine and then stimulated for 3 h with ionomycin and phorbol 12-myristate 13-acetate to induce shedding. Immunoprecipitation with TR-Fc revealed TRANCE proteins in the media (24 and 26 kDa; Fig. 1B, lane 4) and lysate (24,26,46, and 48 kDa; Fig. 1B, lane 2). The observed membraneanchored and soluble forms of TRANCE were thus similar in molecular weight to those seen in COS-7 cells expressing TRANCE. The 24-and 26-kDa proteins in the cell lysate may represent cleaved ectodomains that reside in the intracellular secretory pathway, or that are still bound to uncleaved TRANCE molecules on the cell surface.
To address whether ecto-TRANCE is indeed functional, its activity was tested in osteoclast differentiation and in DC survival assays. Bone marrow precursors cultured in the presence of macrophage colony-stimulating factor or mature bone-marrow derived DCs were incubated with 293T supernatants containing ecto-TRANCE (see Fig. 3C) at an estimated final concentration of 60 ng/ml (see "Materials and Methods"). Fig. 2A demonstrates that bone marrow progenitors incubated with supernatants containing ecto-TRANCE, but not supernatants from cells transfected with vector alone, induced high levels of tartrate-resistant acid phosphatase activity, indicating that ecto-TRANCE mediates the differentiation of osteoclasts from precursors. This activity was dependent on a TRANCE/ TRANCE-R interactions because its effect could be blocked with saturating doses of a soluble receptor, TR-Fc. Fig. 2B shows that ecto-TRANCE-containing supernatants enhanced dendritic cell survival and this effect could also be inhibited by the addition of TR-Fc, indicating that the shed form of TRANCE is functionally active as a survival factor for mature DCs.
Since shedding of many cell surface proteins can be stimulated with phorbol esters and inhibited with metalloprotease inhibitors (12), we tested how these factors affect TRANCE shedding in transiently transfected COS-7 cells. In a pulsechase experiment, a 15-min (data not shown) or 6-h treatment with the phorbol ester TPA increased the amount of ecto-TRANCE released into the supernatant (Fig. 3A, lower panel, lane 1) compared with constitutive shedding for 15 min (data not shown) or 6 h (Fig. 3A, upper panel, lane 1). Treatment of TPA stimulated cells with BB-94 (17), a hydroxamic acid-based metalloprotease inhibitor, strongly decreased TPA-stimulated shedding of TRANCE (Fig. 3A, lower panel, lane 2), but did not detectably affect constitutive shedding (Fig. 3A, upper panel,  lane 2). Serine protease inhibitors (leupeptin, STI, Pefabloc SC, and TPCK; Fig. 3A, upper and lower panels, lanes 3-6), a cysteine protease inhibitor (E-64, 10 M; data not shown), and an aspartate protease inhibitor (pepstatin, 10 M; data not shown) had no apparent effect on the TPA-dependent or constitutive shedding of TRANCE from COS-7 cells.
Titrating the amount of BB-94 in the cell-based assay revealed significant inhibition of TRANCE shedding at 10 nM BB-94, and maximal inhibition at 100 nM (Fig. 3B). This is consistent with the K i of 11 nM that has been determined for the inhibition of TACE by BB-94 in vitro (21). We note that a  5 and 7-9). At the indicated time points, full-length TRANCE was immunoprecipitated from cell lysates with the anti-FLAG mAb (lanes 1-5), and soluble TRANCE was precipitated from supernatants with TR- Fc (lanes 6 -9). B, T cell hybridoma cells (KMLS-8.3.5.1) were stimulated for 3 h with 500 ng/ml ionomycin and 50 ng/ml phorbol 12-myristate 13-acetate following labeling overnight with [ 35 S]methionine/cysteine. Supernatant (lanes 3 and 4) and lysate samples (lanes 1 and 2) were collected and incubated with TR-Fc followed by protein A (lanes 2 and 4), or protein A beads alone (lanes 1  and 3). All samples were reduced prior to electrophoresis. residual component TRANCE shedding is not inhibited even by high doses of BB-94 in the TPA-treated sample. Since BB-94 also does not inhibit TRANCE shedding in unstimulated cells, other proteases besides metalloproteases may play a role in the constitutive release of TRANCE into the supernatant. A similar observation has been reported for the ␤-amyloid precursor protein (␤APP) (23). In fibroblasts lacking TACE, the phorbol 12-myristate 13-acetate-dependent shedding of ␤APP is abolished, while a low level of constitutive shedding of ␤APP, which is not inhibited by the hydroxamate-based metalloprotease inhibitor TAPI-2, is still present. Only a small percentage of total TRANCE was released, as levels of TRANCE in the cell lysate did not decrease with TPA stimulation compared with untreated cells (Fig. 3A, lane 1) or increase in BB-94-treated cells. Taken together, these results provide the first evidence that TRANCE, like TNF-␣, ␤APP, and other shed proteins can be released in response to phorbol esters (12,24,25), and that this release can be inhibited by a hydroxamate-based metalloprotease inhibitor.
To test for a potential role of TACE in the shedding of TRANCE, metabolically labeled full-length TRANCE was immunoprecipitated and incubated in vitro with recombinant TACE. This treatment yielded polypeptides of 23 and 26 kDa (Fig. 3C, lane 3) that were not visible in the untreated sample (Fig. 3C, lane 2). A likely explanation for the relatively inefficient processing of immunoprecipitated TRANCE in Nonidet P-40 by soluble TACE is that TACE and its substrate may both need to be membrane-anchored for optimal cleavage to occur.
The 26-kDa band generated in vitro by TACE co-migrated with the ecto-TRANCE isolated from the supernatant of transfected COS-7 cells (Fig. 3C, lane 1), whereas the 23-kDa product did not. As a control for specificity, immunoprecipitated full-length TRANCE was also incubated with MMP-1, a member of the matrix metalloprotease family (26). MMP-1 produced two polypeptides of 25.5 and 24.5 kDa (Fig. 2B, lane 6). The 25.5-kDa band generated with MMP-1 did not co-migrate with ecto-TRANCE in the supernatant of TRANCE-expressing COS-7 cells (data not shown), or with TRANCE polypeptides resulting from TACE cleavage (Fig. 2B, lane 5). These in vitro cleavage results are consistent with the idea that TACE, or a protease with a similar substrate specificity, may be involved in cleaving TRANCE. Western blot analysis confirmed the presence of TACE in COS-7 cells (Fig. 3D, lane 2) and 293T cells (Fig. 3D,  lane 6).
To compare the in vitro cleavage site for TACE in TRANCE to the N terminus of ecto-TRANCE in the supernatant of 293Ttransfected cells, a soluble fusion protein of the TRANCE ectodomain with human CD8 (hCD8-TRANCE, see Fig. 4C and "Materials and Methods") was incubated in vitro with TACE. This treatment resulted in at least three visible cleavage products (Fig. 4A, lane 2) that were not present in the TACE sample (Fig. 4A, lane 3) or in the hCD8-TRANCE sample (Fig. 4A, lane 1) incubated separately. N-terminal sequence analysis of the 26-kDa band, which co-migrated with ecto-TRANCE, revealed that it contained two polypeptide species. One of the two sequences corresponded to the N terminus of ecto-TRANCE (see  (Fig. 4A, lane 5) confirmed the previously reported cleavage site of TRANCE, and was identical to the cleavage site generated by TACE (Fig. 4B and Ref. 3). These results demonstrate that a major in vitro cleavage site for TACE in the ectodomain of TRANCE is identical to the cleavage site of the protease that releases ecto-TRANCE from 293T cells (see diagram in Fig. 4C).
To further evaluate the cleavage of TRANCE, we determined the cleavage site and kinetics for processing of TRANCE and TNF-␣ peptides by TACE. The TRANCE peptide was cleaved in the correct position by TACE, but the specificity constant was 1000-fold lower than for the TNF-␣ peptide (Table I). MMP-1 did not cleave the TRANCE peptide (data not shown). This is consistent with the observation that TACE cleavage of immunoprecipitated pro-TRANCE produced a protein that co-migrated with TRANCE shed from cells, whereas cleavage with MMP-1 did not (see Fig. 2C). With respect to the peptide cleavage specificity of metalloprotease-disintegrins, we note that both ADAM10 (KUZ/MADM) and TACE cleave a TNF-␣ peptide at the physiological position (9,10,27), while MDC9 has a clearly distinct specificity compared with TACE (21). The correct cleavage of the TRANCE peptide by TACE therefore suggests that TACE, or a metalloprotease with a similar substrate specificity, cleaves TRANCE in cells. The difference in specificity constants of TACE for the TRANCE and TNF-␣ peptides (1000-fold) is similar to the reported difference in peptide cleavage efficiency of TACE for TNF-␣ and the putative TACE substrate L-selectin (2250-fold) (11). A general question raised by these observations is whether TACE substrate recognition in cells involves additional targeting events between the protease and the substrate (11,21,28,29), or alternatively if TRANCE is actually cleaved by another related metalloprotease with a higher specificity constant for the TRANCE cleavage site.
The results presented here demonstrate that a phorbol esterinducible metalloprotease can release ecto-TRANCE from the plasma membrane. The finding that incubation of the hCD8-TRANCE ectodomain or of a TRANCE peptide with recombinant TACE generates a fragment with the identical N terminus as ecto-TRANCE suggests that TACE, or a related metalloprotease, mediates TRANCE shedding. Soluble TRANCE has potent functional activity in promoting dendritic cell survival and osteoclast differentiation. In analogy to the regulation of TNF-␣ function by shedding from the plasma membrane (9, 10), our results suggest that TRANCE shedding may be an important aspect of the functional regulation of this protein.
We propose that released TRANCE may mediate signaling as a soluble cytokine, and that the release of TRANCE may be an important factor in the TRANCE/TRANCE-R/osteoprotegerin signaling axis.